Microsoft announced on October 30, 2026 that Bing will read separate preferences for AI training and AI grounding, replacing the old NoArchive and NoCache tags for AI use. New settings are honored as pages are crawled from February 1, 2027. If you rely on NoArchive or NoCache to keep content out of AI answers, you have until March 1, 2027 to switch.
"Grounding" means using your page to help generate an AI answer, as in Copilot or Bing's AI features. "Training" means using it to build or update a model. Until now Bing users mostly had blunt tools for both. This article sets out the new controls, the dates and a practical checklist.
Key takeaways
- Microsoft announced separate AI training and AI grounding preferences for Bing on October 30, 2026, replacing NoArchive and NoCache for AI use.
- Training uses the robots.txt token __usage_ai-training or a page-level ai-training:disallow; grounding is page-level only, with Allow, DisallowExtended and Disallow values.
- New preferences are honored as pages are crawled from February 1, 2027; after March 1, 2027 NoArchive and NoCache no longer count for training or grounding.
- Both settings default to Allow, and neither affects eligibility for traditional Bing search results.
What happened
In a Microsoft Bing Blog post titled Your content, your choice: clearer preferences for training and grounding, dated October 30, 2026, Principal Product Manager Krishna Madhavan described separate preferences that do not affect eligibility for Bing search results. The post states: "Publishers should not have to choose between being discoverable in Bing and deciding how generative models use their work."
The timeline:
| Date | What happens |
|---|---|
| October 30, 2026 | Microsoft announces the new AI training and AI grounding preferences |
| December 15, 2026 | Temporary grounding transition preferences open in Bing Webmaster Tools, only for sites using NoArchive or NoCache on that date |
| February 1, 2027 | New preferences take effect as pages are crawled |
| March 1, 2027 | Deadline to publish grounding preferences and record the temporary Webmaster Tools preference; after this date NoArchive and NoCache are no longer read for training or grounding |
| September 30, 2027 | Temporary preferences expire, or earlier if a page is recrawled and its new preference is found |
The two new controls
AI training
Training can be set for the whole site, a folder, or a single page.
- robots.txt (site or path): Bing uses the shared user-agent token
__usage_ai-training, the same token defined in the IETF draft that Google also plans to support. Example:
User-Agent: __usage_ai-training
Disallow: /
- Page level, meta tag:
<meta name="Bingbot" content="ai-training:disallow"> - Page level, HTTP header:
X-Robots-Tag: Bingbot: ai-training:disallow
Microsoft says a page-level setting overrides robots.txt. We cover the shared token and Google's plans in more detail in the AI training robots.txt signal.
AI grounding
Grounding is page level only. There is no site-wide robots.txt setting for it. You use the same meta tag or X-Robots-Tag header formats with one of three values:
| Value | What it means |
|---|---|
| Allow (default) | Your content may inform generated responses |
ai-grounding:disallowextended |
Grounding is limited to Microsoft-operated experiences (AI Mode, AI Overview, Copilot, Microsoft 365) and excluded from third-party grounding services such as Web IQ and Foundry API grounding |
ai-grounding:disallow |
Your content is not used for generated responses |
For example, a page that should only appear in Microsoft's own AI experiences would carry <meta name="Bingbot" content="ai-grounding:disallowextended">.
What happens to NoArchive and NoCache
The new controls take over the AI role that NoArchive and NoCache played. Microsoft's post says:
- Both tags are replaced for training and grounding purposes.
- If Microsoft detects the new AI Training or AI Grounding preferences on a site, it disregards NoArchive and NoCache on that site.
- After March 1, 2027, "NoArchive and NoCache will no longer be read as a preference for training or grounding."
- Sites that use these tags and do not record a temporary grounding preference in Bing Webmaster Tools by March 1, 2027 default to Allow for grounding until their pages are recrawled.
That last point is the one to act on. If you rely on NoArchive today and do nothing, your pages could become eligible for AI grounding again.
What does not change
Microsoft's post lists several things that are unaffected or limited:
- Eligibility for traditional Bing search results. Indexing and ranking for traditional search are not treated as grounding.
- The current Bing search syndication API is not affected by the grounding preference.
- An Allow setting does not override a robots.txt rule that blocks Bingbot from crawling.
