What are Link headers, and how do they help AI agents?
A Link header is a pointer your server sends alongside a page, saying "there is a related resource over here". When it points at your sitemap or a description of your site, an agent learns where your map is from the response itself, without reading the page or guessing addresses.
What it actually is
Pages can link to related resources in their HTML, but the web's linking standard also lets a server put the same links in the response headers, the small envelope of information that arrives before the page. Each link carries a relation saying what kind of thing it points to: a sitemap, a description of the resource, the documentation for an API.
Why it matters
Agents are economical. Many make a quick request, look at what came back, and decide what to fetch next. A header that points at your sitemap or a description of your site hands them the next step in the first response, before they have parsed anything, and it works for every kind of file — including the ones that are not HTML and have nowhere to put a link.
What good looks like
Your responses carry a Link header pointing at the discovery resources you actually have — the sitemap, a description of the site, an API catalogue if you publish one — each with the relation that names it. Performance hints and platform boilerplate can sit in the same header; they just do not count as discovery.
Being honest about it
This is listed under Agent & MCP readiness and does not move your score: agents can still find a sitemap through robots.txt or a well-known address. This is optional, and we treat it as a suggestion rather than a requirement: agents can still find a sitemap through robots.txt or a well-known address, so its absence leaves nothing unreadable. It is on the report because it is cheap once your hosting allows custom headers, and it is the one place a machine is told where your map is before it has read a single page.
Where most sites go wrong
The common case is a site that does send Link headers, but only the ones its platform adds by itself: hints to connect to a font server early, or a pointer to the platform's own programming interface. Those are useful for other reasons and tell an agent nothing about where your content is mapped. The other is a header pointing at a sitemap that has since moved.
On the web
State of the Web · November 2026In November 2026, 1% of the sites we could read passed the “Resource-discovery Link headers” check.
Next.js sites pass it most often, at 10%. Shopify sites pass it least often, at 0%.
On the sites scanned in both October 2026 and November 2026, it stayed at 11%.
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.
Goes with
-
Sites that pass this are 24.5 times as likely to pass the “Content usage preferences declared” check too. 69% of the sites that pass the first check pass this one, against 3% of the sites that do not.
-
Sites that pass this are 21.5 times as likely to pass the “Clean content for agents (Markdown)” check too. 72% of the sites that pass the first check pass this one, against 3% of the sites that do not.
-
Sites that pass this are 13.5 times as likely to pass the “security.txt is published” check too. 53% of the sites that pass the first check pass this one, against 4% of the sites that do not.
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