Microsoft Introduces New AI Training and Grounding Preferences for Web Publishers

Microsoft announced newer and clearer preferences enabling web publishers to manage whether their content is collected and utilized for generative AI training and grounded responses across Bing, Copilot, and Microsoft 365. The company stated that web publishers will no longer need to rely solely on legacy NoArchive and NoCache tags to control AI usage while staying indexed for traditional search. Under the new approach, AI Training preferences can be specified at the domain or path level using a shared syntax developed with the Internet Engineering Task Force (IETF) AI Preferences working group, or overridden at the individual page level via meta tags or HTTP headers. In addition, Microsoft introduced independent page-level AI Grounding controls, featuring Allow, Disallow, and a DisallowExtended option that restricts grounding strictly to first-party experiences such as Copilot and Microsoft 365 while blocking third-party syndication APIs. Both settings default to Allow if unconfigured. Microsoft indicated these preferences will take effect as crawlers visit pages starting February 1, 2027, with legacy NoArchive and NoCache rules phased out for AI usage by March 1, 2027.

Key takeaways

• AI Training and AI Grounding are configured independently, defaulting to Allow if no explicit rule is published.
• AI Training syntax aligns with IETF draft standards via robots.txt or page-level tags, while AI Grounding is managed strictly at the page level.
• A DisallowExtended option permits grounding in Microsoft-operated tools like Copilot while excluding third-party API grounding services.
• New controls take effect February 1, 2027, and publishers relying on NoArchive or NoCache must update configurations before March 1, 2027.