From URL to audit report
Four stages, fully automated. Here's what happens when you scan.
Discovery
BFS crawl + sitemap parsing
ClearSight starts from your root URL and discovers every page on your site.
- Follows every internal link via breadth-first search
- Parses sitemap.xml for URLs not linked in navigation
- Same-origin filter — only crawls pages on your domain
- URL normalization deduplicates trailing slashes, fragments, and tracking params
- Configurable page limit and throttle delay between requests
Page scanning
Concurrent Playwright + axe-core
Each page is rendered in a real browser and tested against 50+ accessibility rules.
- Rendered in headless Chromium at 1280x720 viewport
- axe-core runs 50+ WCAG 2.1 Level A and AA rule checks
- Custom engines check link text quality and touch target sizing (48x48px)
- Screenshots captured for visual element inspection
- Bounding boxes computed for every flagged element
- Up to 3 pages scanned concurrently
AI enrichment
ClearSight AI engine
Every issue gets enriched with AI-generated descriptions and fix suggestions.
- Each issue analyzed with full page context for human-readable descriptions
- Actionable fix suggestions with code examples
- Confidence scores (0-100) assigned to each finding
- Falls back to axe-core help text if AI is unavailable
- Up to 2 parallel enrichment calls
- Preliminary results visible while enrichment runs
Results & export
Score, track, and share
Get a scored report with prioritized issues you can track and share.
- Accessibility score (0-100) per page and aggregated per site
- Issues categorized as confirmed violations or potential concerns
- Four severity levels: critical, serious, moderate, minor
- Issue tracking across crawls — new, fixed, recurring
- PDF reports with AI-generated executive summary
- Excel spreadsheets with filterable multi-sheet data
Built for reliability
Three concurrent queues, automatic retry, progressive results, and graceful degradation.
Three independent job queues
BullMQ + Redis manages crawl-discovery, page-scan, and ai-enrichment as separate queues with independent concurrency, retry, and backoff.
Cross-crawl issue tracking
Every issue gets a SHA-256 hash. Re-crawls compare hashes to show what's new, fixed, or recurring. Dismissed issues stay dismissed.
Progressive results
Raw findings are saved before AI enrichment starts. The frontend shows preliminary results immediately, then upgrades them as AI completes.
Fault-tolerant pipeline
Automatic retry with exponential backoff. AI failures fall back to raw descriptions. Stalled jobs are recovered automatically.