Methodology for Assessing the Accessibility of Digital Resources

  • Version: 0.3.8
  • Date: 17 October 2026
Sampling, interface checks, recorded observations and reassessment form an evaluation process.

Download the document (DOCX)

Status: draft for discussion and trial; before approval, the methodology must be trialled on live resources with the participation of users. This document is not an approved standard, a normative legal act, a certificate, or an automatic basis for a statement of conformance. The section on websites and web applications has passed internal trial; for electronic documents, mobile applications, desktop software, digital services and self-service terminals the protocols have been issued, but the trial is not complete (section 10 "Section Readiness Table").

Part I. General part

The general part applies to any digital resource regardless of the readiness of the industry-specific section of Part II dedicated to that resource (section 10).

1. Purpose and principle

The methodology is intended for reproducible assessment of digital resources - websites and web applications, electronic documents, mobile applications, desktop software, digital services and processes, self-service terminals - taking into account the needs of people with impairments of vision, hearing, movement, speech, attention, memory, comprehension and learning. Each resource type has its own assessment procedure, which is set by the corresponding section of Part II; the general part below defines principles common to all resource types. The scope of application of the methodology is described in section 1.1.

The assessment answers four different questions:

  1. which criteria of the applicable standard are met on the tested units and states;
  2. whether different users can complete the key processes;
  3. which barriers go beyond the minimum criteria of the standard, especially in cognitive accessibility;
  4. to what extent the evidence allows the conclusion to be generalised beyond the sample.

Conformance with criteria is not the same as practical accessibility for every individual. WCAG 2.2 is the primary standard for websites and web applications (section A) and states directly that the recommendations do not cover all needs, especially cognitive, language and learning ones; this limitation also applies to other bases used by the sections of Part II. User success also does not prove formal conformance to all criteria. Therefore automated, manual and user assessments are carried out separately and are compared only in the conclusion.

A single accessibility percentage is not used for any resource type. A scanner does not issue a certificate. A numerical score is capable of hiding a blocking defect among many successful checks. WCAG-EM 2.0 allows an aggregate indicator only as a supplement and warns that there is no single sufficiently reliable metric, and WCAG does not provide a rating scheme.

Assessment of organisational maturity is not part of the result of checking a specific digital resource. Policy, allocation of responsibility, training, procurement and monitoring can be assessed separately using the W3C Accessibility Maturity Model with Group Note status, and can be planned with the draft Accessibility Roles and Responsibilities Mapping resource. High maturity is not proof that a product passes the applicable requirements, and a successful product check is not an assessment of the organisation's maturity.

Draft proposal. For monitoring a portfolio of resources of a single owner or authority (section 8), a separate regulation for automated preliminary screening and sequencing of checks is permitted, by analogy with the architecture of simplified control of Uutilsynet (Norway); such a regulation does not replace the conclusion for a specific resource and is not published as its assessment.

The legal basis and the scope of obligation are determined separately for a specific resource, the category of its owner, the applicable jurisdiction and the date of the assessment. National standards may also be adopted voluntarily or through a contract outside their country of origin. The applicable legal, procurement and contractual instruments must be recorded in the scope passport. The Russian profile is included as one national example in section 2.1.

1.1. Scope of application

The object of the methodology is digital resources: websites and web applications, electronic documents, mobile applications, desktop software, digital services and processes with authentication, self-service terminals. The general part (Part I) sets unified assessment principles for any resource type; the unit of assessment, the sampling procedure, configurations and the checklist for a specific resource type are determined by the corresponding section of Part II. As of the date of this version, only section A "Websites and web applications" is ready and has passed internal trial; for section B "Electronic documents", V "Mobile applications", G "Desktop software", D "Digital services and processes with authentication" and E "Self-service terminals and kiosks" appendices with check protocols have been issued, but trial of the sections is not complete. The full readiness table is in section 10.

The methodology is applied by assessors, procurement and acceptance staff, independent auditors, and those responsible for regular monitoring of the resource portfolio of state authorities, local government bodies, organisations subordinate to them, and other resource owners who apply the methodology voluntarily. It is applicable to preliminary diagnostics, release acceptance, independent audit, procurement checks and regular monitoring.

The specific object of the check is the unit of assessment of the resource type indicated in the scope passport (section 3) and determined by the applicable section of Part II. If one product is represented by several resource types (for example, a website, a mobile application, a desktop version and the documents it produces), each resource type is assessed separately according to its own section of Part II, and the assessment result for one type is not automatically carried over to another.

1.2. Terms and definitions

The definitions below - unit of assessment, sampling unit, structured sample, random sample, complete process, state, configuration - are formulated for websites and web applications and are explained in section A; the other sections of Part II use or adapt these concepts as applicable to their resource type.

  • Digital resource - a generic designation of the object of the methodology: a website or web application, an electronic document, a mobile application, desktop software, a digital service or process with authentication, a self-service terminal; the list of sections by resource type and their readiness is in section 10.
  • Unit of assessment - a "resource + representation/state" pair: a single URL with a public and an authenticated role, an open dialog, an error or a mobile layout may yield several units (section A.1).
  • Sampling unit - a unit of assessment included in a structured sample, a random sample or the composition of a complete process according to the rules of sections A.1-A.5; a specific check result in the results register is linked to a sampling unit (section 9.5).
  • Structured sample - a sample including instances of all identified common representations, essential functions, types, technologies and other relevant pages (section A.2).
  • Random sample - a number of randomly selected unique pages, representations or states added on top of the structured sample (section A.3).
  • Complete process - a starting point, all pages and states of the main sequence, as well as frequently used and critical branches (section A.4).
  • State - a specific condition of a unit of assessment: a role, an error, an empty result, a modal window, a notification or another dynamic state differing from a plain URL (section A.1, section 5).
  • Configuration - a recorded set of test run parameters: product version/build, OS, browser, assistive technology, interface language, viewport, zoom, input method and system settings (section A.6).
  • Defect severity - a characterisation of harm in a specific product, assigned independently of the WCAG level A, AA or AAA (section 6).
  • Blocking defect (S1) - a defect due to which the target group cannot complete a critical process, no accessible workaround exists, or there is a safety risk or irreversible loss (section 6).
  • Draft proposal - a separate label for an authorial, non-normative element of the methodology; such an element is not part of the mandatory WCAG requirements unless the customer has made it a contractual condition (section 2).
  • Legally applicable profile - a profile of requirements with a verified basis, edition and date, mandatory for the resource by operation of law or contract (section 3).
  • Voluntary target profile - a profile adopted as a benchmark beyond the legally applicable one, for example WCAG 2.2 AA and the COGA proposals (section 3).
  • Accessibility statement (declaration) - a published accessibility statement for a website, included in the inventory list and checked separately in the conclusion if the website is subject to a duty to publish it (section A.1, section 7).
  • WCAG (Web Content Accessibility Guidelines) - the web content accessibility guidelines developed and maintained by the W3C consortium; this methodology uses version WCAG 2.2.
  • W3C (World Wide Web Consortium) - the international consortium that develops standards and recommendations for the internet, including WCAG.
  • WAI (Web Accessibility Initiative) - an initiative (programme) of the W3C consortium that leads the development of WCAG, WAI-ARIA and related accessibility documents.
  • WCAG-EM (WCAG Evaluation Methodology) - the W3C consortium methodology for a step-by-step assessment of a website's conformance to WCAG requirements; published as a Note document, not as a formal W3C Recommendation.
  • COGA (Cognitive and Learning Disabilities Accessibility Task Force) - the W3C consortium task force on cognitive accessibility; it issues advisory, non-mandatory supplements to WCAG for people with cognitive and learning impairments.
  • ACT Rules (Accessibility Conformance Testing Rules Format) - the W3C consortium format for describing rules of automatic and manual accessibility testing against unified, reproducible criteria.
  • Levels A, AA, AAA - the three steps of WCAG conformance in increasing order of requirement strictness; the methodology targets level AA, for which all applicable criteria of levels A and AA must be met.
  • Assistive technologies (AT) - programmes and devices used by a person with disabilities to work with the resource: screen readers, screen magnifiers, alternative input devices and similar means.
  • ISO - the International Organization for Standardization, which publishes international standards, including in the field of accessibility. An ISO standard may be used voluntarily or incorporated through legislation, procurement or contract in any jurisdiction. Its legal or contractual effect depends on the applicable instrument. In the Russian profile, adoption as a GOST R or reference in a contract does not by itself make every provision mandatory; the applicable legal text or contractual condition must be checked separately.
  • GOST R ISO - the designation of a Russian national standard adopted on the basis of an international ISO (or ISO/IEC) standard, with a degree of correspondence indicated - identical (IDT), modified (MOD) or non-equivalent (NEQ); an identical standard (IDT) means that the text of the GOST R fully matches the international one.
  • PDF/UA (ISO 14289-1) - the international standard for accessible tagged PDF documents. National implementations and related standards must be identified in the applicable profile; the Russian mapping includes GOST R 70176-2022 (section B).
  • PDF/A (ISO 19005) - a family of international PDF standards for long-term archival storage of documents; it does not relate to accessibility and is not identical to PDF/UA.

1.3. Normative references

Common methodological and technical sources:

Russian national profile examples:

The methodology is advisory in nature, is not a normative legal act and does not establish mandatory requirements. An international or national standard may be selected voluntarily or incorporated through an applicable law, procurement instrument, contract or technical specification. The Russian sources in section 2.1 form one national example and do not define a universal profile.

2. Source versions and national profiles

The primary standard for testing is WCAG 2.2, a W3C Recommendation in the current publication of 12 December 2024. The target level, usually A or AA, is set by the customer. For level AA all applicable criteria of levels A and AA must be met, not the average or the majority.

For organising the sample and the report, WCAG Evaluation Methodology 2.0 applies, published on 23 July 2026 as a W3C Group Note. As of the date of this version it is no longer a draft, however a Group Note is not equal to a W3C Recommendation and does not add requirements to WCAG. WCAG-EM 1.0 of 10 July 2014 is retained as the previous version and must not be mistakenly presented as the current one.

For cognitive accessibility, Making Content Usable for People with Cognitive and Learning Disabilities is used. This is an advisory W3C supplement, not part of the WCAG conformance requirements. COGA checks in the report receive a separate label "advisory check" or "Draft proposal" if the customer has made them a contractual condition.

For describing transparent test rules, ACT Rules Format 1.1 can be used. The rule format does not guarantee completeness of implementation and does not turn an individual rule into a certificate of conformance with the entire standard.

2.1. Russian profile

Government of the Russian Federation Resolution No. 102 of 07.02.2026 is effective from 01.03.2026 and forms a special mandatory profile for the official websites of state authorities, local government authorities, and organizations subordinate to them, with respect to accessibility for people with visual disabilities. It is not a general rule for all websites and all types of limitations.

GOST R 52872-2019 has the general subject of accessibility of internet resources, applications, and interfaces. GOST R 70176-2022 applies to PDF documents, not to a website interface or to a conversion process; PDF/A is not equal to PDF/UA. GOST R ISO 21801-1-2022 gives general guidance on cognitive accessibility, but is not an independently measurable website checklist. These standards have different subjects and are not identical to WCAG 2.2 AA. The report maintains a separate mapping for them.

Order of the Ministry of Labour and Social Protection of the Russian Federation No. 647n of 14.11.2025 enters into force on 01.09.2027 and is not in force as of the date of this version. The posted DOCX appendix has unfilled details in the approval stamp, therefore its content can be used only as a source of the project idea of a plain-text adaptation cycle, and not as a verified operative normative text.

3. Assessment scope passport

The assessment unit and the method of determining it depend on the type of resource: for websites and web applications it is the pair "resource + presentation/state" (Section A.1); for other types of resources, the assessment unit, its boundaries, and the inventory method are determined by the corresponding section of Part II. The passport fields below are common to all resource types; an industry-specific section may supplement them with specific fields.

Before testing, the following are recorded:

  • owner, name, and purpose of the resource;
  • exact addresses, versions, or identifiers of the resource and exclusions (for a website: domains, subdomains, and exclusions, Section A);
  • version, build number, deployment date, or control date;
  • public and authenticated areas, user roles;
  • language versions, regional variants, and personalized presentations;
  • responsive variants for screens and orientations;
  • embedded third-party services, widgets, payment, and identification contours;
  • technologies the resource relies on;
  • types of documents, video, audio, and interactive content;
  • the target level of the applicable standard and additional requirements (for websites and web applications: WCAG, Section A.7);
  • the legally applicable profile with a verified basis, edition, and date;
  • the voluntary target profile, for example WCAG 2.2 AA and COGA proposals;
  • permitted test data, security restrictions, and confidentiality constraints;
  • product owner, responsible author or content editor, and an available support service route for reporting a found barrier.

The scope is described so that another specialist could choose the same product and version. Formulations such as "the main site", "the current version", and "typical pages" without an enumeration are insufficient.

Project proposal. If applicable law provides for an exception from requirements (for example, disproportionate burden), the basis is documented in the passport; the methodology does not create such a basis. The assessor reflects in the conclusion the fact of a declared exception and does not confirm its justification as a fact, but records a discrepancy if the stated reason is evidently not consistent with the observed situation. The approach to a documented exception is described by analogy with the Finnish law on digital services (Laki digitaalisten palvelujen tarjoamisesta 306/2019, §§ 8, 10d-10e).

The legally applicable profile and the voluntary profile are assessed separately. If law or a contract refers to a different edition of requirements, a successful check of the standard applicable to the type of resource does not substitute for the legal analysis, and a narrower legal profile does not limit the voluntary target. The conclusion provides a separate mapping for each profile. In this project WCAG 2.2 AA is the recommended voluntary target for websites and web applications, as long as no other target is established for a specific scope on a verified basis.

If the product is represented by several types of resource (for example, a website, a mobile application, a desktop version, and generated documents), the passport is drawn up separately for each resource type, and the assessment of one type is not automatically transferred to another (Sections A, B, C, D, E of Part II). The product's ability to generate accessible documents is checked separately from the accessibility of its interface (Section B).

4. Three layers of testing

4.1. Automated checks

Automated tools are applied to software-determinable features: missing accessible names, some structural errors, document language, part of contrast, duplicate identifiers, and certain ARIA states. The results are re-checked by a person. An automated pass means only that the rule detected no error in the checked area.

The report states the rule set and version, the pages and states checked, the number of elements checked, each confirmed defect, false positives, and the remainder requiring manual checking. A repeated HTML id does not by itself mean a violation of the WCAG 2.2 success criterion 4.1.1 excluded from conformance: its effect on applicable criteria is established separately, for example 1.3.1 or 4.1.2. Automated results are not mixed with manual results into a single percentage.

4.2. Manual check against the standard

The assessor walks through all applicable criteria of the target level on each unit of the sample, checks full pages, processes, supported ways of using technology, and non-interference. A repeated component may not be analyzed again in every identical implementation, but all differing states and templates are checked. Basis: the five WCAG conformance requirements and step 4 of WCAG-EM 2.0.

The route with assistive technology is performed from an agreed starting point to a confirmed result and covers applicable transitions: loading, empty result, error, correction and clearing the error, success, closing, and returning. For each step, the state before the action, the action, the name, role, state, value, and available actions of the element, the focus, and the state after the action are recorded. Verbatim speech output is stored only as additional evidence of a specific configuration, while the expected result is formulated in terms of the conveyed meaning and the possibility to continue the task. Applicable component routes are taken from the six base cards, and the run sequence from the manual checks guide.

