On October 30, 2026, Google, Microsoft and OpenAI published plans to honour one shared robots.txt signal for AI training, and Anthropic published the underlying proposal. None of them is enforcing it yet. Here is what the signal is, when each company says it will start, and what it does not do for a website owner.
The signal
The token is __usage_ai-training. A site names it on the User-Agent line of a robots.txt group, as described by PPC Land:
User-Agent: __usage_ai-training Disallow: / Allow: /press-kit/
That expresses a preference against AI training on every path except /press-kit/. A wildcard group (User-Agent: *) does not count as expressing the preference. The mechanism comes from an Internet-Draft, “Machine-Readable Training Preference Signal”, by Kevin Kelley of Anthropic. The IETF notes it is an individual draft, not endorsed by the IETF, and it expires on May 3, 2027.
Who starts when
- Anthropic: says it has already started implementing support, with no date.
- Microsoft: will honour AI training and grounding preferences from February 1, 2027. Bing Webmaster Tools transition options open December 15, 2026.
- Google: says it intends to support the signal in early 2027.
- OpenAI: aims to implement it “in the coming months”, with no firm date.
What it does and does not cover
- It does not affect inclusion in Google Search or Bing results. Blocking Googlebot or Bingbot still blocks search.
- It is not a crawling directive; Microsoft says its settings govern use of content, not crawl access.
- Google plans to treat a shared-token training opt-out as if Google-Extended were disallowed, unless Google-Extended is addressed explicitly. Google-Extended also covers grounding in Gemini Apps and Vertex AI, so PPC Land reads the opt-out as reaching grounding too; Google’s post does not address that directly. Microsoft limits the signal to training and handles grounding separately.
- OpenAI says the signal will not replace its existing robots.txt controls, which include GPTBot.
- The draft has no enforcement mechanism, and private agreements may override preferences.
What a site owner can do now
- Decide whether you want to opt out of AI training at all. Opting out may be reasonable for paywalled or original reporting; it is a business choice.
- Keep your existing company-specific rules (such as GPTBot and Google-Extended) in place, since the new token does not replace them.
- If you add the new group, test the robots.txt syntax and avoid assuming a wildcard rule covers it.
- Check each company’s documentation as the rollout dates arrive, since details may change.
Related: UK publishers have been pushing Google on opt-outs too, see our report on the UK ruling.