- Preferences do not create licensing or payment arrangements, and neither choice guarantees inclusion, traffic or payment.
- A preference does nothing until Bing crawls the page. IndexNow may increase the chance of a recrawl but does not guarantee a date.
Both preferences "default to Allow when no applicable preference is published," so sites that are happy to appear in AI answers and training do not need to add anything.
Who this affects
- Publishers and content businesses using NoArchive or NoCache to limit AI use. You have the most to do.
- Businesses that want AI visibility, such as local services, ecommerce and SaaS. Check that nothing in your templates or CDN adds NoArchive, NoCache or the new disallow values by accident.
- Agencies and developers who manage meta robots tags,
X-Robots-Tagheaders or robots.txt for clients. - Sites whose content is licensed or paywalled, where AI training and grounding may need different answers.
What to do now
- Decide your policy per section. Training and grounding are separate questions. A publisher might block training but allow grounding in Microsoft's own products; a retailer might allow both.
- Find existing NoArchive and NoCache tags. Check key templates with our meta tag checker, and check server responses for
X-Robots-Tagheaders with our HTTP headers checker. CMS plugins and CDNs sometimes add these headers without anyone noticing. - Add the new tags to templates, not pages one by one. Grounding has no robots.txt option, so put the meta tag or header in your page templates or server configuration.
- Use robots.txt for training if it suits you. Add a
__usage_ai-traininggroup and test specific URLs with our robots.txt tester. Make sure you do not accidentally block Bingbot itself. - If you use NoArchive or NoCache, plan for December 15. From that date, eligible sites can record a temporary grounding preference in Bing Webmaster Tools. Our Bing Webmaster Tools setup guide covers verification if you have not added your site yet.
- Encourage recrawling. After you change tags, submit updated URLs through IndexNow. Our guide to automating indexing with IndexNow and sitemaps explains the setup.
- Recheck after February 1, 2027. Confirm the live pages serve the tags you expect and watch Bing Webmaster Tools reporting. Our post on Bing's Topics Portfolio in AI Performance covers the AI reporting already available.
You can review which AI user agents your robots.txt currently allows with our AI crawler checker. The free tools at /tools need a free account.
Frequently asked questions
What is AI grounding in Bing?
Grounding is when Bing uses your page to help generate an AI response, for example in Copilot, AI Mode or AI Overview. Microsoft treats it separately from training a model and from ranking in traditional search results. Under the new controls you set grounding with a page-level meta tag or X-Robots-Tag header.
Will blocking AI grounding remove my site from Bing search?
No. Microsoft says the new preferences do not affect eligibility for traditional Bing search results, and indexing and ranking for traditional search are not treated as grounding. Blocking grounding can reduce your visibility inside AI answers, though.
Does NoArchive still work for Bing AI?
Until March 1, 2027, Bing still reads NoArchive and NoCache, unless it finds the new preferences on your site, in which case it ignores the old tags. After March 1, 2027, those tags are no longer read as a preference for training or grounding. Switch to the new ai-training and ai-grounding values before then.
What is the difference between Disallow and DisallowExtended?
Disallow keeps your content out of generated responses. DisallowExtended allows grounding only in Microsoft-operated experiences such as AI Mode, AI Overview, Copilot and Microsoft 365, and excludes third-party grounding services such as Web IQ and Foundry API grounding.
Can I block Bing AI grounding in robots.txt?
No. Microsoft says there is no site-wide robots.txt setting for grounding. Use the Bingbot meta tag or the X-Robots-Tag header on each page, ideally through your templates. robots.txt can only carry the AI training preference.
When do the new Bing AI preferences take effect?
Microsoft says new preferences are honored as pages are crawled from February 1, 2027. Publishing a tag earlier is fine, but it only takes effect once Bing recrawls the page after that date.
Sources
More guides for your industry
Free checklist (PDF)
Free marketing agencies website checklist
31 practical checks for marketing agencies: website speed and mobile, Google Business Profile and reviews, email security, booking and enquiry handling, and AI search visibility. Each one says why it matters, and many link to a free tool that checks it.
Written by the MakeItFlow team
We build the website audit these guides are based on. Every fact is checked against the primary sources listed above . How we research and check
Check your own site in 30 seconds
Run a free audit covering speed, SEO, DNS, email authentication and security headers — and get a prioritised fix list.