Project proposal. The formula for counting the result of a criterion across the sample, by analogy with the RGPA portal rule on conformity assessment, if the conclusion aggregates several units: a criterion is considered fulfilled at sample level only if it is fulfilled on all units of the sample where it is applicable; a criterion is considered applicable at sample level if it is applicable on at least one unit of the sample; a criterion is considered not applicable at sample level only if it is not applicable on all units without exception. The formula does not replace the defect card for each unit (Section 5) and applies only to the aggregated row of the control results registry (Template 9.5). The aggregated result of a criterion neither cancels nor weakens the conclusion rules of Section 7 and does not by itself affect the conclusion on conformance of the entire declared scope.

4.3. Testing with users

People with disabilities and perceptual differences perform critical tasks in familiar conditions. Achievement of the goal, comprehension, errors, returns, unsolicited help, data loss, load, and trust in the result are checked. The composition reflects different interaction models, not only diagnoses. One participant does not represent a group.

W3C WAI points out that evaluation with users reveals usability problems that are not detected by a single conformance check. The results cannot be used to claim that a WCAG criterion is fulfilled without checking the normative condition.

Project proposal. The initial research cycle includes at least one participant for each substantially different interaction model in a critical process. The composition is expanded when new classes of barriers appear. This is a planning rule, not a normative quota, and not statistical proof.

Before the session, the participant receives an accessible description of the goal, tasks, risks, compensation, data collected, retention period, and deletion. The participant may take a break, end participation, or withdraw consent to use the recording without explaining reasons. Audio, video, or screen recording requires separate consent. The team provides the familiar configuration, the necessary conditions for communication, and safe typical test conditions without real payment and without third-party data. The moderator stops the task in case of discomfort, a risk of data loss, or a participant's request, offers assistance, and conducts a debrief. After the session, the participant is informed of an available support channel. The research protocol stores the participant code and a minimum of personal data.

The tasks and the rules of conducting user testing for any type of resource (Sections A-E of Part II) are set in the appendix "Protocol of clarity and cognitive accessibility": checks K01-K12, consent and stopping, the observation card, and the rules on decision and storage. Its result is stored separately from the conformity decision and does not replace the test protocols of the Part II sections.

A route that an expert has completed with a screen reader, magnifier, keyboard, or other means belongs to manual checking. It becomes user testimony only in a separate session with a participant for whom the selected configuration is familiar, with consent and observation rules observed. The report does not combine expert observation, an automated signal, and participant experience into a single piece of evidence.

4.4. Differing needs and presentation settings

Contrast, color saturation, screen brightness, motion, and information density are assessed as different characteristics. A distinguishability requirement does not mean a requirement for a bright design of the whole resource. A calm composition is compatible with sufficient contrast; the preferences of people with a single perceptual characteristic may vary, and constraints may combine. The assessor does not choose settings on a person's behalf on the basis of a diagnosis.

Project proposal. If the resource offers settings for colors, motion, or the amount of secondary information, the test passport adds the original presentation, the selected value, the method of changing it, the save result, and the method of returning. The accessibility of the controls by keyboard and assistive technology, the preservation of content, functions, and visible focus are checked. An automated pass of one configuration is not transferred to others; a user preference is recorded separately from criterion conformance.

When a lower-contrast combination chosen by a person is used, the condition of the W3C explanation to WCAG 1.4.3 is taken into account: there is a mechanism for returning to the required values. Without a check of this condition, low contrast cannot be automatically excluded from defects; the presence of a setting does not by itself give grounds to consider the original low-contrast interface conformant. The decision on the criterion and the evidence of the return are recorded separately from the observation about comfort. The limitations of automated checking of settings are stated in the report.

5. Evidence and recording of the result

For each result the following are stored:

  • check identifier and source;
  • URL, user role, state, and step of the process;
  • date, product version, and configuration;
  • preconditions and test data without sensitive information;
  • exact actions;
  • expected result;
  • actual result;
  • status: conforms, does not conform, not applicable, not checked, or requires expert decision;
  • evidence sufficient for reproduction;
  • impact on the user and affected interaction models;
  • the relation to a recurring component and to other pages;
  • the state after the action, and for an error also the state after correction and confirmation that the previous error is no longer presented as current.

A screenshot is useful for a visual defect, but it does not prove focus order, screen reader announcement, or dynamic behaviour. Such cases require a sequence of actions and the observed output. Code helps establish the cause, but the test result describes the user-observable outcome.

For a route with assistive technology, the observed output is recorded in a structured way: accessible name, role, state, value and actions, focus position, change after the action, and the ability to continue the process. An audio recording or a verbatim transcript supplements this record, but the wording of a particular synthesizer is not a universal expected result.

The status "not applicable" requires justification. "Not tested" is not considered a success. If specialists disagree, both interpretations are preserved, along with the normative text, the chosen decision, and the person responsible for it.

6. Defect severity

Severity is assigned independently of the level A, AA, or AAA. The WCAG level reflects the conformance model, while severity reflects the harm in a specific product.

Level Definition Typical example
S2, high A substantial function is unavailable or requires disproportionate assistance; the workaround is unstable or not obvious Form errors are not announced, the user repeatedly loses entered data
S3, medium The task is achievable, but the load, time, or probability of error increases Inconsistent labels, weak orientation in a long process
S4, low A local deviation with limited impact A non-critical decorative icon creates extra noise

Fix priority takes into account severity, how widely the component is used, process frequency, the number of affected groups, the cost of an error, and the probability of recurrence. A widespread S3 in a shared template may have a higher priority than a single S2 on an archived page.

Draft proposal. The remediation plan sets a remediation deadline by the severity level of the defect and agrees that deadline with the resource owner, unless otherwise established by a contract or by the requirements of a supervisory authority: a shorter deadline for S1 and S2, a longer one for S3 and S4. A widespread S1/S2 violation in a shared template or component library cannot receive a longer remediation deadline than an equivalent point violation. As a benchmark rather than a methodology norm, selected national monitoring regimes apply deadlines for the first written response of about 5 working days and for full remediation of about 8-12 weeks (Estonia, TTJA, 2024 digital accessibility report for the public sector; Norway, Uutilsynet, how the authority supervises websites and apps; United Kingdom, GOV.UK, accessibility monitoring of public sector websites and mobile apps from 2022 to 2024).

7. Conclusion rules

The conclusion contains:

  1. the exact scope, version, and period;
  2. the target WCAG profile and any additional criteria;
  3. the method and the composition of the structured, random, and process-based sample;
  4. configurations and access limitations;
  5. separate results for automated, manual, and user testing;
  6. a list of defects by severity, criteria, processes, and need groups;
  7. a limitation of the conclusion to the bounds of the sample;
  8. conditions for retesting;
  9. Draft proposal. If the site is subject to an obligation to publish an accessibility statement (declaration), its presence, the date of the last review, the consistency of the declared status with the test results, and a working feedback channel are checked separately. The completeness of the statement's content in the established form is checked only where such a form is established by law or by the customer. A discrepancy between the declared status and the test results is recorded as a separate finding regardless of the technical defects of the site. The approach is analogous to the accessibility declaration review in Finland and Denmark.

Draft proposal. If the conclusion is published as a statement of conformance with the AA profile, or is used to obtain an accessibility mark or for inclusion in a registry, it undergoes an independent control review: by a specialist or an organization that did not take part in the main testing and that is not affiliated with the contractor or the site owner, by analogy with the mandatory review of the BIK BITV-Test. The reviewer spot-checks at least 10% of the result cards and all defect cards of level S1. For an acceptance test of a release and other cases where the conclusion is not published as a statement of conformance, a second sign-off within the contractor's team is sufficient. Discrepancies are recorded and resolved before the conclusion is published; if agreement cannot be reached, both positions are preserved in the conclusion with the responsible persons indicated, by analogy with the rule of section 5 on disagreement between specialists.

Acceptable wordings:

  • "Non-conformances with the criteria were found in the tested sample";
  • "The tested full process cannot be completed by users working with the keyboard only";
  • "On the tested pages, the criterion was met in the specified configurations";
  • "The sample does not provide grounds to assert the absence of similar defects on the remaining pages".

Draft proposal. For the overall conclusion covering the entire declared scope, the methodology allows three mutually exclusive statuses, by analogy with the statuses of the RGAA accessibility declaration; each status is necessarily accompanied by a full list of non-conformances with the criteria indicated: "conforms" - when all five conformance requirements of this section are met with not a single proven defect in the declared scope; "partially conforms" - when proven defects are found, but none of them is a blocking level S1 defect in a critical process; "does not conform" - when at least one level S1 defect is found in a critical process. The status does not replace the descriptive conclusion of this section and is always accompanied by it.

The wording "the site conforms to WCAG 2.2 AA" is permissible only where there is a basis for the entire declared scope and all five conformance requirements are met. On the basis of an ordinary sample-based assessment, the result of the sample is formulated, not the conformance of the entire site. The absence of errors from automated testing cannot be described as the absence of barriers.

A single proven defect of an applicable criterion on a page or in a full process within the declared scope is sufficient to refute full conformance of that scope to the declared level. Such a defect does not prove the inaccessibility of every element, every page, or every user. A positive sample result, on the contrary, is usually insufficient to assert full conformance of the whole scope.

COGA findings are presented separately: "conforms to the proposal", "partially supports", "a user barrier was found". They neither raise nor lower the WCAG level on their own.

8. Retesting and regression

A fix is tested in the original configuration using the same steps and evidence. Then regression is performed:

  • of all states of the changed component;
  • of all sampled pages where the component recurs;
  • of the full process that includes the change;
  • of adjacent criteria that the fix could have affected;
  • of at least one other assistive technology configuration for a high-risk change.

For a new release, the scope, versions, and technology map are updated. Previously problematic pages, changed templates, critical processes, and a new random part are included in the retest sample. Using exactly the same pages allows comparing dynamics, but does not reveal new areas, so the permanent sample is complemented by an updated one.

Draft proposal. If the methodology is applied as regular monitoring of several resources of the same owner or agency (section 1.1, scope of application "regular monitoring"), the composition of the resource sample for each cycle is built on the following principle: the sample size of a cycle depends on the size of the resource portfolio, the sample is rotated so that part of the objects repeat to compare dynamics and part is updated to cover new areas, and no resource in the portfolio remains untested for longer than the period agreed in the monitoring passport. As an example of such a principle, and not as a mandatory methodology norm, the European Commission Decision (EU) 2018/1524 and the practice of a number of EU Member States use the formula "base number plus an increment for portfolio size" and a rotation rule of at least 50% new and at least 10% repeated objects per cycle (Estonia, TTJA, public sector digital accessibility; Denmark, Digitaliseringsstyrelsen, monitoring and supervision; Spain, Observatorio de Accesibilidad Web, monitoring report). The monitoring passport justifies the chosen volume and rotation parameters for the specific portfolio.

If the meaning of a process, the mandatory steps, the order, the role, the input data, the endpoint, or the legally significant outcome has changed, the process receives a new baseline version and is tested in full. It cannot be considered the same process as before with a local regression. The original and the new typical test conditions are kept separate.

Draft proposal. Blocking fixes and changes to a shared component are not closed based on the original example alone. It is required to test all known occurrences, or to documentively establish a single implementation source and run a representative regression.

9. Templates

Templates 9.1-9.6 are used when formatting a document package as appendices 1-6 to the methodology.

9.1. Assessment passport

Product and owner:
Version, build, period:
Domains and exclusions:
Roles and language versions:
Legally applicable profile and basis:
Voluntary target profile:
Target WCAG profile:
Additional requirements:
Technologies and third-party components:
Critical processes:
Limitations of access and test data:
Version of the master registry, enumeration method, and deduplication:
Size of the structured sample, 10% calculation, seed, and selection date:
Responsible persons:

9.2. Coverage matrix

Unit ID:
URL or way to open the state:
Type: structured / random / full process:
Page template:
Features:
Technologies:
Role and state:
Fixture, login method, session TTL, and reset:
Adaptive variant:
Documents and media:
Criteria and configurations covered:
Reason for inclusion:
Related full processes and states:

9.3. Defect card

ID and short title:
Severity and justification:
Requirement source and direct link:
Page, state, version:
Configuration:
Preconditions:
Fixture, test role, login method, and TTL:
Steps:
Exact sequence of keys/gestures:
Expected result:
Actual result:
Evidence:
Impact on the user:
Reproducibility and related pages:
Risk owner and deadline:
Temporary available workaround:
Proposed acceptance criterion for the fix:
Related component and regression scope:
Re-verification status:

9.4. User protocol

Participant code without extra personal data:
Interaction model and familiar configuration:
Agreed conditions and communication format:
Informed consent and date:
Compensation and conditions:
Separate consent for audio/video/screen:
Right to stop and the data withdrawal decision:
Retention period and deletion date:
Available conditions, breaks, and support:
Risk, incident, and post-session support:
Task goal and initial state:
Observed steps:
Achieved result:
Errors, backtracking, hints, and assistance:
Understanding of consequences:
Data or context loss:
Self-assessment of load and confidence:
Participant quote with consent:
Barrier and link to the defect:
Session status: not conducted / completed / stopped

9.5. Verification results registry

Check:
WCAG criterion or project proposal:
Sample unit:
Product version/build:
OS and version:
Browser and version:
AT and version:
Language, viewport, orientation, and zoom:
Input method and system settings:
Fixture and sequence of keys/gestures:
Status:
Expected result:
Actual result:
Evidence:
Defect:
Link to the AT output or other evidence:
Related component and regression:
Assessor and date:
Applicability note:

9.6. Structure of the final report

  1. Executive summary for decision-making
  2. Scope, version and limitations
  3. Standards and additional rules
  4. Research method and sample
  5. Configurations
  6. Results of automated checks
  7. Results of manual checks
  8. Results of user checks
  9. Defects by severity and by process
  10. Coverage matrix
  11. Limitations of the conclusion
  12. Remediation and re-check plan

10. Section readiness table

The status "ready" means that the section has passed internal validation on live resources and is published in this version of the methodology; the status is not an approval, certification or a statement of conformance. The status "in progress" means a published section with framework provisions that may change following validation. The status "planned" means a section that has not been validated, presented in outline form to preserve the integrity of the methodology; such a section does not replace a separate test protocol that will be issued later.

Automated checking of the project service (section A.7, the automatization map) as of the date of this version is implemented only for websites and web applications. For electronic documents, a reproducible experimental analyser of individual software features has been issued (section B.6); for mobile applications, the capabilities of static package analysis are described (section V.6); and for digital services, an experimental scenario for recording individual steps, focus and software messages (section D.6). These tools are not included in the project service, are not a mandatory automated layer, and do not confirm the passing of a check. For desktop software and self-service terminals there is no automated checking: assessment is carried out manually unless separately announced upon update of the relevant section. Instrumental measurement of terminal dimensions, force, sound and contrast remains a manual check and does not yield an automated status.

Resource type Section Status
Electronic documents B in progress: test protocols issued (annex, version 0.1), validation not complete
Mobile applications V in progress: test protocols issued (annex, version 0.1), validation not complete
Desktop software G in progress: test protocols issued (annex, version 0.1), validation not complete
Digital services and processes with authentication D in progress: test protocols issued (annex, version 0.1), validation not complete
Self-service terminals and kiosks E in progress: test protocols issued (annex, version 0.1), validation not complete

Part II. Sections by resource type

Each section of Part II applies Part I to a specific type of digital resource: it defines the unit of assessment, the configurations and, where possible, the checklist. A section that is absent in expanded form in this version is still subject to Part I; only industry-specific detail is missing (section 10).

A. Websites and web applications

Section status: ready.

