One robots.txt setting will soon block all AI training
A shared token called _usage_ai-training would replace blocking each AI company's bot one by one.
Site owners are getting one switch to say no to AI training on their content. Marie Haynes (@marie_haynes) says a new robots.txt control, written as _usage_ai-training, will instruct all participating AI tools not to train on your content. Robots.txt is a small text file on your site that tells automated bots what they are allowed to do.
there will soon be a new robots.txt control that instructs all AI tools not to train on your content: _usage_ai-training
Right now, keeping your pages out of AI training means handling each company separately. Every AI developer runs its own crawler with its own name, so blocking them is a list you have to keep adding to. The shared token replaces that list with a single instruction.
Google's wording, quoted by Haynes, is that site owners "will be able to use a shared 'policy token' within robots.txt to set broad preferences across participating AI providers."
Who has signed up
Barry Schwartz (@rustybrick) says Google, Microsoft, OpenAI and Anthropic are all onboard. That covers the companies behind Google's AI features, Copilot, ChatGPT and Claude.
So Google, Microsoft, OpenAI and Anthropic are all onboard
Schwartz also notes the controls are announced for 2027 and are under consideration by the IETF, the standards body that agrees on how the internet's plumbing works. So this is a plan, not a setting you can use today.
What this means for your site
Nothing changes on your site this week. There is no token to add yet, and the sources do not say when in 2027 it starts or what the exact rollout looks like.
The decision you will eventually have to make is whether you want AI companies training on your pages at all. Blocking training is not the same as blocking AI answers from citing you, and the sources here only cover training. If you want your pages kept out of AI answers in search, that is a separate set of controls Google has already listed.
For most small business owners, the honest answer is do nothing for now. Your pages being used to train a model does not cost you a click today, and the visibility you get from being quoted by AI tools is usually worth more than the principle. Publishers with a lot of original research, data or writing they sell access to have a stronger reason to block.
If you want to be ready, find out whether your site already has a robots.txt file. Type your domain followed by /robots.txt in a browser and see what comes up. Most website builders generate one automatically and let you edit it in settings. Knowing where yours lives means adding one line in 2027 takes a minute rather than an afternoon.
Be the person who already knew
One email a day. Under three minutes. You will hear about a Google change before it shows up in your traffic.