Hybrid detection
axe-core is used as a baseline and extended with AccessAudit custom WCAG checks.
About AccessAudit
Built to make accessibility findings easier to understand.
AccessAudit is an independent accessibility evaluation product focused on automated WCAG checks, visual evidence, AI-assisted explanations and guided verification.
- Built around WCAG reporting and validation
- Uses automation, AI assistance and human review
- Designed for developers, teams and research use
Product direction
A hybrid accessibility workflow, not a single scanner.
AccessAudit combines automated checks, visual context, AI assistance and guided verification so accessibility findings are easier to inspect and act on.
Visual evidence
Findings can be reviewed through screenshots, markers, affected elements and grouped issue context.
AI assistance
AI helps explain findings, draft remediation guidance and support selected semantic checks.
Workflow
From detection to verification.
The goal is not only to list issues, but to help users understand what was found, where it appears and what still needs review.
Detect
Run automated WCAG checks and custom rules against the target page.
Locate
Review the issue list, affected elements and visual evidence.
Understand
Use plain-language explanations, WCAG context and remediation guidance.
Verify
Use guided checks and manual review where automation is not enough.
Why it exists
Accessibility reports should be actionable.
Many tools can report issues. The harder part is helping users understand what matters, how to verify uncertainty and what to fix first.
Practical accessibility
The goal is to help teams move from long issue lists to clear decisions and concrete fixes.
Honest automation
Automated checks are useful, but not enough. AccessAudit separates detection, AI assistance and human review.
Evidence over claims
Benchmark reports, visual markers and guided verification make findings easier to inspect and validate.
AI transparency
Clear separation between scanning, AI assistance and human review.
AccessAudit is designed so users can understand where automation ends, where AI assistance begins and what still needs manual verification. AI helps explain and interpret findings, but it is not treated as a final compliance authority.
What may be sent to AI
For explanations, AccessAudit may send the rule ID, WCAG criterion, severity, selector, short HTML snippet, page URL and surrounding issue context.
When visual data is used
For selected vision-assisted checks, such as alt text quality review, the system may use screenshot or image context together with the existing alternative text.
What users should avoid
Users should not intentionally scan pages containing passwords, secrets, payment data, private customer records or other sensitive content.
What AI does not decide
AI output is advisory. Final WCAG conformance, legal risk and acceptance decisions still require human judgement and project context.
Model usage
Which models are used and why
The primary detection layer remains deterministic. AI is used for explanations, remediation guidance and selected semantic or vision-assisted checks that require interpretation.
Vision-assisted review
Used for selected semantic checks where visual context matters, especially reviewing whether existing alt text is descriptive and useful.
Explanations and remediation
Used to generate plain-language explanations, likely user impact, WCAG context, caveats and draft fix suggestions for developers.
Primary detection layer
Most findings come from axe-core and AccessAudit custom rules. AI is an assistance layer, not a replacement for deterministic checks or expert review.
Stack and capabilities
Built for practical WCAG review.
Most detection comes from axe-core and custom rules. AI is used for explanations, draft fixes and selected semantic assistance, not as a legal compliance engine.
Next.jsaxe-corecustom WCAG rulesClaude Sonnet visionClaude Haiku guidancevisual overlayguided verificationPDF reportsmonitoring
Validation phase
Want to review the current prototype?
The application is live and currently open during validation. Feedback from developers, researchers and accessibility practitioners is especially useful at this stage.