What is already applicable from the general part. The section uses Part I without change: purpose and principle (section 1), source versions and national profiles (section 2), scope passport (section 3), the three check layers (section 4), evidence and result registration (section 5), defect severity (section 6), conclusion rules (section 7), re-check and regression (section 8) and templates (section 9). Below are the specifics of inventory, sampling, configurations and the checklist for websites and web applications.

A.1. Inventory

The assessor identifies:

  1. common pages and views: home, sign-in, registration, search, sitemap, contacts, help, legal information, accessibility statement and feedback;
  2. essential functions: obtaining a service, creating a document, filling in a form, submission, payment, confirmation, uploading, publishing, restoring access;
  3. page and state types: templates, roles, errors, empty results, modal dialogs, notifications, tables, editors, panels and dynamic states;
  4. technologies: HTML, CSS, JavaScript, WAI-ARIA, graphics, media players, document formats and third-party components;
  5. sensitive and high-risk functions: authentication, personal data, legal and financial actions;
  6. differing adaptive, language and personalised variants.

This scheme follows step 2 of WCAG-EM 2.0.

Before samples are selected, a general register of the scope is created. The unit is not only a URL but the pair "resource + view/state": a single URL with a public and an authenticated role, an open dialog, an error or a mobile layout may yield several units. The register is built from the sitemap and application routes, the CMS export, an allowed crawl of links, the list of processes, and the owner's data. For each unit the following are recorded: a stable ID, a canonical URL or way of opening it, role, language, template, function, technology, state, date of discovery and the source of the listing.

Deduplication is performed before random selection. Exact duplicates of a canonical URL with the same content and state are merged while retaining all original addresses. Pages of the same template are not merged if the function, author, data type, language, role, interactive state or technology differ. The merging decision is recorded so that another assessor can reconstruct the general list.

A.2. Structured sample

The structured sample includes instances of all identified common views, essential functions, types, technologies and other relevant pages. One instance may cover several attributes, but the matrix must show this. The size depends on complexity, heterogeneity, the number of authors, technologies, interactivity and the required confidence. A fixed number of pages that takes no account of diversity is not a justified sample.

A.3. Random sample

Under step 3.2 of WCAG-EM 2.0, the number of randomly selected unique pages, views or states amounts to 10% of the size of the structured sample. They are added on top of it. If the structured sample contains 80 units, the random sample contains 8, and the total before adding the pages of complete processes equals 88. The share is counted from the structured sample, not from the total number of pages of the site.

Draft proposal. When the result is fractional, rounding up is applied so that the random part does not become less than 10%. For a structured sample of 1 to 9 units, at least one random unit is included if material other than the structured sample exists.

Algorithm for the random part:

  1. freeze the version of the general register and the structured sample;
  2. remove all structured units and documented duplicates from the candidates;
  3. calculate 10% of the number of structured units; rounding up is a draft proposal;
  4. record the algorithm, the seed, the date and the order of the original list;
  5. select unique units without replacement and add them on top of the structured sample;
  6. store the selected and the rejected as inaccessible units with the reason for replacement;
  7. if the random part revealed a new type, technology, state or a new class of defect, add the corresponding instances to the structured part, recalculate the random part and repeat the comparison until representativeness is achieved.

If there are fewer candidates than the calculated volume, the entire remainder is included. With an empty remainder the random part equals zero and the reason is recorded. A unit that cannot be checked remains in scope with the status "not checked"; replacement in the sample does not remove this limitation and does not justify a positive conclusion about it.

The selection must not be manually tailored for convenience or changed after viewing the results. A pseudorandom algorithm with a seed fixed in advance is acceptable as a draft proposal; its reproducibility is necessary for repetition. The general list, the method of obtaining it, the algorithm and the listing limitations form part of the evidence.

Draft proposal. The scope of the check (section 3) is defined by the customer in the assessment passport. Within the already agreed scope, the final composition of the structured and random sample is determined by the assessor and is not agreed with the customer page by page, by analogy with the rule of BIK BITV-Test on forming a sample. The customer is entitled to provide a list of critical processes and technologies for the inventory (section A.1) and to indicate objective access limitations, but is not entitled to exclude specific pages, states or instances from an already formed sample on grounds of inconvenience of checking. Any change to the sample at the customer's initiative after the start of the check is recorded separately with the initiator and the justification indicated, and cannot lead to the exclusion of an already identified defect from the conclusion.

A.4. Complete processes and states

If the selected unit is part of a process, the starting point, all pages and states of the main sequence, and frequently used and critical branches are included. For each transition the actions and data are recorded. A single URL is insufficient, because the state may depend on the role, the filling in, an error or a server response. At minimum, the following are tested if they exist: the success path, mandatory fields, one invalid value, correction, back, cancel, re-authentication, timeout, progress preservation, and final confirmation. This extends the base sequence of WCAG-EM for the assessment of real risk and is marked as a Project proposal.

The authorized standard test conditions (in testing: fixture) include the permitted test environment, the test role and rights, anonymized data, the way to obtain the initial state without passing the password to the assessor, the session lifetime, MFA or CAPTCHA conditions, expected server responses, the reset procedure, and the cleanup owner. Real payments, submission of applications, and modification of other people's data are not performed. Records and screenshots mask personal, payment, and authentication data; temporary accounts and data after a run are revoked or cleaned up.

WCAG 2.2 requires that, when a claim of process conformance is made, all its pages conform to the target level. And WCAG-EM 2.0 requires including all units of the selected process in the sample.

Section D extends the approach to end-to-end processes with authorization to the cases where a process continues beyond a single type of resource (for example, it starts on the website and finishes in a mobile application).

A.5. When the whole site is tested

Full testing is preferable if the product is small, the number of presentations is limited, or it cannot be meaningfully divided. Sampling reduces the volume of testing but limits the conclusion. It does not prove the absence of defects outside the tested units. WCAG-EM 2.0 warns that a single sample cannot be used to claim conformance of the whole site, because errors may remain on untested pages.

A.6. Configurations and assistive technologies

Before the start, the technological support baseline is fixed: operating systems, browsers, screen sizes and orientations, zoom, high-contrast mode or system colors, on-screen keyboard, screen readers, keyboard-only operation, voice or switch input, magnification, and reduced-motion settings.

Project proposal. The minimum matrix for a public adaptive website includes:

  • a current desktop browser with a keyboard;
  • a current desktop browser with a screen reader common in the target environment and language;
  • a mobile browser with the built-in screen reader;
  • zoom up to 200% and a reflow check at a width equivalent to 320 CSS pixels;
  • the system reduced-motion setting;
  • a no-sound mode and a check of visual equivalents;
  • an alternative method of pointer input if a critical function uses complex gestures or dragging.

Versions are selected according to the real audience and are fixed in full. For each run the following are mandatory: product version/build, OS and version, browser and version, assistive technology and version, interface language, viewport and orientation, zoom, input method, system motion/contrast settings, date, result, and a link to the evidence. Omitting a field is allowed only with a reason. Success in one combination does not transfer to all platforms. WCAG links "accessibility supported" means of using a technology with tested compatibility with assistive technologies and user agents: the definition of accessibility supported.

Project proposal. The methodology owner maintains and updates at least annually a separate reference list (the "tool registry"): specific versions of browsers, operating systems, and assistive technologies, including at least one screen reader common among the languages and audiences of the assessed product, which the assessors use by default in the absence of other instructions from the customer, by analogy with the tool list of BIK BITV-Test. A deviation from the reference list in a specific assessment is recorded in the assessment passport with a reason. The list fixes the assessment environment and is not a purchasing recommendation.

A.7. Basic set of 34 control checks

These 34 items serve as the basic route of testing for frequent risks and are not a complete checklist of WCAG 2.2 A/AA. To claim an AA profile, the assessor must pass all 55 current A and AA criteria per the appendix "WCAG 2.2 A/AA Check Protocols" and link each result to a sample unit, state, and evidence. The excluded criterion 4.1.1 is not part of the 55; criteria 2.2.6 and 2.5.6 are at the AAA level and are not part of the mandatory AA profile.

Each check is performed on the applicable sample units and states. In the "Source" column, WCAG means the requirement being tested, and COGA means an advisory supplement.

№ Control check Source Expected result
2 Headings, lists, regions, and tables WCAG 1.3.1 Visual relationships are programmatically determinable; reading order and cell relationships are understandable
3 Meaningful sequence WCAG 1.3.2 The programmatic order preserves meaning when read linearly
4 Instructions not relying only on form, color, or position WCAG 1.3.3, 1.4.1 The user identifies the required element without perceiving a specific color, shape, or coordinate
5 Text contrast WCAG 1.4.3 Normal text is at least 4.5:1, large text at least 3:1, taking the criterion exceptions into account
6 Contrast of components and graphics WCAG 1.4.11 Necessary boundaries, states, and meaningful graphical objects reach 3:1 against adjacent colors
7 Text enlargement WCAG 1.4.4 At 200% there is no loss of content or function, except for the provided exceptions
8 Reflow WCAG 1.4.10 At the equivalent of 320 CSS pixels there is no two-dimensional scrolling for ordinary content and nothing is clipped
9 Changing text spacing WCAG 1.4.12 The spacings specified by the criterion do not cause loss of content or function
10 Keyboard accessibility WCAG 2.1.1 All functions are performed through the keyboard interface without pointer emulation
11 No keyboard trap WCAG 2.1.2 Focus can be moved into the component and out of it in a standard or documented way
12 Control of moving and updating content WCAG 2.2.2 The user can stop, hide, or control the frequency of applicable motion and updating
13 Flashing WCAG 2.3.1 There is no flashing more than three times per second, or it is below the established thresholds
14 Focus order WCAG 2.4.3 Focus movement preserves meaning and manageability, including modal windows and return
15 Link purpose WCAG 2.4.4 The purpose of each link is understandable from the text and the programmatically associated context
16 Focus visibility and obstruction WCAG 2.4.7, 2.4.11 Keyboard focus is visible and not completely hidden by fixed panels or other author content
17 Alternative to dragging WCAG 2.5.7 A function requiring drag is available via a single-pointer action if drag is not essential
18 Target size WCAG 2.5.8 The target is at least 24 by 24 CSS pixels or meets one of the criterion exceptions
19 Match of visible label and accessible name WCAG 2.5.3 The accessible name contains the visible label text, and voice commands select the expected element
20 Language of the page and of fragments WCAG 3.1.1, 3.1.2 The main language and the differing fragments are programmatically determinable
21 Predictable behavior WCAG 3.2.1, 3.2.2 Receiving focus does not cause a change of context; changing a component setting does not cause it automatically, except in cases where the user was warned before using the component
22 Consistent navigation, identification, and help WCAG 3.2.3, 3.2.4, 3.2.6 Repeated elements retain their order and naming, and help mechanisms are located in a consistent place
23 Form labels and instructions WCAG 3.3.2 When input is requested, the necessary labels or instructions are provided. Permanently visible labels are an additional project rule; the result under WCAG is determined by the normative condition
24 Error detection and correction WCAG 3.3.1, 3.3.3 The error is named in text, a known way of correction is suggested if it exists and does not jeopardize security; programmatic association and screen reader announcement are checked separately under 1.3.1, 4.1.2, and 4.1.3 in the specific configuration
25 Prevention of irreversible errors WCAG 3.3.4 For the operations listed by the criterion, one of the options is completed: cancelling a submission; checking the input and the possibility of correcting it; or reviewing, confirming, and correcting before final submission
26 Exception of repeated entry WCAG 3.3.7 Previously entered data within the process is auto-populated or available for selection, except for the criterion exceptions
27 Accessible authentication WCAG 3.3.8 Logging in does not require a cognitive test without an accessible alternative, assistance, or object recognition, except as provided
28 Name, role, value, and state WCAG 4.1.2 For a user component, the name and role are programmatically determinable, user-settable states/values are available to be set, and changes are reported to user agents; the actual output is recorded in the declared AT configuration
29 Status messages WCAG 4.1.3 Messages meeting the status definition are programmatically determinable in such a way that AT can present them without receiving focus; the actual output and support are checked separately
30 Separate media checks WCAG 1.2.1, 1.2.2, 1.2.3, 1.2.4, 1.2.5 For each applicable type, the following are recorded separately: an alternative for prerecorded audio/video, captions for prerecorded video, audio description or a level A media alternative, live captions, and audio description for prerecorded video at level AA
31 Downloadable document and the full process WCAG 1.1.1, 1.3.1, full page and process For the document itself, the applicable criteria and the format profile are checked separately; for the page and the process, the accessibility of obtaining it and the equivalence of the alternative are checked separately. GOST R 70176-2022 applies to PDF, and PDF/A does not prove PDF/UA
32 Clarity and literalness of content COGA Design Guide Frequent words, short meaningful blocks, literal labels, and separate instructions are checked with target users; the result is marked as advisory
33 Error prevention, back, and data preservation COGA Objective 4 The user understands the consequences, can go back and correct data without losing work already done; the check is advisory
34 Attention, memory, and help COGA Summary The critical path is not overloaded, progress and context are visible, help is available at the moment of difficulty, and the process does not require holding information in memory; the check is advisory

The check of the time limit under WCAG 2.2.1 is performed by a separate card of the full registry: for each limit set by the content, switching off, adjusting, or extending it and the exact criterion exceptions are checked. Preservation of entered data after a timeout is a separate cognitive Project proposal and does not substitute for 2.2.1.

Items 32-34 are performed per the appendix "Protocol of clarity and cognitive accessibility", where K01-K12 are defined as authored scenarios. Their result is stored separately from the decision on the 55 WCAG A/AA criteria and does not automatically raise or lower the WCAG level.

B. Electronic documents

Section status: in progress.

What already applies from the general part. Purpose and principle (section 1), the three layers of testing (section 4), evidence and result recording (section 5), defect severity (section 6), conclusion rules (section 7), and re-testing (section 8) apply unchanged. The scope passport (section 3) is supplemented with the document registry from B.2. The templates of section 9 are used with web-specific fields replaced by the file identifier, page, or document object and the viewing configuration.

The procedures for all control checks B-01-B-17 are issued in the appendix "Electronic Document Check Protocols". The basic set is a route of frequent checking, not a claim of full conformance to the standard. The legally applicable profile and the editions of the sources are still fixed in the scope passport.

Grounds for the section:

  • EN 301 549 V4.1.1, section 10 extends the requirements to non-web documents, including downloadable documents associated with web pages, and does not exclude documents solely because of encryption, electronic signature, or a watermark;
  • WCAG2ICT explains the application of WCAG 2.2 criteria to non-web documents and software. It is an informative W3C document, not a separate conformance standard;
  • ISO 14289-1:2014 specifies the application of ISO 32000-1 for accessible electronic PDF documents (PDF/UA-1). The mere presence of a PDF/UA claim does not replace an observable check;
  • the official card for GOST R 70176-2022 gives the exact title of the standard: "Internet resources and other information presented in electronic digital form. PDF format files. Accessibility requirements for people with disabilities and other persons with limited abilities". The card confirms the PDF scope and is not used here to assert the provenance of the standard's text;
  • W3C techniques for PDF clarify verifiable implementations: PDF3 for reading order and bypass, PDF6 for tables, PDF7 for recognition of scanned text, PDF9 for headings, PDF10 and PDF23 for forms, PDF16 and PDF19 for language, PDF18 for the document title, PDF11 for links, PDF21 for lists and PDF2 for bookmarks.

B.1. Object and unit of evaluation

The object of this section is PDF files, word processor documents, presentations, spreadsheets and fillable forms that the resource provides to the user for reading, downloading, filling in, signing or further processing. The scope includes attachments of public and authenticated services, documents inside archives if the user has to open them, as well as documents generated by the resource as a result of a user action.

