Google plans to support shared AI training opt-out via robots.txt

Google plans to support a shared way for website owners to opt out of having their content used to train AI systems, using robots.txt. The announcement concerns a machine-readable instruction that site operators can publish on their own servers, giving automated systems a common signal about their training preferences. It extends the role of robots.txt, commonly used to communicate crawler access rules, into the area of AI training control.

A shared approach could make it easier for publishers to express a preference once in a familiar place rather than rely on separate controls for different AI services. It also depends on systems recognizing and honoring the signal: robots.txt communicates a site’s instructions but does not itself prevent access. The supplied information does not specify the proposed syntax, which AI systems would follow it, how training use would be distinguished from other crawling, or when support would begin. Those details will determine how reliably the control works in practice and how much control it gives website owners over reuse of their content.