Google, Microsoft, OpenAI and Anthropic Back One AI Training Opt-Out; 0 of 31 Canadian Sites Had It
One shared robots.txt token states a training preference for participating providers. A next-day check found none of 31 large Canadian sites using it, while four providers set their own rollout clocks.
Tech · · 1 view
Web publishers have one robots.txt token for stating an AI-training preference to participating providers: __usage_ai-training. The IETF document is an Internet-Draft with no formal standing in the standards process. The token expresses a use preference; it does not prevent a crawler from fetching a page.
| Provider | Timing stated |
|---|---|
| Anthropic | Implementation started; support across services will take time |
| OpenAI | Aims for the coming months |
| Aims for 2027 | |
| Microsoft | February 1, 2027 |
AI Magazine checked 31 large Canadian websites on October 31. None carried the new token; 24 named none of the 11 AI crawler names or training tokens the outlet checked. It is a one-morning snapshot of 31 sites.
One shared token is cleaner than four crawler names. A publisher's next check is when each service has implemented what it is now agreeing to read.
Based on: The AI training opt-out is one robots.txt line now. Most Canadian sites we checked name no AI crawler, AI Magazine, October 31, 2026; Making it easier for websites to express AI training preferences, OpenAI, October 30, 2026; Supporting a shared AI training control for websites, Google Search Central, October 30, 2026; Your content, your choice: clearer preferences for training and grounding, Microsoft Bing, October 30, 2026; Machine-Readable Training Preference Signal, IETF Internet-Draft, October 30, 2026; New draft: draft-kelley-aipref-training-preference-signal-00, Kevin Kelley, IETF AI Preferences working group, October 30, 2026.