The default unit of evaluation is the document as a whole in a specific published version. For a compilation of independently created parts, an independent part may be a separate sampling unit if its boundaries, authorship and purpose are fixed. At the same time, the title, default language, protection, bookmarks, general navigation and link to an alternative version are checked for the file as a whole. A spreadsheet sheet or a presentation slide does not become an independent unit of evaluation merely because the tool displays it separately.

The version of a unit is determined by the checksum of the downloaded file, the date of retrieval, a direct address or a reproducible way of generating it, and for a changeable document also by the revision number and the date inside the document. Two files with the same name but different checksums are different versions.

Out of scope for this section are the viewing or editing software interface (Section G), page accessibility before the link to the file (Section A, except for the link to an alternative in B-15), media material as an independent object, and the paper original. If a document is embedded in a page, the document itself is checked under Section B, while the shell, buttons and the path of retrieval are checked under Section A. Content that merely looks like a document externally but is an ordinary web page is checked under Section A.

B.2. Inventory of the resource's documents

Before selecting samples, a master register is created. The assessor walks through the navigation and site map, search, service pages, document and publication sections, material cards, print and download links, attachments in authenticated accounts and services, as well as the results of generating certificates, receipts, reports and applications. If a content management system export or an owner file list is available, they are cross-checked against the walkthrough. Forms, logins and actions with legal consequences are not performed without a permitted fixture from Section A.4.

For each instance the following is recorded:

  • stable ID, document type and purpose;
  • source page, link text, direct address or exact generation sequence;
  • file name, declared and actual format, MIME type, size, publication date and retrieval date;
  • version number, date inside the document and cryptographic checksum;
  • language, intended audience, owner or source, template or formatting family;
  • role in a critical process, need for reading, filling in, signing or printing;
  • number of pages, slides or sheets, presence of forms, tables, images, columns, scans and protection;
  • presence of an alternative version and links between versions;
  • accessibility without authorization, role and permitted test account, if authorization is needed;
  • discovery source and the reason for exclusion from scope, if the document is excluded.

Deduplication is performed by checksum and content version, not only by name or address. All pages that link to a duplicate are preserved. Documents of the same template are not merged if they differ in purpose, author, language, generation method, presence of a form, data type or risk level. An instance that cannot be downloaded remains in the register with the status "not checked" and a reason.

B.3. Sampling

The structured part includes at least one instance of each detected combination of material characteristics: format; purpose; source or template; language; document for reading, filling in or signing; presence of tables, images, columns, forms, protection, scans; participation in a critical process. All documents in which an error may deprive the user of an entitlement, a payment, a deadline, safety or the ability to complete a critical process are included. A single instance may cover several characteristics, which is reflected in the coverage matrix.

The random part is added on top of the structured part. Its size equals 10% of the number of units in the structured part, rounded up. If the structured part contains from 1 to 9 units and there is a remainder outside it, at least one random unit is added. This numeric rule and the rounding are a Project proposal by analogy with A.3.

Reproducible selection is performed as follows:

  1. freeze the master register, the structured part and the order of candidates;
  2. remove exact duplicates and already selected units;
  3. record the pseudorandom selection algorithm, the seed, the date and the encoding of the source list;
  4. select without replacement the calculated number of units; if there are fewer candidates, include the entire remainder, and with an empty remainder the random part equals zero with the reason recorded;
  5. save the result and the reasons for replacing unavailable units;
  6. if a new type, template, risk or defect class is found, add its instance to the structured part, recalculate the random part and repeat until stable.

Project proposal. All documents are checked if the master register contains no more than 20 unique documents; if this is required by a contract or acceptance; if every document participates in a critical process; or if the set cannot be meaningfully split into repeating families. A conclusion from the sample is not automatically extended to unchecked documents. To claim conformance of the whole set, evidence is required for each unit included in it.

B.4. Configurations and assistive technologies

Before checking, a configuration matrix is fixed. It includes at minimum:

  • a viewer or editor with a screen reader program;
  • viewing with magnification of at least 200%;
  • a high contrast mode or system colors, if supported by the environment;
  • reflow or an equivalent narrow mode without mandatory horizontal scrolling, if the format and program support it;
  • keyboard-only operation for links, bookmarks, table of contents, forms and embedded elements;
  • extraction and copying of text to check the text layer and the order;
  • the print view, if printing is part of the purpose.

For each run, the operating system and version, viewer or editor program and version, screen reader program and version, interface language, magnification, contrast mode, input method, whether reflow is enabled, date, result and evidence are recorded. The names of specific products are determined by the target environment and the test passport, and not by this section. Success in one configuration is not carried over to others.

B.5. Basic set of control checks

No. Check Source Expected result
B-02 Headings WCAG 1.3.1, 2.4.6, PDF9 Visual headings are marked programmatically, levels are not substituted by formatting, and they describe the following section
B-03 Lists WCAG 1.3.1, PDF21 Lists and their items are programmatically determinable, and nesting and numbering are not lost during reading
B-04 Tables WCAG 1.3.1, PDF6 Data cells are associated with row and column headers; order and merged cells do not distort the meaning
B-05 Alternative text for images and diagrams WCAG 1.1.1 Meaningful objects have an alternative equivalent in purpose, and decorative ones are excluded from programmatic reading
B-06 Language of the document and of fragments WCAG 3.1.1, 3.1.2, PDF16, PDF19 The main language and the differing fragments are programmatically determinable and correspond to the content
B-07 Document title EN 301 549, 10.2.4.2, PDF18 The editable programmatic title describes the topic or purpose; for PDF the viewer is instructed to show the title instead of the file name
B-08 Bookmarks, table of contents and navigation Project proposal; advisory technique PDF2 (for a single document WCAG criterion 2.4.5 does not apply, see EN 301 549, 10.2.4.5) In a long document there is a working programmatic way to reach the main sections; the items correspond to the structure
B-09 Links WCAG 2.4.4, PDF11 The purpose of a link is clear from the text and the programmatic context; the link is keyboard accessible and leads to the stated target
B-10 Contrast WCAG 1.4.3, 1.4.11 Applicable text, form elements and meaningful graphics reach the required contrast, taking the exceptions of the criteria into account
B-11 Text as text, not as an image of text WCAG 1.4.5 Real text is used to convey information, except in the permitted cases of essentiality or customisation
B-12 Fonts, text extractability and scans WCAG 1.1.1, 1.4.5, PDF7 Text is extracted in the correct encoding and order; a scan has a verified text layer, and recognition errors do not change the meaning; fonts do not turn characters into unreadable codes
B-13 Fillable fields and their labels WCAG 2.1.1, 3.3.2, 4.1.2, PDF10, PDF23 Fields have programmatic labels, name, role, value and state; logical tab order, input and correction are available without a mouse
B-14 Protection does not hinder assistive technologies EN 301 549, section 10 Encryption, signing, copy restrictions and other protection do not block reading, navigation and filling in in the stated configuration
B-15 Link to an alternative version on the page WCAG, conformance requirements 1 and 4, G136 The alternative is available via an explicit link next to the document, contains the same information and functions, has the same language, currency and target conformance level
B-16 Magnification, high contrast and reflow WCAG 1.4.4, 1.4.10, WCAG2ICT In supported formats, under magnification, system colors and reflow, content, order and functions are not lost; a capability not provided by the format is not treated as a violation
B-17 Not only color, shape or position WCAG 1.3.3, 1.4.1 Instructions, required fields, statuses and values in tables and diagrams are conveyed not only by color, shape, size or position

B.6. Automatability map

An automated result is a preliminary signal. The presence of a tag or property does not prove its semantic correctness, and the absence of an extractable feature may be a limitation of the format or the analyser. The status "pass" for B-01 to B-17 is set by the assessor after the full protocol. | Check | Automatability | What the machine checks | |---|---|---| | B-01 | partially | For PDF: presence of a tag tree and of a tagged-document flag; correctness of roles and order is not established | | B-02 | partially | For word processor documents and presentations: presence of built-in styles or heading placeholders; meaning and levels are checked manually | | B-03 | manual only | The automated layer cannot reliably distinguish a semantic list from visual formatting across all formats | | B-04 | partially | For word processor documents and presentations: a programmatic header row marker; for spreadsheets only the header of a structured table, ordinary ranges remain unchecked | | B-05 | partially | For PDF: presence of an alternative text field on Figure elements; for word processor documents, presentations and spreadsheets: filling in the graphic object description field; quality and decorative nature are not assessed | | B-06 | partially | Presence of a programmatic language in the PDF catalog or in the content of an office document; correspondence with the text language and completeness of fragment markup are not assessed | | B-07 | partially | For PDF: a non-empty metadata title and an instruction to display it instead of the file name | | B-08 | manual only | The presence of bookmarks without checking completeness, order and purpose is not sufficient for a result | | B-09 | manual only | Purpose, context, keyboard accessibility and actual intent require observation | | B-10 | manual only | Extracting colors from real documents proved unreliable for the basic cross-format layer | | B-11 | manual only | The machine does not establish whether an image of text is essential or an adjustable exception | | B-12 | partially | For PDF: the number of pages with extractable text and the fact of the absence of a text layer; recognition accuracy, encoding, order and fonts are checked manually | | B-13 | manual only | The full order, announcements, input, errors and output of a screen reader are checked in the configuration | | B-14 | partially | For PDF: presence of encryption and permission for text and graphics extraction; actual operation of the assistive technology is checked manually | | B-15 | manual only | Equivalence, currency, language, reachability and feedback of versions require comparison | | B-16 | manual only | Losses at magnification, with system colors and on reflow are observed in the stated configuration | | B-17 | manual only | The machine does not establish whether meaning is conveyed by color or position alone; a check without color perception is required |

B.7. Evidence and severity

The result card additionally contains the document ID and checksum, the source page and direct address, format and version, the number of the page, slide, sheet, section or object, the configuration and the exact method of reproduction. For programmatic structure a fragment of the tag tree or of the object properties is attached; for a visual defect a snapshot with context and measured values; for reading order the sequence of the actual output; for a form the sequence of keys, focus, the announcement and the observable output of the screen reader. A snapshot without programmatic structure does not prove a semantic defect, and an analyzer report without the source object does not prove its meaning.

Severity is assigned according to section 6 of Part I, and not according to the check number. Typical reference points:

  • S1: an inaccessible document blocks a critical service, a right, a payment or a mandatory action and there is no equivalent workaround;
  • S2: a significant part of the content or of a form is inaccessible to the target group, but a difficult or external workaround exists;
  • S3: the defect interferes with a separate part of the document, with navigation or with understanding, but the main task remains performable;
  • S4: limited inconvenience without loss of substantial information or function.

One template defect is registered with a list of all confirmed instances and an assessment of prevalence. Severity cannot be reduced merely because the violation was found automatically or is repeated in many files. An inaccessible mandatory document does not become accessible because a contact phone number is present, if obtaining an equivalent is not part of the same deadline and process.

B.8. Limitations and transition to the "ready" status

This section does not prove the accessibility of the viewer application interface, the correctness of an electronic signature, the legal force of an alternative or the comprehensibility of professional content without the involvement of a subject-matter specialist and target users. The automated layer does not check the quality of semantics and does not support all varieties of office documents. Programmatic markers may be absent; the presence of metadata by itself does not confirm the quality of the structure.

The status will remain "in progress" until all of the following conditions are met simultaneously:

  1. protocol B-01-B-17 has been applied by at least two assessors to an independent corpus of PDF, word processor documents, presentations, spreadsheets and forms;
  2. the consistency of the results has been measured and ambiguous steps have been eliminated;
  3. the manual layer has been trialed with a screen reader, magnification, high contrast and reflow in fixed configurations;
  4. the user layer has been conducted with participants for whom the documents are genuinely part of the task;
  5. the automated layer has been re-verified on a marked reference set, and false positives and misses have been published;
  6. normative references and legally applicable profiles have been re-verified as of the date of issue.

Until these conditions are met, the protocol is issued and suitable for reproducible project-level testing, but the result cannot be presented as a certification, a universal compliance claim or a completed validation of the section.

V. Mobile applications

Section status: in progress.

What already applies from the general part. Purpose and principle, source versions, the scope passport, the three testing layers, evidence, defect severity, the rules of conclusion and re-testing (sections 1-8) apply without changes. Below, the specifics of a mobile application are established. Procedures V-01-V-22 are issued in the appendix "Mobile application test protocols".

Grounds for the section: EN 301 549 V4.1.1, section 11, which extends the requirements for non-web software to mobile applications; WCAG2ICT as an informative guide to applying WCAG 2 to non-web documents and software, including mobile applications; and national mappings where applicable. The Russian national example is GOST R 52872-2019, whose exact title includes applications for stationary and mobile devices and other user interfaces. WCAG2ICT is not an independent conformance standard and does not define a full set of requirements for a product.

V.1. Object and unit of assessment

The object is a mobile application. The unit of assessment is the application in a specific version and build on a specific mobile platform and device model, in a fixed configuration, role, language and state. The configuration includes the settings of V.4. A role, language or state forms a separate unit of assessment if it changes the available functions, the interface or the expected result. The result key contains the application identifier, version and build, platform, device model, configuration identifier, role, language and state.

The sampling unit is a screen, a dialog, a persistent screen state or a complete user scenario within the unit of assessment, and not a URL. Loading, an error, an empty result, an open keyboard, a system permission request, a notification, enlarged text and an active web view form separate sampling units if they change the available actions or the programmatic representation. The result of card V-NN is recorded for the combination of the unit of assessment and the sampling unit; the result of one configuration, role, language or state is not transferred to another automatically.

A hybrid application remains the object of section V. The native container, transitions, system dialogs and the final interaction are checked per V. An embedded web view is not treated as a separate web page for the purposes of EN 301 549 V4.1.1, section 11, however the applicable procedures of section A are used to check the semantics and the behavior of the web content. The result is recorded once per V-20 without double counting. If the application opens a standalone page in an external program, that page becomes a separate object of section A, while the accessibility of the transition and of the return remains in section V.

Not included in the result of section V: the accessibility of the platform as a separate product; the web version of the service; a desktop application; the content of a downloaded file; the hardware properties of the device; the quality of connectivity, security, performance and functional correctness, unless they create an accessibility barrier. A downloaded document is checked per section B, an end-to-end process with authorization is also checked per section D, and an external web resource is checked per section A. Results are not transferred automatically between platforms, devices and versions.

V.2. Inventory of screens and states

Before sampling, a master inventory is compiled:

  1. entry points, main tabs, menus, navigation levels and screens;
  1. roles, account states, settings, locales and first-run variants;
  1. empty, loading, successful, error, offline and recovered states;
  1. forms, the on-screen keyboard, autofill, search, filters and data selection;
  1. gestures, drag and drop, device motion actions and their alternatives;
  1. system permissions, notifications, app links, transitions to external tools and the return;
  1. embedded web views, media, downloaded documents and other boundaries with sections A, B and D;
  1. text, zoom, contrast, color, motion, time and orientation settings;
  1. actions with an external keyboard, switches and other alternative input;
  1. critical scenarios, background operations and states after interruption or restart.

Sources of discovery include the scope passport and critical scenarios, the description of functions and navigation, all available entry points and roles, a sequential walkthrough of the interface, the first run, notifications and external transitions, as well as supplementary results of static analysis and the accessibility tree. A single source is not considered sufficient for completeness. Before selection and review of results, the date, the master inventory version and the versions of the sources used are recorded.

