For thirty years, docs were written for humans, scored by Google, and graded on search rankings. That contract changed in 2024-2026. The new audience — coding agents, browsing agents, and answer engines — reads the same pages very differently. None of them care about your branding or your fonts; all of them care about whether your content is reachable, parseable, and tagged.
Anthropic shipped v2.1.105 of Claude Code. The release note: "Improved WebFetch to strip <style> and <script> contents from fetched pages so CSS-heavy pages no longer exhaust the content budget before reaching actual text." Agents now ignore your CSS. Pages designed around styling lose the content-design contract.
↗ anthropics/claude-code CHANGELOGOpenAI's agentic browser. A full Chromium running every site through the OWL layer. Atlas users see your post-JS DOM, including content rendered via React or Vue — but only if your site responds to the browser's slower, instrumented requests. The headless / raw-HTTP gap matters more now.
↗ OpenAI: Building ChatGPT AtlasPerplexity's Chromium-based agent browser. Fetches your docs as a logged-in human would. Different sample of your traffic than the declared PerplexityBot — and they read differently.
↗ Comet engineA new layer of well-known endpoints emerged: /.well-known/mcp/server-card.json, /.well-known/agent-skills/index.json, /.well-known/agent-card.json. Agents discover capabilities programmatically. The docs site that publishes a manifest gets called by agents trying to do real work; the docs site that doesn't is invisible to that path.
Other AI-readiness tools tell you whether an endpoint exists. Useful, but one-dimensional. Docs Lens shows you what three real agent reader populations actually receive when they fetch your page — and where they diverge from each other and from what you see in a browser. The divergence is the lesson.
No gamified Level 0 / Not Ready rhetoric. The grade is a footer-level shareability handle. The headline is a verdict sentence and an audit trail.
Every profile names the agents it represents — Claude Code WebFetch, ChatGPT Atlas, Perplexity Comet — with linked release notes. No hand-waving.
Each finding shows the literal HTTP request and response that produced it. Skeptical? curl the URL yourself. Every claim ties to a real fetch or a Playwright snapshot.
We're not trying to replace these tools — they're good. We're the visual-first complement: the tool you run when you want to see what agents see, not just whether an endpoint passes a probe.
Probes ~13 endpoints + WebMCP detection. Punitive scoring (Level 0 / Not Ready) — most docs sites start at 8/100.
We run 38 checks but score honestly: missing optional endpoints don't penalize. We also show how three different agent reader populations actually read your content, not just whether endpoints exist.
22 checks, polished UI, 87/100 for ekline. The AI-readiness audit Fern is best known for.
Same audit surface, plus the visual hook: side-by-side rendered page vs three agent reads (Raw HTTP, Headless, Snippet) per page. Fern has a single agent-class lens; we show divergence.
Open-source CLI by the AFDocs spec authors. Comprehensive. Best for CI integration.
We complement it. Our agent-fix prompt explicitly points users at npx afdocs check --fixes --verbose for deeper local detail. We're the visual-first companion.
Pick a docs URL, watch three agent populations read it, get a prompt to fix what's broken.