Google and OpenAI plan one shared line for websites to say no to AI training
A single signal in a site's robots.txt file will let owners opt out of AI training at many companies at once, but Google aims to support it in 2027.
If you run a website, telling AI companies to stay away is about to get easier. On October 30, Google and OpenAI each said they will honor one shared signal for AI training.
Today the job is messy. Every AI company sends its own robot to read the web. OpenAI says site owners have had to name each robot and keep separate rules for every one of them.
The new plan puts one shared line in a file called robots.txt. That is a small text file on a website that tells robots what they may and may not do. With one line, a site owner can say no to AI training at all the companies that take part.
The idea comes from a working group at the Internet Engineering Task Force, a group that writes common rules for the internet.
Google says that if a site uses the shared signal to opt out, it will treat that as a no for its own AI training setting, called Google-Extended. The old company-by-company settings keep working too. So you can still say yes to one company and no to another.
There is a catch on timing. Google says it aims to support the signal in 2027. OpenAI says it aims to add it in the coming months.
There is also a trade-off. Google says sites that allow Google-Extended can show up in answers inside its Gemini apps and get clicks from them. Saying no everywhere could cost you that.
Open it and check the claim yourself. That is the point.