For each position a stable identifier, name and purpose, the main type, the entry point, preconditions, role, expected state, available control methods, related permissions and transitions are recorded. Positions are considered identical if they have the same stable identifier, entry point, role and state, and their available actions and programmatic representation match. Duplicates are merged before sampling, and the log preserves the original identifiers, the reason for merging or excluding, and the resulting position. An unreachable screen remains in the inventory with the status "not checked" and a reason. Absence from the inventory is a limitation of scope, not a pass.

V.3. Sampling of screens and scenarios

The mandatory part includes: the first run; the main screen and each main navigation section; sign-in and sign-out, if present; application settings; a form with text, selection and an error; a system permission; a notification; empty, loading, successful and error states; search or filter; a critical scenario from the scope passport; return after interruption; an embedded web view, a document and media, where present. Each mandatory scenario is walked from the initial state to the result, including intermediate screens and system transitions. The structured part includes the mandatory part and at least one instance of every identified combination of essential characteristics: role; language; main sample unit type; state; critical or rare function; input method; permission; external transition; media or document. For the random part, each unit is assigned exactly one main type: screen, dialog, state, or scenario. A single instance may also cover other characteristics, which is reflected in the coverage matrix.

Project proposal. The random part is added on top of the structured part separately for each main type. The group denominator is the number of structured units of that type; the volume equals 10 percent of the denominator, rounded up. If the structured part of a type has from 1 to 9 units and a candidate remains after exclusions, at least one random unit is selected. A group with a zero denominator is not formed, and the absence of the type is checked as an inventory defect.

Reproducible selection is performed as follows:

  1. freeze the versions of the master registry, the structured part, and the candidate order before looking at the results;
  2. record the stable identifier and main type of each unit, then remove from the candidates the structured units and documented duplicates;
  3. record for each group the denominator, the calculated volume, the pseudorandom selection algorithm, the seed, the date, and the encoding of the source list;
  4. sort the candidates by stable identifier, shuffle with the chosen algorithm using the fixed seed, and select without replacement the first units in the calculated quantity;
  5. if there are fewer candidates than the calculated volume, include the entire remainder and record the shortfall; with an empty remainder the random part equals zero with the reason stated;
  6. leave an unavailable unit in scope with status "not verified" and a reason, and if a replacement is needed take the next element of the same shuffled list; a replacement does not lift the limitation on the unavailable unit;
  7. save the selected units, the unavailable units, the replacements, and the excluded duplicates; if a new type, state, function, or defect class is discovered, add an instance of it to the structured part, recompute the affected groups, and repeat until stable.

A re-check of the same assessment unit uses the same frozen registry, algorithm, and seed. A new registry or a new version of the application receives a separate sampling version; the previous result is retained for comparison. A screen is checked together with the applicable states. A shared component is allowed to be registered once with a list of all screens where the same defect is confirmed, but a critical scenario must not be reduced to a reference to an individual component.

V.4. Configurations

The minimal matrix includes two common mobile platforms and each platform's built-in screen reader. For each platform, the following are checked: the normal configuration; the maximum system font size; the screen magnifier; the high-contrast or invert mode, if the platform provides one; portrait and landscape orientations; an external keyboard; switch control; and reduced motion. If a combination is not supported by the device or the application, the exact reason and the status "not verified" or NA are recorded, but it is not counted as a pass.

For each run, the following are recorded: device model, platform and update version, application version and build, screen size and density, language and region, orientation, font size and scale, color mode, motion settings, screen reader version, input method, permissions granted, network state, and the initial state of the data. An emulator is allowed for preliminary checks, but it does not replace confirmation on a physical device for gestures, motion, vibration, the screen magnifier, switch control, and system notifications.

V.5. Baseline set of control checks

No. Check Source Expected result
V-02 Reading order, focus, and virtualized content WCAG 1.3.2, 2.4.3, EN 301 549 V4.1.1, 11.5.2.13-11.5.2.15, protocols G-02 and G-04 Traversal follows the meaningful sequence, does not land on hidden or empty objects, reaches lazily loaded elements, and focus after a transition or update ends up in a predictable place
V-03 Complex gestures and a single-pointer alternative WCAG 2.5.1, EN 301 549 V4.1.1, 11.2.5.1 An action requiring multiple points or movement along a path is performed with a single pointer without a path, except for an essential gesture
V-04 Pointer cancellation WCAG 2.5.2, EN 301 549 V4.1.1, 11.2.5.2 A normal action does not complete on touch-down alone, or it can be cancelled, aborted, or reversed within the exceptions of the criterion
V-05 Dragging and an alternative without dragging WCAG 2.5.7, EN 301 549 V4.1.1, 11.2.5.7, protocol G-12 Any function that involves dragging is available with a single pointer without dragging, except for an essential action or a platform action not modified by the application
V-06 Target size and spacing WCAG 2.5.8, EN 301 549 V4.1.1, 11.2.5.8 The target is at least 24 x 24 CSS px or satisfies one of the exact exceptions of the criterion; the actual activation area is measured on the device
V-07 Text enlargement and reflow WCAG 1.4.4, 1.4.10, EN 301 549 V4.1.1, 11.1.4.4, 11.1.4.10 and 11.7, protocol G-09 Text enlargement and reflow at a width of 320 CSS px or a height of 256 CSS px are checked separately, with no loss of content, functionality, or unintended two-dimensional scrolling
V-08 Portrait and landscape orientations WCAG 1.3.4, EN 301 549 V4.1.1, 11.1.3.4 Content and functions are available in both orientations, except where a specific orientation is essential
V-09 Device motion and an accessible alternative WCAG 2.5.4, EN 301 549 V4.1.1, 11.2.5.4, protocol G-13 A function triggered by device or user motion has a user interface control and can be disabled, except for the supported accessibility interface or essential motion
V-10 Contrast and independence from color WCAG 1.4.1, 1.4.3, 1.4.11, EN 301 549 V4.1.1, 11.1.4.1, 11.1.4.3 and 11.1.4.11, protocol G-10 Normal text has a contrast of at least 4.5:1, large text at least 3:1, significant non-text elements at least 3:1 with the exact exceptions; meaning is not conveyed by color alone
V-11 Notifications, errors, progress, and status messages WCAG 3.3.1, 4.1.3, EN 301 549 V4.1.1, 11.4.1.3 and 11.5.2.15, protocol G-07 The status message is programmatically determinable for presentation by assistive technology without receiving focus; the actual announcement is recorded as a separate integration subresult
V-12 Labels, instructions, errors, and form correction WCAG 2.5.3, 3.3.1, 3.3.2, 3.3.3, protocol G-14 The field has a stable visible and programmatic label, the necessary instruction, and a programmatically associated error; the accessible name contains the visible text, and correction does not require searching for the message outside the context
V-13 Input purpose, focus occlusion, and re-entry WCAG 1.3.5, 2.4.11, 3.3.7, EN 301 549 V4.1.1, 11.1.3.5, 11.2.4.11 and 11.3.3.7 The programmatically determinable purpose of supported fields, the absence of full focus occlusion by author content, and the prohibition on re-entering previously provided data are checked separately, with the exact exceptions
V-14 Time limits, motion, and flashing WCAG 2.2.1, 2.2.2, 2.3.1, EN 301 549 V4.1.1, 11.2.2.1-11.2.2.2 and 11.2.3.1, protocol G-13 The limit can be turned off, adjusted by at least ten times, or extended on warning at least 20 seconds before, at least ten times, if there is no exception; applicable motion is controllable; no more than three flashes per second, or the general and red thresholds are observed
V-15 Accessible authentication WCAG 3.3.8, protocols D-09-D-12 Sign-in, recovery, one-time codes, and re-authentication do not require a cognitive test without the assistance or alternative provided by the criterion; the whole process is additionally checked per section D
V-16 External keyboard WCAG 2.1.1, 2.1.2, 2.1.4, 2.4.3, 2.4.7, EN 301 549 V4.1.1, 11.2.4.3 and 11.2.4.7, protocols G-03 and G-11 All functions that do not depend substantially on a path are available via the external keyboard; the order preserves meaning, focus is visible, and there are no traps or unmanageable single-character shortcuts
V-17 Switch control WCAG 2.1.1, 2.4.3, 2.4.7, 4.1.2, EN 301 549 V4.1.1, 11.2.1.1, 11.2.4.3, 11.2.4.7 and 11.5.2, protocols G-01 and G-03 Sequential scanning reaches every action in meaningful order, highlighting is noticeable, the action is performed, and the current state is available without a sensory gesture
V-18 System accessibility settings EN 301 549 V4.1.1, 11.6.2 and 11.7, protocol G-08 The prohibition on impairing documented platform functions, except at the user's request, and the following of user preferences, for which an essentiality exception is permitted, are checked separately
V-19 System dialogs, permissions, and return WCAG 2.4.3, 3.2.1, 3.2.2, EN 301 549 V4.1.1, 11.2.4.3, 11.3.2.1 and 11.3.2.2, protocols G-06 and G-07 The application preserves the meaningful focus order and does not cause an unexpected change of context; Project proposal: the clarity of the request, the absence of a dead end, and data preservation are checked separately with the profile enabled, and the accessibility of the system dialog belongs to the platform
V-20 Embedded web views applicable exact normative clauses EN 301 549 V4.1.1, 11.1-11.4, mapped to each performed procedure of section A Web content passes the applicable procedures of section A by the corresponding 11.x clauses, the container and transitions pass section B, and the result is not counted twice
V-21 Audio and video WCAG 1.2.1, 1.2.2, 1.2.3, 1.2.4, 1.2.5 For each applicable type, the required text and audio alternatives are separately available, and playback control is available in the configurations of section V.4
V-22 Error prevention, confirmation, cancellation, and re-entry WCAG 3.3.4, 3.3.7, protocol G-19 and section D A significant action allows review, correction, confirmation, or cancellation within the criterion; data available during the process does not require re-entry without an exception

Project proposal. The recommendations of the selected platform for target size and the spacing between elements are recorded in the passport as an additional profile. They may be stricter than V-06, but they do not replace the threshold, the exceptions, or the units of WCAG 2.5.8. A static size in the markup is only a characteristic: the actual activation area and spacing are measured on the device.

The G protocols named in the table are performed per the appendix "Desktop Software Test Protocols" with mobile specifics: a window is replaced by a screen or state, keyboard focus is separated from screen reader focus, platform system gestures are not attributed to the application, and the result is confirmed on a device from the matrix of V.4. A reference to G does not exempt the B card and the recording of the mobile configuration.

V.6. Automatability map

Static analysis of the application package is used only for the preliminary selection of manual checks. It can detect in the accessible binary markup: images and image buttons without a text description and without an explicit decorative status; text sizes in absolute or density-independent units instead of scalable units; explicitly set sizes of interactive targets; and a prohibited orientation in the manifest. Such a signal does not see the interface created by code, the actual touch area, inherited properties, composite elements, display conditions, or the meaning of a label. As a result, the number of layouts found and successfully decoded, decoding errors, and unresolved resource references are recorded separately. Zero layouts means the absence of static coverage, not the absence of defects. A nested image inside a composite element is not counted as a separate violation without checking the accessible target as a whole.

An automatic traversal of the accessibility tree on a device can record visible nodes, name, role, state, value, bounds, available actions, traversal order, and tree changes after a step. It does not prove the quality of the name, the meaningful sequence, the actual announcement, the absence of a trap in all states, the accessibility of a gesture, or the success of a scenario. The tree is captured separately for each platform, configuration, and checked state.

Checks Static package Tree on the device Manual conclusion
V-06 explicit sizes in the markup target bounds measurement, exceptions, and actual operation are mandatory
V-07, V-08 fixed text sizes and orientation in the manifest comparison of state trees loss of content and essentiality are checked manually
V-03-V-05, V-09-V-11, V-13-V-16, V-18-V-22 does not confirm may give separate characteristics only manually, per the card

Machine characteristics are used to select elements that require checking. They may be incomplete for interfaces created by code and for composite elements. Automatic checking is not a mandatory part of version 0.1 of the protocol. The absence of a signal is not counted as a pass, and a found candidate receives a result only after manual confirmation.

V.7. Evidence and severity

The minimal record contains the object and version, the platform and device model, the configuration of V.4, the screen and state IDs, preconditions, action, expected and actual result, focus before and after, the accessible name and announcement, a screenshot, and a link to other evidence. For a semantics defect, the dump of the corresponding accessibility tree fragment is attached; for order and messages, an audio recording with a transcript; for a gesture, motion, switch, keyboard, orientation, and enlargement, a short recording of the sequence. If static analysis is used, the number of layouts found and decoded, the analysis errors, and the unverified resource references are additionally recorded. Secrets, one-time codes, the contents of personal fields, and extraneous notifications from other applications are not included in the evidence.

Severity is assigned per section 6 of part I by the observable consequence in the scenario, not by the origin of the signal. A blocking unreachable step of a critical scenario usually corresponds to S1 when no available workaround exists; an external or late workaround may correspond to S2; difficulty with reliable independent completion is S2 or S3 depending on the consequences. One shared defect is registered once with a list of platforms, screens, and states. A result on one platform does not reduce the severity of the defect on another.

V.8. Limitations and transition to the "ready" status

This section does not prove the accessibility of the platform itself, the hardware, an external service, a document, or a web resource. An emulator, static analysis, and the accessibility tree do not replace checking on a physical device with the actual output of the screen reader. Checking only one platform, orientation, role, or version limits the scope to the corresponding variant.

The status will remain "in progress" until all of the following conditions are met simultaneously:

  1. V-01-V-22 have been applied by at least two assessors to independent applications of two common mobile platforms, including a native and a hybrid application;
  2. critical scenarios, permission states, notifications, errors, interruptions, and returns have been checked, as well as the boundaries with sections A, B, and D;
  3. the manual layer has been passed in the full matrix of V.4 on physical devices, including both built-in screen readers, maximum text, magnifier, contrast or invert, both orientations, external keyboard, switches, and reduced motion;
  4. the consistency of the assessors' results has been measured and ambiguous protocol steps have been eliminated;
  5. the user layer has been conducted with people for whom the selected applications solve real tasks;
  6. the static analyzer and the tree traversal have been verified on an annotated reference set, and false positives, misses, and boundaries have been published;
  7. normative references, legally applicable profiles, and platform recommendations have been re-verified as of the release date.

Until the conditions are met, the protocol is issued and suitable for reproducible design review. The result must not be presented as a certification, universal conformance, or a completed validation of the section.

G. Desktop software

Section status: in progress.

What already applies from the general part. Purpose and principle, source versions, the scope passport, the three verification layers, evidence, defect severity, conclusion rules, and re-verification (Sections 1-8) apply unchanged. Below, the specification for desktop software is established. Procedures G-01 to G-29 are issued in the annex "Desktop software verification protocols".

Section grounds: EN 301 549 V4.1.1, where section 11 is devoted to non-web software, section 12 to information about products and services, and section 5.10 to authoring tools; WCAG2ICT as informative guidance on applying WCAG 2 to non-web documents and software; and national mappings where applicable. The Russian national example is GOST R 52872-2019, whose exact title explicitly includes applications for stationary and mobile devices and other user interfaces. WCAG2ICT is not an independent conformance standard and does not define a complete set of product requirements.

G.1. Object and unit of assessment

The object is a desktop application. The unit of assessment is the application in a specific version and build on a specific operating system and hardware architecture, in a fixed configuration, role, and state. The configuration includes the parameters from G.4. A role or state forms a separate unit of assessment if it changes the available functions, the interface, or the expected result. The result key contains the application identifier, version and build, platform and architecture, configuration identifier, role, and state.

