# canaisee

A free, public diagnostic for the post-browser web. Paste any URL and
canaisee grades it on how readable it is to AI agents: a single headline
grade (A+ through F) backed by a transparent, versioned sub-score breakdown.

Think of it as a browser network tab, but for agents.

## Why this exists

AI systems like ChatGPT, Claude, Gemini, and Perplexity increasingly answer
questions about your site without sending a human visitor. Whether those
answers are accurate depends on whether the model can actually read the
page: whether your robots policy lets it in, whether your content survives
a plain HTTP fetch, whether your markup gives it hooks, and whether you
expose the agent-native surfaces — content negotiation, MCP, A2A, llms.txt,
.md mirrors — that make agent access cheap.

canaisee measures those signals and returns a single headline grade plus a
per-check breakdown so you know exactly what to fix.

## How the grade is computed

canaisee v1.5 splits the composite into two tiers (scoring math carried forward from v1.3):

- **Hygiene** (85% for content sites, 75% for interactive) — table stakes:
  crawler accessibility, content readability, and semantic structure.
- **Frontier** (15% / 25%) — ahead-of-the-curve agent-native signals:
  `Accept: text/markdown` negotiation, `/.well-known/mcp.json`, WebMCP
  tool declarations, A2A agent cards, `Content-Signal` response headers,
  rich structured data, and the older `/llms.txt` and `.md` mirror
  signals.

A site with perfect Hygiene and zero Frontier lands at B (85). Frontier
features push a site into A and A+. See the published
[rubric](https://canaisee.com/rubric) for the exact weights.

## What this tool cannot measure

canaisee measures signals that should correlate with AI readability. It
cannot directly verify whether a specific model has ingested your content,
whether ChatGPT cites your site in answers, or whether your pages appear in
any given training corpus. For outcome-level questions see the companion
"What does AI say about your site?" demo on the Evangent site.

## Operating the tool

- Canonical URL: https://canaisee.com/
- Crawler identity: `CanAISeeBot/1.0 (+https://canaisee.com/about-bot)`
- Rate limits, cooldowns, and takedown flow are documented at
  [/about-bot](https://canaisee.com/about-bot).
- Source code and rubric history live at
  https://github.com/8gara8/agentread under the MIT license.