Practical guide / 2026-09-09

Automated signal is a start. Human review makes the next test clear.

A short practical guide for teams deciding what to review, what evidence to record and what a result can—and cannot—establish.

By Virtually(Creative) Accessibility. Resource by Virtually(Creative). Reviewed September 9, 2026. This resource is educational, not legal advice or a conformance certification.

1 / Scope

Start with the task, not a dashboard.

Choose the page, journey and state where a missed barrier would matter. A checkout, signup, account or public-service flow may include dialogs, errors, validation, loading, confirmation and responsive states that a single URL does not describe.

Record the environment, access assumptions, browser, assistive technology and standard or technical target requested. Separate responsive web, native apps and documents rather than treating them as one surface.

2 / Evaluate

Use tools for signal. Use people for context.

Automated checks can find some detectable patterns quickly. They cannot by themselves explain whether a person can understand, operate or complete the task. Add keyboard review, screen-reader output, focus behavior, names, instructions, errors and status changes.

When you find a barrier, describe what happened. Do not convert a skipped check, a count or a tool result into a universal conformance or legal conclusion.

3 / Report

Make the next action reproducible.

A useful finding names the reviewed location and state, observed behavior, user impact, reproduction steps, recommended direction and limitation. It gives the person implementing the fix enough context to return to the same experience.

Retesting is a separate activity with its own environments, changed findings, cycles and report updates. A focused example is not a full audit, ACR, accessibility statement or legal certification.

Read next: the written demonstration and the evidence methodology.

Source

Keep the boundary.

The W3C evaluation overview explains why evaluation uses more than automated tools. Apply the source to its scope.

Discuss an audit
  1. 01

    Scope what matters

    Define the task, state, environment and limits before drawing conclusions.

  2. 02

    Review with human checks

    Interpret interaction, context and task completion alongside automated signal.

  3. 03

    Report the limit

    Say what the evidence supports, what it does not and what should be tested next.