The sampling unit is a window, dialog, mode, or complete user scenario inside the unit of assessment, not a URL. Different editors, workspaces, and modes form separate sampling units if they change the available functions or the interface. The result of card G-NN is recorded for the combination of the unit of assessment and the sampling unit; the result of one configuration, role, or state is not automatically carried over to another.

For an application that creates content, two independent properties are assessed: the accessibility of its own interface under Section G and the ability to create, check, save, and export an accessible result. The resulting file is separately verified under Section B. The presence of an export command or a built-in check does not prove the accessibility of the result. EN 301 549 V4.1.1, section 5.10 requires supporting the creation of an accessible result within the accessibility information supported by the output format; preserving such information on conversion where equivalent mechanisms exist; offering a remedy for a problem the tool is able to detect; and, where templates are provided, providing and labeling at least one template that supports the creation of accessible content.

Not included in the result of Section G are: the accessibility of the web version, of the mobile application, of the operating system as a separate product, of hardware, of installations without a user interface, of a network service in essence, and of the content of a created file. These objects are verified under the corresponding sections. Security, performance, and functional correctness are taken into account only where their violation creates an accessibility barrier.

G.2. Inventory

Before sampling, a register is compiled:

  1. of main and rare functions, roles, and launch paths;
  2. of main and auxiliary windows, modal and non-modal dialogs, messages, panels, menus, ribbons, tabs, and context areas;
  3. of viewing, editing, reading, presentation, collaboration, read-only, and recovery modes;
  4. of keyboard commands, single-character combinations, drag-and-drop operations, and actions requiring a pointer;
  5. of interface and accessibility settings, including contrast, colors, scaling, text, motion, and time;
  6. of help, documentation, compatibility information, and support channels;
  7. for editors: types of content created, built-in checks, templates, formats for opening, saving, and exporting.

For each item, the entry point, preconditions, expected result, available control methods, and link to the sampling unit are recorded. Omitting a window or mode from the inventory means a limitation of scope, not a "conforms" result.

G.3. Scenario sampling

The mandatory part includes: starting and finishing work; opening, creating, modifying, saving, and reopening data; navigation in the main window; menus, panels, and settings; at least one modal dialog; an error and recovery; notification of progress and completion of a long-running operation; help; every critical scenario from the scope passport. For an editor, the following are additionally included: creation of semantic structure, alternative text, a table, a language, running an accessibility check, and export followed by verification of the file under Section B.

The structured part includes the mandatory part, plus at least one instance of each identified combination of material attributes: role; editor or workspace; main type of the sampling unit; mode; critical or rare function; input method; format for opening, saving, or exporting. For calculating the random part, each unit is assigned exactly one main type: window, dialog, mode, or scenario. One instance may cover several other attributes, which is reflected in the coverage matrix.

Project proposal. The random part is added on top of the structured part separately for each main type. The group denominator is the number of structured units of that type; the group size equals 10 percent of the denominator, rounded up. If the structured part contains from 1 to 9 units of a type and at least one candidate of that type remains after exclusions, at least one random unit is selected. A group with a zero denominator is not formed, and the absence of a type in the structured part is verified as an inventory defect.

A reproducible selection is performed as follows:

  1. freeze the versions of the master register, the structured part, and the order of candidates before viewing the results;
  2. store for each unit a stable identifier and main type, then remove from the candidates the structured units and documented duplicates;
  3. record for each group the denominator, the calculated size, the pseudorandom selection algorithm, the seed, the date, and the encoding of the source list;
  4. sort the candidates by stable identifier, shuffle with the selected algorithm using the fixed seed, and select without replacement the first units in the calculated number;
  5. if there are fewer candidates than the calculated size, include the entire remainder and record the shortfall; if the remainder is empty, the random part of the group equals zero with the reason indicated;
  6. leave an inaccessible unit in scope with the status "not verified" and the reason, and take a replacement, if one is needed to fulfil the sampling size, from the next element of the same shuffled list; the replacement does not remove the limitation on the inaccessible unit;
  7. store the selected units, the inaccessible units, the replacements, and the excluded duplicates; if a new type, mode, function, format, or defect class is found, add an instance of it to the structured part, recalculate the affected groups, and repeat until stable.

Re-verification of the same unit of assessment uses the same frozen register, algorithm, and seed. A new register or a new version of the application receives a separate sampling version; the previous result is retained for comparison. These rules specify A.2-A.3, B.3, and D.3 for desktop software.

G.4. Configurations

Configurations are selected according to the real audience and the platforms supported by the product. A national or public-sector operating system is included when required by the applicable legal, procurement or contractual profile. For each configuration, the following are recorded: operating system and build, hardware architecture, application version and build, interface language, platform accessibility API, screen reader version, screen magnifier, high-contrast mode and system colors, system scaling, font size, reduced-motion mode, input method, and display parameters.

At minimum the following are verified: keyboard-only operation; operation with a screen reader corresponding to the platform; magnification with a screen magnifier; high contrast and system colors; enlarged system font and scale. For an application that edits text data and exposes text ranges, at least one configuration includes a braille display with routing keys. For other applications, braille output, alternative input, and additional key combinations are added where a confirmed audience or declared support exists. EN 301 549 V4.1.1, 11.5-11.7 links non-web software to documented platform accessibility services, prohibits unintentionally disrupting their operation, and requires taking user settings into account under the conditions established by the standard.

G.5. Basic set of control checks

No. Check Source Expected result
G-02 Virtualized content and objects outside the visible area EN 301 549 V4.1.1, 11.5.2.9-11.5.2.10, 11.5.2.13 Structure, text ranges, and focus position are available during scrolling and loading without loss of context.
G-03 Full keyboard operation and absence of traps EN 301 549 V4.1.1, 11.2.1.1-11.2.1.2, WCAG 2.2, 2.1.1, 2.1.2 All functions are available from the keyboard, focus can be moved out of every component, and a non-standard method is documented.
G-04 Visible order and focus stability EN 301 549 V4.1.1, 11.2.4.3, 11.2.4.7, 11.2.4.11, 11.5.2.13 Focus is visible, not completely obscured, follows the logical order, and is tracked programmatically. Project proposal: an update does not reset the working context without cause.
G-05 Menus, panels, ribbons, and complex widgets EN 301 549 V4.1.1, 11.2.1, 11.4.1.2, 11.5.2.5-11.5.2.17 Complex widgets are accessible from the keyboard and expose structure and states through the platform accessibility interface.
G-06 Dialogs and focus management EN 301 549 V4.1.1, 11.2.4.3, 11.5.2.13-11.5.2.15 The order and change of focus are programmatically determinable. Project proposal: on opening, focus lands in the expected place, stays in the modal dialog, and after closing returns to the logical context.
G-07 Notifications, states, errors, and progress EN 301 549 V4.1.1, 11.3.3.1-11.3.3.3, 11.4.1.3, 11.5.2.15, WCAG 2.2, 4.1.3 A significant change is announced programmatically without unnecessary focus movement; the error is named, associated with the object, and linked to a remedy.
G-08 System accessibility settings EN 301 549 V4.1.1, 11.6.2, 11.7 The application does not disrupt the platform accessibility functions and takes into account documented contrast, color, text, pointer, and cursor settings; reduced motion is checked if it is documented by the platform, otherwise as a project proposal.
G-09 Scaling and reflow without loss of functions EN 301 549 V4.1.1, 11.1.4.4, 11.1.4.10, WCAG 2.2, 1.4.4 Text is enlarged up to 200 percent or up to the platform limit, if it is lower. At a viewport width equivalent to 320 CSS px for vertical scrolling, or a height of 256 CSS px for horizontal scrolling, information and functions are preserved without scrolling in two dimensions, except for parts that require a two-dimensional layout for use or meaning.
G-10 Contrast and independence from color EN 301 549 V4.1.1, 11.1.4.1, 11.1.4.3, 11.1.4.11, WCAG 2.2, 1.4.1 Normal text has a contrast of at least 4.5:1, large text at least 3:1, and components, states, and graphical objects needed for distinction at least 3:1 against adjacent colors, taking the exceptions of WCAG 2.2 into account. Color is not the only means of conveying meaning. Project proposal: shape or position is likewise not used as the only indicator if without them the state is not programmatically determinable.
G-11 Hotkeys and conflicts EN 301 549 V4.1.1, 11.2.1.4, 11.6.2, WCAG 2.2, 2.1.4 Single-character commands can be disabled, remapped, or restricted to focus; a critical conflict with a platform or assistive technology command is eliminated by remapping, disabling, or an accessible alternative. Detectability and the full conflict matrix are project rules.
G-12 Alternative input, dragging, and target size EN 301 549 V4.1.1, 11.2.1.1, 11.2.5.7-11.2.5.8, WCAG 2.2, 2.5.7, 2.5.8 For dragging there is a keyboard method and a single-pointer method without dragging, except for the normative exceptions. The target has a size of at least 24 x 24 CSS px or satisfies one of the exceptions of EN 11.2.5.8. Project proposal: for actions that are not dragging, the method without precise movement is checked separately.
G-13 Time limits, motion, and flashes EN 301 549 V4.1.1, 11.2.2.1-11.2.2.2, 11.2.3.1 For a time limit, one of the conditions of EN 11.2.2.1 is fulfilled, motion and updating are stopped per EN 11.2.2.2, and the interface creates no more than three flashes in any one second or remains below the general and red flash thresholds.
G-14 Text input and editing with assistive technology EN 301 549 V4.1.1, 11.5.2.10, 11.5.2.13-11.5.2.14, 11.5.2.17 Text, format, caret, and selection are determinable; the user can read, select, and modify text in an accessible way. For an editor of text data, braille ranges and caret routing are checked in all editable areas.
G-15 Text editor EN 301 549 V4.1.1, 11.2.1.1, 11.5.2 Structure and actions are available from the keyboard and through the accessibility interface. The project scenarios cover table and object properties, search and navigation, spelling and grammar, comments, revisions, conflict resolution, forms, and protected areas.
G-16 Spreadsheet editor EN 301 549 V4.1.1, 11.2.1.1, 11.5.2.6-11.5.2.17 Structure and actions are available from the keyboard and through the accessibility interface. The project scenarios cover cell and range properties, user headers, formula tools, filter state, chart series and categories, source data, fields, hierarchies, totals, groups, filters, and pivot table slices.
G-17 Presentation editor EN 301 549 V4.1.1, 11.2.1.1, 11.5.2 Structure and actions are available from the keyboard and through the accessibility interface. The project scenarios cover the slide and object hierarchy, reading order, numeric and keyboard operations with geometry, notes, presentation, structural reading of tables and charts, and access to source data.
G-18 Documentation, help, and compatibility boundaries EN 301 549 V4.1.1, 12.1-12.3 If digital information or documentation is provided, it is available in its own format; the documentation on the product or the official page describes the accessibility features, their use, and how to obtain support. Project proposal: the owner publishes the supported configurations.
G-19 Error correction, undo, and confirmation EN 301 549 V4.1.1, 11.3.3.1-11.3.3.4, 11.3.3.7 The error is named and correctable. Information previously entered or provided to the user and requested again in the same process is filled in automatically or is available for selection, except in cases of significance of re-entry, security, or loss of currency. Project proposal: the remaining unchanged fields are preserved, and any action with a risk of data loss can be undone or confirmed before the result.
G-20 Built-in checking and hints EN 301 549 V4.1.1, 5.10.4 If the tool is able to detect a problem, it offers the author a way to correct it. Project proposal: the check distinguishes manual items and allows accessible navigation to the object and the correction.
G-21 Semantic structure, language, and links EN 301 549 V4.1.1, 5.10.1-5.10.3 The author can assign headings, lists, order, the language of the document and of a fragment, the link purpose, and the link text.
G-22 Alternative text and complex graphics EN 301 549 V4.1.1, 5.10.1-5.10.3 The author can set an alternative or a decorative status; for a complex object, an extended description or data is available.
G-23 Table semantics EN 301 549 V4.1.1, 5.10.1-5.10.4 The author can assign headers and cell relationships. If the tool is able to detect a problematic structure, it offers a remedy; a mandatory warning about a complex table is a project proposal.
G-24 Content contrast and accessible templates EN 301 549 V4.1.1, 5.10.4-5.10.5 If the tool is able to detect insufficient contrast, it offers a remedy; if it provides templates, at least one template that supports the creation of accessible content is available and labeled. Verification of the created file under Section B is a project rule.
G-25 Multimedia alternatives EN 301 549 V4.1.1, 5.10.1-5.10.3 The author can add, change, check, and play back subtitles, transcripts, and audio descriptions where they apply.
G-26 Preserving semantics on opening and re-saving EN 301 549 V4.1.1, 5.10.3 In a restructuring or re-encoding conversion, the supported accessibility information is preserved if the output technology has equivalent mechanisms.
G-27 Export of an accessible result EN 301 549 V4.1.1, 5.10.3-5.10.4 The export preserves the supported accessibility information where equivalent mechanisms exist, and a detectable problem receives a remedy. The explicit PDF/UA-1 profile, an accessible check before a potentially destructive export, and passing the applicable checks of Section B are project rules.
G-28 Accessibility regression control Section 8 of Part I Critical components, scenarios, and exported files are re-verified after a change. This is a project proposal, not a separate requirement of the standard.
G-29 Built-in read aloud EN 301 549 V4.1.1, 11.2.1.1, 11.5.2, 11.6.2, 12.3 If the feature is claimed, it is available from the keyboard, documented, and does not break the documented platform accessibility features. Project proposal: the feature announces the reading position. The mere presence of the feature is not in itself a universal requirement.

G.6. Automatization map

Object of inspection Permissible automated inspection Mandatory manual part
Keyboard and focus Recording the event sequence as evidence Performing every function, absence of a trap, order and return logic
Color, contrast, scaling Measuring pixels and window parameters System modes, overlaps, preservation of functions and meaning
Notifications Dumping accessibility interface events Timeliness, comprehensibility, absence of superfluous focus
Produced file Feature-specific analyzer of section B Creation, correction, export, and full verification of the result under section B
Documentation Search for the stated information Format accessibility, accuracy of instructions, and their execution

Inspection of the accessibility tree confirms only the observable software representation. It does not replace manual work with the keyboard, a screen reader, a screen magnifier, and system modes. As of the version date, the project service does not provide a separate automated check for desktop software (section 10).

G.7. Evidence and severity

The minimum defect card contains the configuration, the sampling unit, the steps, the expected and actual result, the severity under section 6, and an artifact. A visual defect requires a screenshot of the window with the area marked; for semantics, a dump of the corresponding accessibility tree fragment; for the sequence, a recording of the scenario with the keyboard or a screen reader with a textual transcript of the key events. For a produced result, the source and exported files and the protocols of section B are attached. One artifact may supplement another, but does not replace reproducible steps.

A defect is usually blocking when it prevents completing a critical scenario in the stated configuration and has no available workaround. A defect is usually substantial when it deprives the user of part of the information or requires a disproportionate workaround. The final severity is established only by the rules of section 6, taking into account frequency, coverage, workaround, and consequences.

G.8. Limitations and transition to the "ready" status

This section does not prove compatibility with all combinations of platforms and assistive technologies. The result is limited to the versions, builds, configurations, and sampling listed in the passport. When assessing office suites, the following are checked: virtualized content, alternative input, specialized editor scenarios, collaboration, creation of multimedia alternatives, the re-save cycle, and regression control.

