You’re getting a Bat-Signal for AI training: one robots.txt rule telling every participating AI company not to train on your content.

Google, Bing and OpenAI promised to look up, but they didn’t. Yet.

Before you copy and paste: It’s a normal robots.txt group naming a shared token instead of a crawler. Search Engine Journal traced it to an individual IETF draft by Anthropic’s Kevin Kelley, so it’s a proposal, not a standard.

Google makes it a package deal: From early 2027, Google will treat the opt-out like blocking Google-Extended, its AI training control, unless your file names Google-Extended.

That also cuts your pages out of Gemini Apps answers and their clicks. Search, AI Overviews and AI Mode have separate controls.

Bing lets you split the bill: From February 2027, Bing will honor the shared rule and a page-level grounding tag, which decides if your pages can inform AI answers.

Block training, allow grounding, and Copilot can still cite you.

Still limiting AI use with NoArchive or NoCache? Bing ignores them after March 1, 2027.

Publish the new settings, and from December 15, log a temporary grounding preference in Bing Webmaster Tools.

OpenAI’s in, GPTBot stays: OpenAI will honor the shared rule within months, and your rules for GPTBot, its training crawler, keep working.

Careful with the other off switch: Google’s AI Overviews and AI Mode opt-out lives in Search Console. Cyrus Shepard’s small test found it costly:

• AI impressions earned an estimated 0.9% CTR
• Organic clicks dropped too while the site was out

Until someone answers the Bat-Signal, your old rules do all the work.