Microsoft announced on October 30 new Bing AI content controls that let site owners make clearer choices about training, grounding, or both, while remaining eligible for Bing search results. The controls separate three decisions that publishers previously had to handle with blunter tools: whether content may train AI models, whether it may ground AI answers, and whether grounding is limited to Microsoft’s own products. Traditional Bing rankings are not affected by the new preferences.
Three choices, kept separate
The first option is an AI training setting, which allows or prevents content from being used to train AI models. The second is an AI grounding setting, which allows or prevents content from helping generate AI answers. The third is limited grounding, which allows content to support Microsoft’s own AI products but not its third-party grounding services, set independently.
That separation is the point of the update. A publisher could prevent Microsoft from using its articles to train models while still allowing Copilot to cite them in answers, keeping its AI visibility. The reverse is also possible, but it carries a cost: blocking AI answers from using content can reduce AI visibility, even though ordinary search rankings stay unchanged.
The rollout calendar
Microsoft set out a timetable that gives publishers several months to prepare. Temporary grounding controls become available in Bing Webmaster Tools on December 15, 2026, for eligible sites that are using the older NoArchive or NoCache directives. Those temporary controls are a bridge for sites that already restrict AI use and need time to move to the new system.
Microsoft begins honouring the new preferences as it crawls pages on February 1, 2027. Publishers using the old directives have until March 1, 2027, to configure the new preferences to maintain their restrictions, and temporary grounding preferences expire on September 30, 2027. Missing the March deadline risks losing restrictions that publishers assumed were still in place.
Part of a shared industry move
Bing’s announcement is part of a wider shift toward a shared AI training preference in robots.txt, and Microsoft is going further than a single shared token. Alongside the joint signal, Bing adds granular, page-level training preferences and a separate grounding preference. Google announced adoption of the same shared proposal on the same day, with OpenAI and Anthropic also on board.
Bing Webmaster Tools already offers AI performance reporting that shows citations, cited URLs and grounding queries across supported AI experiences. That reporting becomes more useful once publishers can act on what they see: a site grounding answers it would rather not feed can now simply adjust the grounding setting directly, instead of relying on a broad block.
Why It Matters
For publishers, the update turns a single awkward decision into precise ones. Training and grounding are different uses of content, and treating them as one choice forced publishers to give up visibility they wanted or accept training they opposed. Bing’s split lets a newsroom stay citable in Copilot answers while keeping its reporting out of training data entirely.
The change lands as Bing’s AI answers become a larger surface. Sponsored results have already been spotted inside Bing AI Overviews, showing that AI answers are becoming commercial as well as informational. As that surface grows, the controls around what feeds it matter more, and early adopters will be better placed when the new system takes effect in 2027.
Background
The new preferences build on Bing’s existing controls for managing how website content appears in AI-generated answers, adding granularity rather than starting from scratch. Microsoft announced the changes through its Bing blog, with full documentation in Bing Webmaster Tools, and the company said the controls are designed to keep sites eligible for Bing search results whatever preferences they ultimately choose.
Source: Roundtable report.
0 Comments