The "ready" status is possible after the trial of G-01-G-29 on at least two substantially different desktop applications, including a content authoring tool, on two operating system families; where a national profile requires a specific operating system, that system is included as an additional applicable configuration; independent reproduction of the sampling and the severity; verification of the produced files under section B; elimination of contradictions in the protocol; and publication of an anonymized trial report. Until then, this section has the status "in progress".

D. Digital services and processes with authorization

Status of the section: in progress.

What already applies from the general part. The purpose and principle, the area passport, the three inspection layers, evidence, defect severity, the rules for issuing a conclusion, and re-verification (sections 1, 3-8) apply without change. This section develops the approach of part I to full processes (section A.4), including cases where a process continues beyond a single resource type.

The templates of section 9 apply to a process with the page address replaced by the process and variant identifier, role, step, and state; the list of fields for the path and evidence record is specified in D.7. The procedures for the control checks D-01-D-20 are published in the appendix "Protocols for testing digital services". The base set defines the process inspection route, but does not replace the full applicable requirements profile for each site, document, application, or other resource type included in the process.

D.1. Object and unit of assessment

The object of this section is an end-to-end digital process: sign-in, access recovery, service arrangement, application submission, upload or receipt of a document, consent, payment, or any other sequence that takes the user from the declared entry point to the result. The process is considered as a whole, independently of the number of domains, providers, and resource types.

The unit of assessment is one process in a fixed role and path variant. It includes the initial point, all steps and states of the main sequence, the result, and the available way to return or continue. Success, error, correction, cancellation, going back, session expiry, extension, re-authentication, and recovery after interruption are separate variants, if they exist. The roles of applicant, representative, recipient, payer, and employee form different units if they change the available steps, data, or result.

The process boundary is not cut off when moving between a website, an electronic document, a mobile application, a message with a one-time code, or a notification through another channel. For each transition, the following are recorded: the initial state, the channel, the action, the expected target, the actual state, preservation of context, and the way back. WCAG 2.2 Conformance Requirement 3 extends the conformance claim to all web pages of the full process, and not only to the chosen successful step.

The following are not covered by this section: substantive identity verification of the person, the reliability of account protection, the correctness of accrual, the legal force of an application, the accessibility of staff work in the internal system, and the accessibility of the physical service point. These properties may be verified separately. Each included website is also checked under section A, each received or uploaded document under section B, and an application or device under the corresponding section of part II. A positive result for one resource type does not transfer to the entire process.

D.2. Inventory and criticality of processes

Before sampling, a master register of the resource's processes is created. Sources for the listing: the service catalog, navigation and the site map, public help, support scenarios, owner schemas, the test bench, and permitted test roles. For each process, the following are stored: a stable ID, the name and result, the entry point, roles, whether authorization is mandatory, steps, resource types and channels, result variants, legal or financial consequences, documents used, timeouts, the process owner, and the version and source of the information.

The following are necessarily identified, if they exist:

  1. sign-in, sign-out, re-authentication, access recovery, and identity confirmation;
  2. search and selection of a service, arrangement, draft saving, submission, and cancellation of an application;
  3. consents, permissions, and change of significant settings;
  4. payment, refund, confirmation, and receipt of an operation document;
  5. upload, replacement, deletion, and receipt of documents;
  6. notifications, status tracking, correction, and appeal of the result;
  7. transitions to another website, application, document, or confirmation channel.

Project proposal. A process is considered critical if, without it, the user cannot obtain the main service, complete a mandatory action, dispose of money or data, recover access, or obtain the result within the established deadline. The owner provides a list of critical processes, but the assessor matches it against the master register and records the omissions. A process cannot be excluded only because it is hard to reproduce: in the absence of permitted conditions, it remains in scope with the status "not verified".

D.3. Sampling of processes and states

The structured sample includes all critical processes and at least one process for each identified result, role, authentication method, resource type, confirmation channel, document type, and class of legal or financial consequence. One process may cover several characteristics, which is reflected in the coverage matrix.

The random part is added on top of the structured one according to the rules of A.3: 10% of the number of structured units, rounded up, and at least one unit when a remainder exists, are a project proposal. Before selection, the version of the master register, excluded duplicates, the order of the list, the algorithm, and the seed are recorded. The selection is performed without replacement. If the random part revealed a new process type, state, or transition, it is added to the structured part, and the size of the random part is recalculated.

The selected process is checked as a whole, and not by a single screen. All web pages of the main sequence are included according to WCAG-EM step 3.c and all content of each page according to WCAG 2.2 Conformance Requirement 2. Project proposal. For each applicable process, the successful path, empty required fields, one invalid value, correction, going back, cancellation, session expiry, extension, re-authentication, recovery, and the final confirmation are checked. If a state cannot be safely reproduced, it is not imitated on live data, but is marked "not verified" with a reason and an agreed way of future verification.

D.4. Test conditions and configurations

A test with authorization is conducted only under a scenario agreed in advance with the resource owner. The owner provides a test bench or permission for a safe test, test accounts for the required roles, anonymized data, a way to reset the state, expected responses, the boundaries of permissible actions, and a responsible person. Passwords are not passed to assessors in evidence; secrets, one-time codes, addresses, and identifiers are masked.

Real personal data, live payment details, and data of other people's accounts are not entered. Real payments, submission of applications, acceptance of consents, modification of other people's data, and other actions with external consequences are not performed. For such steps, an isolated bench, a cancellable test operation, or a stop before the action with the status "not verified" is used. The owner's approval does not turn a real operation into a safe test one.

The configuration matrix is formed according to A.6 without fixing program names. Project proposal. It includes at minimum a current desktop browser with keyboard, a desktop configuration with a screen reader, a mobile configuration with a built-in screen reader, magnification and a narrow viewport, system colors or high contrast, and an alternative input method for a process with a complex gesture. If the process moves between devices or channels, the whole supported combination is checked, and not each component separately.

For each run, the following are recorded: the process version, the address or entry method, the role, the test set, the OS and version, the browser or application and version, the assistive technology and version, the language, the viewport and orientation, the scaling, the input method, system settings, the network state, the start time, the session duration, the result, and the evidence. Success of one configuration does not prove success of the others.

D.5. Base set of control checks

№ Check Source Expected result
D-02 Keyboard operation WCAG 2.1.1, 2.1.2 Every step, control, and exit is available from the keyboard without a trap
D-03 Order and transfer of focus WCAG 2.4.3, 4.1.2 The focus follows the logical sequence, after a transition or an error it lands in a predictable place, and the name, role, and state are programmatically available
D-04 Steps, going back, and cancellation WCAG 2.4.6, Making Content Usable: make each step clear, let users go back, project proposal The user understands the current step and the remaining path, and can go back or cancel the action without undeclared loss of what was entered
D-05 Labels and instructions WCAG 3.3.2, 2.5.3 Before input, the purpose of the field, the format, the mandatory nature, and the constraints are clear
D-06 Detection and description of errors WCAG 3.3.1, 3.3.3 The erroneous field is identified, the error is described in text, and a known correction is suggested, unless this would create a security or purpose risk
D-07 Status messages WCAG 4.1.3 Loading, saving, filtering, sending, success, and error are announced to assistive technology without mandatory focus transfer
D-08 Time limit and extension WCAG 2.2.1 An applicable time limit can be turned off, adjusted, or extended after timely warning, taking the criterion's exceptions into account
D-09 Loss warning and preservation of entered data WCAG 2.2.5 (level AAA), Project proposal Before logout, cancellation, or session expiry the user is warned; after signing in again the data is preserved if this is permissible for security reasons
D-10 No re-entry WCAG 3.3.7 Information already entered in the same process is filled in automatically or available for selection, except for essential, protective, and expired cases
D-11 Accessible authentication WCAG 3.3.8 Sign-in does not require a cognitive function test, or a criterion-compliant alternative, an assistance mechanism, or a permitted exception is available
D-12 CAPTCHA and alternatives WCAG 1.1.1, 3.3.8 The purpose of the non-text check is described, alternatives for different types of perception are available, and authentication has a path without an impermissible cognitive test
D-13 One-time codes WCAG 3.3.8, Understanding 3.3.8 The code can be pasted or transferred by a supported mechanism; field separation, the countdown, and re-issuance do not create an inaccessible test
D-14 Re-authentication and continuation WCAG 2.2.5 (level AAA), Project proposal After re-authentication the user returns to an understandable state and continues without re-entering data, if this is permissible for security reasons
D-15 Prevention of significant errors WCAG 3.3.4 A legal, financial, or data-modifying action is reversible, is checked before completion, or has a review, confirmation, and correction step
D-16 Consistent help WCAG 3.2.6 A repeated help mechanism is placed in the same relative order, if the pages of the process contain it
D-17 File upload WCAG 3.3.2, 3.3.1, 4.1.3, 2.5.7, Project proposal Before selection the format, size, and number of files are known; selection, replacement, and deletion are available; progress, success, and error are announced
D-18 Result in an accessible form WCAG 2.4.4, Section B, Project proposal The result is found, unambiguously named, and available within the same process; the document is checked against Section B or an equivalent accessible version is provided
D-19 Transition between resources and return WCAG 2.2, conformance requirement 3, Project proposal The user is warned about a change of resource or channel, context and data are not lost, and re-entry and return are available

| D-20 | Notifications and confirmations through other channels | WCAG 1.3.3, 1.4.1, Project proposal | The channel, deadline, and purpose of the message are clear in advance; meaning is not conveyed by color, position, or a single type of perception alone; there is a safe way to repeat the action |

D.6. Automatization map

The automated layer applies to every web step under Section A and can confirm an individual programmatic feature or violation. It does not establish that the process as a whole is clear, securely completed, and accessible in all channels. The assessor assigns the "pass" status for D-01-D-20 after the full route.

Check Automatization What the machine checks
D-02 partially Replays the key sequence and captures the active element; reachability, visibility, and absence of a trap are confirmed manually
D-03 partially Records the focus sequence, programmatic name, role, and state; meaning and the expected transfer point are assessed manually
D-04 manual only The clarity of the indicator, of return, of cancellation, and of the consequences of loss require a semantic assessment
D-05 partially Finds programmatic labels, requiredness, and relationships of instructions; clarity and sufficiency are checked manually
D-06 partially Triggers permitted test errors and captures the relationship between the message and the field; the accuracy of the description and of hints are checked manually
D-07 partially Captures the appearance of text and of programmatic message regions; timeliness and clarity are confirmed manually with assistive technology
D-08 partially Detects explicit counters and warnings and can measure time on the test bench; exceptions, extension, and recovery are checked manually
D-09 manual only The risk of loss, the sufficiency of the warning, and the permissibility of preservation depend on the process and on security
D-10 partially Compares repeated fields and values in the test route without recording content; the applicability of exceptions is assessed manually
D-11 manual only The assessor determines the cognitive load, the permitted exception, and the equivalence of the alternative
D-12 manual only The presence of a widget does not prove the accessibility of the CAPTCHA and of the alternatives for different limitations
D-13 partially Checks the possibility of pasting and the programmatic properties of the fields on the test bench; obtaining the code, the deadline, and clarity require a manual run
D-14 manual only Continuation after re-authentication is checked on a permitted test account
D-15 manual only The legal or financial consequence, reversibility, and sufficiency of confirmation are assessed against the process
D-16 partially Matches the position and purpose of repeated help links; their operability and relative order are confirmed manually
D-17 partially Checks the programmatic properties of the field, the constraints, and the messages on a permitted file; the actual upload is performed only on the test bench
D-18 partially Checks the presence and purpose of the link to the result; the accessibility of the content is established under Section B and manually
D-19 partially Records domains, resource types, redirects, and return; preservation of context and clarity of the transition are checked manually
D-20 manual only Delivery, accessibility of the other channel, the deadline, and the possibility of repetition cannot be derived from a web page

D.7. Evidence and severity

A path record contains the process and variant IDs, the role, the configuration, the test set, the step number, the initial state, the action, the expected and the actual result, the address or channel, the focus before and after, the programmatic message, the time, a snapshot, and a link to other evidence. Secrets and the contents of personal fields are not recorded. For a transition between resources, evidence is attached from both sides; for a timeout, the warning time, the extension action, and the state after it; for an error, the entered anonymized value class, the relationship of the field and the message, the correction, and the outcome.

Severity is assigned according to Section 6 of Part I, taking the result of the whole process into account:

  • S1: the target group cannot complete a critical process, a mandatory step is unavailable, data or money are irreversibly lost, and no accessible workaround exists;
  • S2: completion is possible only through a difficult, external, or untimely workaround, or a significant branch of the process is unavailable;
  • S3: an individual step, message, or return is hindered, but the process can be reliably completed without outside help;
  • S4: limited inconvenience does not change the outcome and does not create substantial risk.

An unavailable mandatory step blocks any conclusion that the whole process passes, even if the remaining screens are accessible. A single defect in a shared component is recorded once with a list of the affected steps and processes, but severity is determined by the most severe confirmed consequence. The availability of telephone support does not automatically reduce severity if the help is not equivalent in deadline, cost, independence, and outcome.

D.8. Limitations and transition to the "ready" status

This section does not prove security, legal correctness, accuracy of calculation, delivery of a message by a telecom operator, or the accessibility of an employee's internal work. A public route without authorization allows only the applicable external steps and the operability of the protocol to be checked, but does not replace the checking of sign-in, one-time codes, session expiry, re-authentication, payment, and final submission on a permitted test bench.

During the assessment, the accessible name of the field, the start of the sequential navigation, the preservation of focus after filtering, and the programmatic message about a dynamic result are recorded (D-02, D-03, D-05, and D-07). Checking only the public steps without authorization does not confirm the accessibility of the closed part of the process.

The status remains "in progress" until all of the following conditions are met simultaneously:

  1. D-01-D-20 are applied by at least two assessors to independent processes of sign-in, recovery, service, document upload, and a legally or financially significant action on permitted test benches;
  2. successful, erroneous, cancelled, expired, and recovered states are checked, as well as transitions between resource types and channels;
  3. the consistency of results is measured and ambiguous protocol steps are eliminated;
  4. the manual layer is passed in the configuration matrix of D.4, including a screen reader program and an alternative input method;
  5. the user layer is conducted with people for whom these processes are a real task;
  6. the automated scenario is verified on annotated states, and its gaps and false conclusions are published;
  7. normative references and legally applicable profiles are re-checked as of the issue date.

Until the conditions are met, the protocol is issued and suitable for a reproducible design review. The result must not be presented as a certification, universal conformance, or a completed trial of the section.

E. Self-service terminals and kiosks

Section status: in progress: the test protocols are issued (annex, version 0.1), the trial is not complete.

What already applies from the general part. The purpose and the principle, the source versions, the scope passport, the three verification layers, the evidence, the defect severity, the conclusion rules, and the re-verification (Sections 1-8) apply without change. Below, the terminal specification is established. The procedures E-01-E-22 are issued in the annex "Test protocols for self-service terminals and kiosks".

E.1. Object, assessment unit, and sample unit

Object of the section - a terminal or kiosk installed in a public place with closed functionality: a touchscreen, physical keys, card and document readers, a receipt printer, audio output, and the space accessible to the user. The object includes the enclosure, the fixed controls and dispensing mechanisms, the screen interface, the built-in accessibility features, and the documented help channel. The accessibility of the premises and of the route to the terminal is recorded as context and a limitation, but is not presented as the result of testing the digital interface or the equipment.

Assessment unit - one physical instance of a terminal at a specific installation site, with the model and physical configuration recorded, the screen interface version, the set of connected devices, the language, and the service mode. Two identical enclosures with a different installation height, interface version, or set of peripherals are different assessment units.

