Why link a Markdown version of a page for AI agents?
A link in the head of the page that points at a Markdown copy of it. Some AI agents do not ask a site for Markdown with a request header; they follow a link, so a page that offers Markdown only through content negotiation is invisible to them.
What it actually is
A Markdown copy of a page is the content without the styling and scripts, which is the cleanest thing a language model can read. There are two ways to offer it. One is to answer a request that asks for Markdown, which our Markdown negotiation check looks at. The other is a link element in the head with the relation alternate and the type text/markdown, naming the address of the copy. Agents that browse like a person find the second.
Why it matters
An agent that cannot find the clean copy reads the full HTML with its navigation and scripts, and spends its attention there. A link in the head is a signpost to the better version, and it costs the page nothing.
What good looks like
A head link with the alternate relation and the Markdown type, pointing at a copy that exists and carries the same content as the page. The copy should be the page, not a stub or a different page.
Being honest about it
This is an Advanced check, optional and never a fail. It stays out of the headline score. It complements the Markdown negotiation check rather than replacing it: a site can offer Markdown either way, and we look at each separately.
Where most sites go wrong
The link points at a copy that has since moved, or at a Markdown file that is out of date and tells an agent something the page no longer says. The other is the type declared wrongly, so a machine does not recognise it as Markdown.
On the web
State of the Web · November 2026In November 2026, 1% of the sites where it applies (88% of the sites we could read) passed the “Markdown version linked from the page” check.
Next.js sites pass it most often, at 5%. Webflow sites pass it least often, at <1%.
On the sites scanned in both October 2026 and November 2026, it stayed at 8%.
The month-by-month line counts the sites scanned in each calendar month, so it can differ from the figure above, which uses each site’s latest scan from the last three months.
Not enough history yet.
Based on the sites Bot Appétit has scanned. Most of these sites come from lists we scan for research. Not a random sample of the web.
Not sure where yours stands?
A free scan checks this and everything else in about 20 seconds — no signup. Or if it's already flagged and you'd rather it were simply handled, we do that too.
Flagged for this one? We handle it. No pitch, just a plain-English chat.
Book a call