Content signals
What we check
Content signals are a newer convention, started by Cloudflare in 2025, for saying not just which bots may visit but what they may do with what they read. A Content-Signal line in robots.txt gives three answers, each yes or no: search (show it in search results), ai-input (use it to answer questions), and ai-train (use it to train models).
We read any Content-Signal lines in your robots.txt, wherever they sit; if two disagree, the last one wins. We report other signals by their own names. We also look for an /ai.txt file: it counts if it loads, is plain text or Markdown, and isn’t empty. The report says it in words, for example “robots.txt says yes to search results and AI answers, no to AI training.”
This check is reported, never scored. What a site permits is its own choice, not something to fix.
Why it matters
robots.txt only says who may come in. Content signals say what happens next, which is the question most site owners actually care about: being found and quoted is good, being used for training may not be.
Not every AI company reads these signals yet, so they don’t change what an assistant does today. They do put your wishes on record in a place machines can read.
How to pass
There’s nothing to pass. If you want to state your preferences, add a line to robots.txt, under the group it applies to:
User-agent: *
Content-Signal: search=yes, ai-input=yes, ai-train=no
Allow: /If you’re on Cloudflare, its managed robots.txt can add these lines for you.