What is Content-Usage, and how does a site state AI preferences?
A line in robots.txt, or a response header, that says what your content may be used for: AI training, AI answers, or search. It comes from the IETF's AI preferences work, which is still an Internet-Draft, and it is a statement of intent, not a lock.
What it actually is
robots.txt answers who may fetch a page. Content-Usage answers a second question: what they may do with it. The proposal defines a small vocabulary for the common uses, training a model, using content in an AI answer, and search, and a site can say yes or no to each. It can be written as a line in robots.txt, optionally for one path, or sent as a header on responses. It is a sibling of Content-Signal, which does a similar job in robots.txt.
Why it matters
Without a stated preference, every machine applies its own default, and those differ. A site that wants to appear in AI answers but not feed training has no standard way to say so except by stating it. Writing it down puts your position on the record in a place a machine can read.
What good looks like
A preference that matches what you actually want, stated in robots.txt or in the header, and consistent with what the rest of the site says. If you also publish Content-Signal, the two should agree.
Being honest about it
This is an Advanced check and a proposal. No AI operator is known to honour it yet, so declaring a preference changes nothing about how any machine behaves today. It never fails a page and stays out of the headline score. Treat it as putting your position on the record, not as protection.
Where most sites go wrong
The usual slip is contradiction: a robots.txt that welcomes AI crawlers while a Content-Usage line forbids AI use, or the reverse. Whoever reads both has to guess which you meant. The other is a header present but carrying nothing the proposal recognises, which declares no preference.
On the web
In October 2026, <1% of the sites we scanned passed the “AI usage preferences (Content-Usage)” check.
| Result | Share of sites |
|---|---|
| Pass | <1% |
| Worth a look | 98% |
| To fix | 0% |
| Couldn't judge | 2% |
Next.js sites pass it most often, at 1%. WordPress sites pass it least often, at 0%.
Not enough history yet.
Based on the sites Bot Appétit has scanned: ones people chose to scan, plus 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