Sample unit - one operation in one essential state on a specific assessment unit: for example, start without authorization, service selection, input, error, waiting, cancellation, time expiry, confirmation, printing, card return, or contact with the operator. One result of each applicable card E-01-E-22 is linked to a sample unit. The object, the assessment unit, and the sample unit are not interchangeable.

The screen interface is tested under the applicable protocols V and G, which the cards E reference, with the terminal specifics: the user does not install programs and does not change equipment, the screen and the controls are fixed, the audio output may be heard by those nearby, and the physical actions and the return of the medium form a single process. There is no need to restate the general screen criteria in Section E, but the reference does not release one from recording the result on the terminal.

E.2. Inventory and freezing of the registry

Before sampling, a register of operations, states and physical parts is compiled. Discovery sources are recorded separately:

  1. the list of the owner's services and configurations;
  2. a full walkthrough of the start screen, help menu, languages and accessibility features;
  3. operation, installation and peripheral documentation;
  4. observation of all connectors, keys, readers, trays, speakers and approach zones;
  5. an anonymized list of the owner's frequent and critical operations, if provided.

A register entry contains a stable operation identifier, type, start and end, mandatory states, physical parts used, whether there is a financial or legal consequence, and the discovery source. Duplicates are merged only when the purpose, sequence, states, interface version and peripherals match; all original identifiers are preserved. The same label with a different sequence or outcome is not a duplicate.

Before selection the register is frozen: the version number, date and time, the assessment unit identifier, the number of entries, the export file and its SHA-256 are saved. After the freeze, new operations receive a new register version. The sample, the results and the report always reference one version.

E.3. Selection of operations and states

Operations are distributed by type: obtaining information, identification and login, selection and input, card or document reading, financially or legally significant confirmation, printing or issuing, help, cancellation and recovery. The denominator of each type equals the number of unique operations of that type in the frozen register.

The structured part includes all critical operations and at least one operation of each non-empty type. For each selected operation, the reachable states are included: initial, success, error, waiting, cancellation, timeout, return to start and peripheral failure. A state that cannot be safely triggered remains in the denominator and receives "not tested" rather than disappearing from the sample.

If candidates of a type remain after the structured selection, one random candidate is added. The algorithm is reproducible:

  1. identifiers are normalized as exact register strings and sorted in ascending order of Unicode code points;
  2. the seed is written as terminal:<register-version>:<assessment-unit-identifier>:<assessment-date>;
  3. for each candidate, the SHA-256 of the sequence seed, LF (U+000A), the identifier encoded in UTF-8 without BOM is computed; the test vector terminal:v1:unit-01:2026-09-24 + LF + operation-01 gives 0540dd05aaf10931680b2db3c6e1057af23b17f6aca4137bf63c21d0aacff1ca;
  4. candidates are sorted by hash, then by identifier, and the first one is taken without replacement.

If there are fewer candidates than required, all of them are taken and the plan, the fact and the reason for the shortage are published. An unavailable selected operation is not silently replaced: it is assigned "not tested", then the next candidate of the same type is taken according to the already computed order as an additional unit. The original denominator and the coverage share are not recalculated. Manual addition of an operation is allowed only in addition to the sample, with a reason, and does not change the random result.

E.4. Test conditions and safety

Testing is performed in person on a live terminal or on a demonstration unit that reproduces the enclosure, installation height, peripherals and software version. The owner is provided in advance with the list of operations, measurements, test media, permissible errors and actions with the operator. Permission and restrictions are recorded in the area passport.

Real payments, contracts, applications and personal data are not used. Fictitious data, test cards and documents, a demonstration account or a stop before irreversible submission are applied. Testing must not block the terminal, interfere with other users, open the enclosure, change the owner's settings, or disclose the sound and data of a real visitor. A condition that cannot be fulfilled safely receives "not tested" with a reason.

E.5. Configurations

The minimal matrix contains five user configurations without binding to device brands and software:

  1. a wheelchair user: frontal and, if provided, lateral approach, reach, screen overview, space for knees and feet;
  2. a blind user with personal headphones: start-up without sight, speech, physical parts, input, errors, privacy and result;
  3. a low-vision user: text size and contrast, distinguishability of focus and controls, glare, magnification if available;
  4. a user with motor impairments: single pointer, no mandatory precise gesture, grip, wrist rotation and simultaneous actions, permissible force;
  5. a user with a hearing impairment: visual equivalents of sound, operation status, warnings and communication with support without mandatory speech.

For each configuration, the assessment unit, interface version, language, input method, accessibility features, headphones where applicable, approach position, lighting, background noise and all deviations from the normal mode are recorded. An assessor's measurement does not replace the user layer with people for whom the operation is a real task.

E.6. Basic set and automability map

The basic set is limited to twenty-two cards and covers equipment, closed functionality and the on-screen interface:

Code Check Main basis and linkage
E-02 Completeness of the speech output EN 301 549, 5.1.3.1, 5.1.3.3, 5.1.3.6, 5.1.3.7, 5.1.3.15; linkage with V-21
E-03 Speech control, language and absence of overlap EN 301 549, 5.1.3.4, 5.1.3.5, 5.1.3.10, 5.1.3.14
E-04 Delivery of the speech output and privacy EN 301 549, 5.1.3.2, 5.1.3.8, 5.1.3.9
E-05 Volume and reset of the level EN 301 549, 5.1.3.11-5.1.3.13
E-06 Visual equivalent of sound EN 301 549, 5.1.5 and 11.1.4.2; linkage with V-21
E-07 Tactile distinguishability, key states and medium orientation EN 301 549, 5.5, 5.6.1, 8.4.1, 8.4.3, 8.5
E-08 Key repeat and non-simultaneous actions EN 301 549, 5.7-5.9
E-09 Operation without precise grip, rotation or high force EN 301 549, 8.4.2
E-10 Text, contrast and colour EN 301 549, 5.1.4, 5.6.2, 8.7, 8.8, 11.1.3.4, 11.1.4; linkage with V-07, V-08, V-10 and G-10
E-11 Keyboard operation, focus and closed interface EN 301 549, 5.1.6, 11.1.3.2 and 11.2.1, linkage with V-02, G-03 and G-04
E-12 Gestures, cancellation, movement and target size EN 301 549, 11.2.5, linkage with V-03, V-04, V-05, V-06 and V-09
E-13 Time limits, movement and flashing EN 301 549, 11.2.2 and 11.2.3.1, linkage with V-14
E-14 Biometrics, speech input and accessible authentication EN 301 549, 5.1.7, 5.3, 11.3.3.8; linkage with V-15
E-15 Labels, input purpose, errors, re-entry and confirmation EN 301 549, 5.1.8, 11.1.3.5, 11.2.4.6, 11.3.3.1-11.3.3.4, 11.3.3.7; linkage with V-12, V-13 and V-22
E-16 Screen height, information and controls EN 301 549, 8.3.2, 8.3.3, 8.3.5, 8.3.6, 8.3.14
E-17 Reach depth and free space EN 301 549, 8.3.7, 8.3.9-8.3.11
E-18 Media, receipt and result confirmation EN 301 549, 5.1.3.16, 5.4 and 8.4.3
E-19 Operation status and multichannel notification EN 301 549, 5.1.3.15, 5.1.5, 11.4.1.3, linkage with V-11 and G-07
E-20 Help, contact with the operator and information about accessibility features EN 301 549, 12.3; linkage with the applicable resource type
E-21 Telephone handset and standard connections EN 301 549, 8.2 and 8.6
E-22 Software accessibility of the open part of the interface EN 301 549, 11.4.1.2 and 11.5.2.5-11.5.2.17, linkage with G-01

As of the version date, all twenty-two cards have the "manual" mode. Meters for size, distance, force, sound and contrast give separate values, and screen and event recording helps gather evidence, but no instrument determines applicability, performs a physical operation or assigns the final status. There is no automatic testing of terminals in the project's service.

For closed parts, the requirements of clause 5.1 of EN 301 549 are the priority. The clauses of section 11 apply to the on-screen software interface within its own preconditions; the open part is not excluded only because it is in the same enclosure. The basic set defines a route for frequent checking and does not confirm a complete legally applicable profile.

E.7. Evidence, statuses and severity

This appendix establishes one rule for the final status of a composite card: a violation takes priority over "not tested", which in turn takes priority over a pass; NA is permissible only when all conditions are NA. Internal sub-results are preserved, but the card does not receive several competing final outcomes.

The minimal evidence package includes the assessment unit passport, the frozen register and the sample, photographs of the installation with measurement points, values and measurement units, a recording of the on-screen and speech path, the state log, the configuration, test data without personal values, the actual result and the defect identifier. For a violation, the specific card condition and the normative clause are given; a project proposal is not recorded as a normative violation.

Severity is assigned according to section 6 of part I, taking the whole operation into account:

  • S1: the user cannot complete a critical operation independently, protected information is disclosed, the medium is not returned, or a risk of money or safety arises without an accessible workaround;
  • S2: the operation is completed only with substantial assistance from another person, an external channel or an unequal workaround;
  • S3: an individual step, message or physical action is substantially impeded, but the operation is reliably completed;
  • S4: limited inconvenience does not change the outcome and does not create a material risk.

E.8. Limitations and transition to the "ready" status

The check does not prove the safety of the equipment, the correctness of the calculations, the conformity of the building to construction requirements, the accessibility of the operator's internal work, or the legal validity of the operation. The demonstration terminal confirms the protocol only within the reproduced enclosure, installation, peripherals, version, and states. A completeness check based on source texts without a physical device is not a user-based evaluation.

The status remains "in progress" until all of the following conditions are met simultaneously:

  1. E-01 to E-22 are applied by at least two evaluators to at least two independent models of stationary terminals and to all provided operation types;
  2. successful, error, cancelled, expired, and restored states, peripheral failure, media return, and support contact are checked;
  3. physical dimensions, force, volume, contrast, and time are measured repeatedly, and evaluator discrepancies are resolved;
  4. the manual layer is passed in all configurations of E.5;
  5. the user layer is conducted with people for whom the selected operations are a real task;
  6. the denominators, the actual sample, the unchecked states, and the limitations of the demonstration or operational test bench are published;
  7. the normative references and the legally applicable profiles are rechecked as of the issue date.

Until the conditions are met, the protocol is issued and fit for reproducible design review. The result must not be presented as certification, universal conformance, or a completed evaluation of the section.

Main sources

  1. W3C, Web Content Accessibility Guidelines (WCAG) 2.2, Recommendation of 12 December 2024.
  2. W3C, WCAG 2.2 Conformance Requirements.
  3. W3C, WCAG Evaluation Methodology 2.0, Group Note of 23 July 2026.
  4. W3C, Website Accessibility Conformance Evaluation Methodology 1.0, earlier Working Group Note of 10 July 2014.
  5. W3C, Making Content Usable for People with Cognitive and Learning Disabilities, Working Group Note of 29 April 2021.
  6. W3C WAI, Involving Users in Evaluating Web Accessibility.
  7. W3C, Accessibility Conformance Testing Rules Format 1.1, Recommendation of 5 February 2026.
  8. W3C WAI, ARIA Authoring Practices Guide.
  9. W3C WAI, Tutorials.
  10. Government of the Russian Federation, Resolution No. 102 of 07.02.2026, effective from 01.03.2026 within the scope it establishes.
  11. Rosstandart, official cards GOST R 52872-2019, GOST R 70176-2022, and GOST R ISO 21801-1-2022.
  12. Ministry of Labour of the Russian Federation, Order No. 647n of 14.11.2025, entering into force on 01.09.2027.
  13. Estonia, TTJA, "Eesti avaliku sektori digiligipääsetavus 2024. aastal", annual report on the digital accessibility of the public sector.
  14. Estonia, TTJA, overview of regular accessibility monitoring.
  15. Norway, Uutilsynet, "Slik fører vi tilsyn med nettsteder og apper".
  16. Norway, Uutilsynet, "Forenklet kontroll og statusmåling med nettsteder".
  17. United Kingdom, GOV.UK, "Accessibility monitoring of public sector websites and mobile apps from 2022 to 2024".
  18. France, DINUM, RGAA portal, "Évaluation de la conformité".
  19. France, DINUM, RGAA portal, "Déclaration d'accessibilité".
  20. Germany, BIK BITV-Test (DIAS GmbH, BIK Prüfverbund network), "Beschreibung des Prüfverfahrens".
  21. Denmark, Digitaliseringsstyrelsen, "Monitorering og tilsyn".
  22. Denmark, Digitaliseringsstyrelsen, "Overordnede resultater for 2024".
  23. Spain, Observatorio de Accesibilidad Web, accessibility monitoring report for websites.
  24. Finland, Act on the Provision of Digital Services 306/2019.
  25. Finland, Traficom, inspection procedure for compliance with the Act on Digital Services.
  26. ETSI, EN 301 549 V4.1.1 "Accessibility requirements for ICT products and services", sections 5, 8, 10-12.
  27. W3C, "Guidance on Applying WCAG 2 to Non-Web Information and Communications Technologies (WCAG2ICT)".
  28. ISO, ISO 14289-1:2014 (PDF/UA-1).
  29. GOST R 70176-2022 "Internet resources and other information presented in electronic digital form. PDF format files. Accessibility requirements for people with disabilities and other persons with limited life activity".
  30. Bank of Russia, information letter No. IN-01-59/184 of 29.12.2020 with recommendations on the accessibility of remote digital channels.
  31. W3C, Accessibility Maturity Model, Group Note of 4 November 2025.
  32. W3C WAI, Accessibility Roles and Responsibilities Mapping, draft resource on the distribution of roles and responsibilities, version of 24 July 2025.

The links and source statuses were checked as of the version date of the document. During a repeated audit they should be checked anew.

Mapping to the numbering of edition 0.2

The table maps the numbering of edition 0.2 (version of 11 September 2026, the document "Site Accessibility Evaluation Methodology") to the numbering of this edition, so that earlier references to a specific section can be found in the new structure. Sections B, V, G, D, and E were absent in edition 0.2: their content is new, except that section D develops the material of previous section 4.4. The earlier section 4.4 is retained as section A.4, while section D provides the new cross-resource process detail.

Section of edition 0.2 Section of this edition
1.1. Scope of application 1.1. Scope of application
1.2. Terms and definitions 1.2. Terms and definitions
1.3. Normative references 1.3. Normative references
2. Versions of methodological sources 2. Source versions and national profiles
2.1. Russian profile 2.1. Russian profile
3. Evaluation scope passport 3. Evaluation scope passport
4. Product research and sampling A (introductory part of the section "Websites and web applications")
4.1. Inventory A.1. Inventory
4.2. Structured sampling A.2. Structured sampling
4.3. Random sampling A.3. Random sampling
4.4. Complete processes and states A.4. Complete processes and states (section D - a new development of the topic beyond a single resource)
4.5. When the whole website is checked A.5. When the whole website is checked
5. Configurations and assistive technologies A.6. Configurations and assistive technologies
6. Three layers of checking 4. Three layers of checking
6.1. Automated checks 4.1. Automated checks
6.2. Manual checking against the standard 4.2. Manual checking against the standard
6.3. Checking with users 4.3. Checking with users
7. Basic set of 34 control checks A.7. Basic set of 34 control checks
8. Evidence and recording of the result 5. Evidence and recording of the result
9. Defect severity 6. Defect severity
10. Conclusion rules 7. Conclusion rules
11. Repeat check and regression 8. Repeat check and regression
12. Templates 9. Templates
12.1-12.6 (six templates) 9.1-9.6 (the same sequence of templates)
13. Main sources Main sources (unnumbered, after Part II)
- 10. Section readiness table (new)
- B, V, G, D, E (new resource-type sections)