Microsoft has introduced new Bing AI content controls that let publishers decide separately whether their content can train AI models and whether it can help generate AI answers. Announced on October 30, the settings are described by Krishna Madhavan, Principal Product Manager at Microsoft Bing, as newer, simpler and clearer preferences. Google, OpenAI and Anthropic are also adopting similar AI training controls.
Training and Grounding, Handled Separately
The change splits two things that were often lumped together. AI training covers whether Microsoft can use a site's content to train AI models. AI grounding covers whether that content can help generate AI answers. A third option, limited grounding, lets Microsoft's own AI products use the content but not its third-party grounding services.
That separation means a publisher can stop Microsoft from training on its articles while still letting Copilot cite them. Blocking grounding can reduce visibility in AI experiences, though traditional Bing rankings are not affected.
Key Dates
- December 15, 2026: Temporary grounding controls open in Bing Webmaster Tools for eligible sites that use the older NoArchive or NoCache directives.
- February 1, 2027: Microsoft starts honoring the new preferences as it crawls pages.
- March 1, 2027: Deadline for publishers using the old directives to set the new preferences if they want to keep their restrictions. Sites that have not recorded a temporary grounding choice by then are reported to default to allowing grounding until recrawled.
- September 30, 2027: Temporary grounding preferences expire.
Part of a Wider Shift
The controls build on the AI Performance report in Bing Webmaster Tools, which shows citations, cited URLs and grounding queries across supported AI experiences. Secondary coverage also points to a shared robots.txt signal for opting out of AI training, backed by Google, Microsoft, OpenAI and Anthropic. It is an individual IETF draft rather than a finished standard, and it works as a preference, not an access control, so crawlers that ignore it are not stopped.
What Publishers Should Do
Start by finding every NoArchive and NoCache tag on your site, including those added by CMS plugins or CDNs. Decide training and grounding separately for each content type, since the visibility trade-offs differ. Publish page-level settings early and record any temporary grounding preference before March 1, because a setting does not take effect until the page is recrawled. Keep crawler-specific rules and WAF or CDN protections in place for real access control.