What Google announced
On 30 October 2026, Google Search Central said it plans to support a shared AI-training control in robots.txt.
John Mueller, a Google Search Advocate, wrote the post.
The control comes from an Internet-Draft at the Internet Engineering Task Force, or IETF. Google links to that draft. The draft’s record says the IETF has not endorsed it, and it has no formal standing as a standard.
Google says a website could set one training preference for participating AI providers, instead of a separate rule for each company.
The post says Google aims to support the control, with other AI providers, in 2027. Google’s crawler page, updated the same day, says Google intends to support the signal in early 2027.
How the shared line is written
The draft names the token __usage_ai-training. Its example refuses training on the whole site and allows it under one folder.
In that file, Disallow means a preference against AI training on a path. Allow means a preference for it.
If the file never names the token, the draft says no preference was expressed. A User-Agent line of only an asterisk does not match this token.
The draft says the token is not an order to stop crawling. A site that wants to limit fetching still uses the crawler’s own name. A private agreement may override the preference in the file.
The draft defines AI training as using a page to change the learned settings of a generative model. Training a model that only classifies, ranks or scores sits outside that definition.
Google’s post says the new control uses that definition while discussion of the wider vocabulary continues.
The draft calls itself work in progress. It expires on 3 May 2027 unless a newer draft replaces it. That expiry is the draft’s clock. It is not the day Google turns the control on.
- User-Agent: __usage_ai-training
- Disallow: /
- Allow: /press-kit/
What Google would do with the line
Google’s post describes one reading. If a site uses the shared token to opt out of training, and the file does not set Google-Extended, Google plans to treat that as a disallow of Google-Extended.
Google says it has offered Google-Extended since 2023. The crawler page says the token covers two uses of pages Google has already crawled.
One use is training future Gemini models for Gemini Apps and the Vertex AI API for Gemini. The other is grounding, which means using those pages inside an answer.
The same page says Google-Extended does not remove a site from Google Search, and Google does not use it as a ranking signal.
A shared opt-out that Google reads as a Google-Extended disallow would cover training and grounding. On Google’s description, it would not drop the site from Search.
A site can still write its own Google-Extended rule. Google says that specific rule keeps working beside the shared line.
Google also says a site that specifically allows Google-Extended is eligible to appear in Gemini Apps answers and to receive clicks from that grounding.
The draft says a company may treat its own token as more specific than the shared token. Google’s plan to keep an explicit Google-Extended rule fits that idea.
Why it matters in the Philippines
A Filipino blog, news site, school site or small business could state one training preference in robots.txt, instead of a different name for every company.
Until Google switches the control on, that line does not change what Google does. The Google control that works now is Google-Extended.
A newsroom can ask Google not to train on its stories and still stay in Search. Those are separate choices on Google’s crawler page.
A school site should decide with the people responsible for the pages before it edits the file.
The line is a preference. A company applies it only if that company chooses to read the token.
What to do next
- Open Google’s 30 October post, the crawler page and the draft before you edit a live file.
- To set a Google choice today, use the Google-Extended token and follow the example on Google’s crawler page.
- Do not expect __usage_ai-training to change Google before the 2027 support Google describes.
- If your file already names Google-Extended, keep that rule. Google says the specific instruction still applies.
- If you want pages eligible for Gemini grounding, Google says a specific allow for Google-Extended is the path it describes.
- Check the crawler page again before you rely on the shared line. Google said it will post updates.
How this was checked
The announcement, the 2027 aim and the Google-Extended reading are from Google’s Search Central post of 30 October 2026, opened on 31 October 2026.
The token name on Google’s crawler page, the early 2027 note, and what Google-Extended covers are from that page, last updated 30 October 2026.
The example lines, the training definition and the 3 May 2027 expiry are from the Internet-Draft, opened the same day.
Google’s post says other providers can use the signal. It does not name those providers or their dates.
Put this into practice
Read Google’s post beside the draft, and keep each claim with the page that made it.
Sources & further reading
Source note. The linked sources support the news in this article. The explanations and suggested next steps are from Pinoy AI Works.
