Four AI Companies Support Unified Website Training Preference Signals

OpenAI, Google, Microsoft, and Anthropic have stated they will support a common AI training preference signal for website publishers. Site operators can set a “policy token” in their robots.txt file to express their choice to exclude content from generative AI model training to all participating AI companies at once, reducing the workload of maintaining separate rules for each company. Existing proprietary control mechanisms from each company will remain available for use. Implementation timelines are not synchronized: Anthropic states it has already adopted the signal; OpenAI plans to implement it over the coming months; Microsoft is scheduled to respect the new preferences during crawling starting February 1, 2027; and Google plans to provide support in 2027. Microsoft will also introduce separate controls allowing publishers to independently choose whether to permit content for AI training and for citation in generated answers; these choices will not affect a website’s eligibility for standard Bing search results. Meta has expressed support for the common signal but has not committed to a timeline for adoption.