Agent readability score
Score your website against the Vercel Agent Readability Spec. Get a 0-100 score across 38 checks covering discovery, structure, and context for AI agents.
Last updated
Why does the Vercel Agent Readability Score matter?
Google spent decades training webmasters to rank in search results. ChatGPT, Claude, and Perplexity don’t have that lineage — they crawl what loads cleanly and cite what parses correctly. A Pew Research study (July 2025) found Google users who saw an AI-generated summary clicked on source links only about 8% of the time, roughly half the rate of searches without one. Princeton’s GEO research (KDD 2024) measured that adding cited sources and statistics to a page can lift its inclusion in AI answers by about 40%. A site that scores well here is one agents can read, summarise, and send qualified traffic to. Sites that score poorly get paraphrased badly or skipped entirely.
How is the agent readability score calculated?
Your score is calculated as the percentage of checks that pass: score = (passed checks / total checks) x 100. Per-page checks run on every discovered page, so systemic issues compound. A site with many pages has a larger denominator.
What do the score ratings mean?
What does the agent readability scanner check?
- 15 site-wide checks: llms.txt, robots.txt, sitemap, AGENTS.md, HTTPS, OpenAPI
- 23 per-page checks: HTTP headers, meta tags, JSON-LD, headings, text ratio, markdown mirrors, content negotiation, code blocks, JS rendering
How do I improve my agent readability score?
Every failed check on your scan result comes with a concrete fix - what file to add, what header to set, what markup to emit. Work top-down: site-wide issues (llms.txt, robots.txt, sitemap) compound across every page, so they’re usually the highest-leverage thing to fix first. Pro-tier monitoring re-scans weekly and emails you if the score regresses.