Google, Microsoft and OpenAI Say They Will Honor One Shared AI Training Opt-Out

A single robots.txt line could let your website say no to AI training at several companies at once, but support does not start until 2027 for Google and Microsoft.

Google, Microsoft and OpenAI have each announced plans to honor one shared robots.txt line that tells AI companies whether they may train on a website's content, according to Search Engine Journal. None of the three supports it yet. (robots.txt is the small text file on your site that tells automated visitors what they may use.)

The line comes from a draft written by Anthropic's Kevin Kelley for the IETF's AI Preferences working group. It adds no new directive. You write a normal robots.txt group using the token `__usage_ai-training` with Allow and Disallow rules. It covers training only, and does not control crawling. A site with no such group is treated as having no preference.

Google says it aims to support the signal in early 2027. If a site uses the shared line without naming Google-Extended, Google will treat it as if Google-Extended were disallowed, Search Engine Roundtable reports. Google-Extended does not affect inclusion in Search, and AI Overviews and AI Mode use a separate Search control. OpenAI says it aims to honor the signal within the next few months.

Microsoft goes furthest. It adopts the shared line for training and adds a separate per-page AI Grounding preference. It begins honoring the new options on February 1, 2027, and stops recognizing its old NoArchive and NoCache tags for training or grounding after March 1, 2027. Search Engine Journal says sites still using the old tags on December 15 can set a temporary preference in Bing Webmaster Tools. The draft is not an official IETF standard, and Search Engine Roundtable notes disputes remain.