← All news

AI

One robots.txt line to opt out of AI training (2026-11-02)

Website owners may soon have a simpler way to express a preference against AI training. The October 30 announcement covered in the SAZ Labs short says OpenAI, Google, and Microsoft will honor a shared robots.txt instruction. The aim described in the script is to replace the need for separate training rules for each company’s crawler with a common signal.

The short describes a new token called usage AI training, with a group set to disallow. It also says an ordinary wildcard rule does not count. Those details indicate that the proposal has its own intended signal rather than treating every existing broad rule as a training preference. The brief does not supply a complete, copy-ready syntax example, so this article does not invent one.

The announcement is about future support. At the time of the short, it was not live. OpenAI described support as arriving in the coming months; Microsoft’s stated start was February 1, 2027; and Google planned support during 2027. These are separate timelines, so the shared direction should not be mistaken for simultaneous enforcement by all three companies.

Another distinction is what the preference covers. The script says it applies to training, not crawling. Bots can still visit pages, even when the owner has expressed the training preference described in the announcement. Saying no to training and blocking a crawler from visiting are therefore separate outcomes in this account. The shared instruction should not be understood as a promise that all automated access stops.

For a website owner, the useful takeaway is to keep the intended policy and its implementation timeline together. The proposal offers a common way to communicate a training choice, but the source describes support that has yet to arrive and a scope limited to training. It does not claim universal adoption by every crawler or immediate protection at the time of publication. Watch the original short below for the planned dates and the distinction between training preferences and continued crawling, which are central to understanding what was actually announced.

Watch on YouTube