Free tool · no signup · runs now
Markdown negotiation checker
Agents fetch Markdown when you offer it — fewer tokens, cleaner citations. The checker negotiates the same way the audit does.
Part of the AI agent discovery checker group — also included in the free full scan.
Checks site-wide files at the domain root — any page URL on the domain gives the same result.
One free scan · anonymous preview · sign in to save the full report
Serves text/markdown with Vary: AcceptStatic example for Markdown negotiation checker; run the check above for your own result.What it checks
See whether agents asking for Markdown get Markdown — with Vary: Accept done right.
Markdown body for the Accept header, Vary: Accept on the response.
Reads the no-JS response only, never runs JS. Follows up to 5 redirects.
What good looks like
Good looks like text/markdown served for the Accept header with Vary: Accept set — without the Vary, a cache can hand HTML to the next agent that asks. HTML-only responses fail; an unreachable page reports na.
How to fix it
Return text/markdown (frontmatter plus body plus JSON-LD) for the Accept header with Vary: Accept.
What the full scan adds
- All 34 checks in one pass — one URL, every category, a single report.
- Evidence on every row: the values found, why it fails, the fix, and a deep link per check.
- Your score against the global distribution — see where you stand.
- Sign in and the report is kept for 1 year; anonymous reports expire after 7 days.
Official source: Cloudflare, markdown for agents · Rule: agent-readiness v1 · checker library
Tracking a Google update? Run the update impact check or read the search update radar.
Frequently asked questions
Is the Markdown negotiation checker free?
Yes — one free run without an account, with an anonymous preview. Sign in to save the full report and keep the evidence.
How often should I run it?
After every deploy that touches routing, templates or the checked file — and after each Google core update, to confirm nothing regressed.
Does the check run JavaScript?
No — it reads the no-JS response only, exactly what a crawler or AI agent sees when it fetches the URL.
How is this different from a full audit?
This page runs one check with its full criteria and fix notes. The free full scan runs every category at once and keeps the evidence per URL — use it once you know which checks matter.