How AccessCheck Works

AccessCheck uses 5 AI-powered analysis layers to detect ~80% of WCAG 2.2 violations โ€” far more than any single automated tool.

The 5 Layers

๐Ÿ” Layer 1: Code Analysis 30% coverage

Technology: axe-core (Deque Systems) โ€” the industry standard accessibility rules engine.

Injects axe-core into the page, runs 100+ automated tests against WCAG 2.2 AA criteria. Catches: missing alt text, ARIA violations, heading structure, color contrast on solid backgrounds.

What it misses: Contrast on images/gradients, keyboard usability, form UX, content readability.

๐Ÿ‘๏ธ Layer 2: AI Computer Vision +10โ€“15%

Technology: GPT-4o โ€” OpenAI's multimodal AI.

Takes a full-page screenshot and analyzes it for visual accessibility issues that axe-core cannot see: text contrast on images and gradients, touch target sizes on mobile, overlapping elements, visible keyboard focus indicators.

Example finding: "White price text on light product photo โ€” contrast ratio less than 3:1."

๐ŸŽฎ Layer 3: AI Agent Simulation +10โ€“15%

Technology: Playwright browser automation + AI orchestration.

Simulates real user interactions: tabs through all focusable elements to verify keyboard navigation order, submits forms empty to check error message accessibility, verifies that all interactive elements are reachable without a mouse.

Example finding: "5 form fields have no labels โ€” screen reader users cannot identify what to enter."

๐ŸŒณ Layer 4: Accessibility Tree +5โ€“10%

Technology: Chrome DevTools Protocol โ€” direct browser accessibility tree access.

Inspects the browser's native accessibility tree โ€” the structure that screen readers actually use. Checks: landmark roles (main, nav, banner), duplicate accessible names, semantic HTML usage, heading hierarchy.

Example finding: "No <main> landmark โ€” screen reader users cannot skip to main content."

๐Ÿ“ Layer 5: Content Analysis +5โ€“10%

Technology: AI language model (GPT-4o-mini).

Analyzes all page text for: reading level (Flesch readability), generic link text ("click here", "read more"), alt text quality (does it describe the image?), form instruction clarity, mixed language detection.

Example finding: "14 links use generic text like 'read more' โ€” screen reader users lack context."

Total Coverage: ~80%

Combined, the 5 layers detect approximately 75โ€“85% of WCAG 2.2 Level AA success criteria. Manual expert review is recommended for complete compliance assessment.

Standards We Check

Comparison

ToolCoverageFix InstructionsSpeedPrice
AccessCheck~80%โœ… + code50 secโ‚ฌ49
WAVE / Lighthouse~25โ€“30%MinimalInstantFree
axe DevTools~30โ€“40%GoodInstantFree
Manual Consultant80โ€“95%Expert2โ€“6 weeksโ‚ฌ500โ€“5,000

FAQ

Is this a legal certification?
No. AccessCheck is an automated software tool. It detects technical violations and provides fix guidance. It does not replace human expert review or legal counsel.
What WCAG version do you check?
WCAG 2.2 Levels A and AA โ€” the standard required by the EU Accessibility Act (Directive 2019/882).
How many pages do you scan?
The free preview scans 1 page (Layer 1 only). The full audit scans up to 8 key pages across all 5 layers, with smart crawl that identifies page types (homepage, product, cart, blog).
Can I use this for ADA compliance in the US?
AccessCheck tests against WCAG 2.2, which is the basis for ADA compliance. However, US courts and regulations have specific requirements beyond WCAG. Consult legal counsel for ADA compliance.

โ† Back to scanner ยท Automated tool ยท Not legal certification ยท WCAG 2.2 ยท EN 301 549 ยท Directive 2019/882