TL;DR
- Microsoft Bing said on 30 October 2026 that new AI Training and page-level AI Grounding preferences replace NoArchive and NoCache, honoured as pages are crawled from 1 February 2027.
- After 1 March 2027, NoArchive and NoCache stop counting as training or grounding preferences; temporary grounding settings open in Bing Webmaster Tools on 15 December 2026.
- The new DisallowExtended option keeps content in AI Mode, AI Overview, Copilot and Microsoft 365 grounding but out of third-party services such as Web IQ and the Foundry grounding API.
- Temporary grounding preferences expire on 30 September 2027 or at recrawl; Microsoft says none of the settings change eligibility for traditional Bing results.
Microsoft Bing announced on 30 October 2026 that it is replacing NoArchive and NoCache as its AI controls with two separate preferences, one for AI training and one for AI grounding, and that sites still relying on the old tags must act by 1 March 2027 to keep their choices. The new grounding control works page by page and adds a middle option, DisallowExtended, which keeps content available to Microsoft's own AI experiences but not to the grounding services Microsoft provides to third parties. The details come from a post on the Microsoft Bing Blog by Krishna Madhavan, Principal Product Manager at Microsoft Bing.
The training half of the announcement uses the shared IETF robots.txt token that Google also said on the same day it plans to honour; RA covers that token in a separate article on Google's AI training opt-out. This article is about the part only Microsoft announced: the grounding preference, the page-level syntax, and the migration away from NoArchive and NoCache.
What Microsoft changed
Microsoft says it was the first company to give publishers controls for AI training and grounding, in 2023, through the NoArchive and NoCache tags. The new preferences replace those two tags for training and grounding purposes. In Microsoft's words, they let publishers "make choices about training, grounding, or both while remaining eligible for Bing search results".
The announcement covers web content collected through Microsoft's crawlers and used for training or to inform generated responses, including web-grounded responses in Bing, Copilot and Microsoft 365.
The two preferences are independent. Setting one does not change the other, and both default to Allow when nothing applicable is published:
- AI Training uses the shared preference in robots.txt for a whole site or a path. Microsoft also extends it to individual pages, and a page-level training setting overrides the robots.txt setting.
- AI Grounding is a separate Microsoft preference set only at page level. There is no site-wide robots.txt grounding preference.
Microsoft's own example of a combination is AI Training set to Disallow with AI Grounding set to Allow. That keeps content eligible to inform generated responses without letting it train generative models.
The three grounding options, including DisallowExtended
Microsoft defines grounding as using content as input to help inform AI-generated responses. It says indexing and ranking pages in traditional search results is not grounding. A page can carry one of three grounding settings.
| AI Grounding setting | What Microsoft says it does |
|---|---|
| Allow (default, no action needed) | Content can be used to help inform generated responses |
| DisallowExtended | Content can inform responses in Microsoft-operated experiences such as AI Mode, AI Overview, Copilot and Microsoft 365, but not grounding services Microsoft provides to third parties, such as Web IQ and the grounding API in Foundry |
| Disallow | Content is not used to help inform generated responses |
DisallowExtended is the new piece. It lets a publisher accept citation in Microsoft's own answer products while declining reuse through the grounding services Microsoft offers to other companies. Microsoft names Web IQ and grounding through API access in Foundry as examples of the third-party services that are excluded. The post does not give a full list.
Microsoft attaches a warning to both restrictive options. Selecting DisallowExtended or Disallow "will significantly reduce your content's visibility in AI-generated experiences, without affecting eligibility for Bing search results". The post also says that allowing grounding makes content eligible to inform responses, "with links and citations providing a path to the original content", and that neither choice "guarantees inclusion, traffic, or payment".
How to publish the page-level settings
Microsoft gives two ways to set each preference on a page: a meta tag or an HTTP response header. Its examples, which all disallow, are:
- Training, meta tag: <meta name="Bingbot" content="ai-training:disallow">
- Training, header: X-Robots-Tag: Bingbot: ai-training:disallow
- Grounding, meta tag: <meta name="Bingbot" content="ai-grounding:disallow">
- Grounding, header: X-Robots-Tag: Bingbot: ai-grounding:disallow
For the middle option, Microsoft says to use ai-grounding:disallowextended in the same places. For training at site or path level, it points to the shared robots.txt group, "User-Agent: __usage_ai-training" followed by "Disallow: /" for the whole site or "Disallow: /example-path/" for a path. Allow is the default and needs no line, though an explicit "Allow: /" is optional.
The HTTP header route matters for files that cannot carry a meta tag, such as PDFs. That is RA's observation, not something the Bing post states.
What the new preferences do not change
The post lists several limits. They are the most likely source of mistakes, so they are worth reading closely:
- The preferences govern use of content, not crawl access. An Allow setting does not override a robots.txt rule that blocks Bingbot from crawling.
- They do not change eligibility for traditional Bing search results, subject to normal crawl and indexing requirements.
- They do not establish licensing or payment arrangements.
- Publishing a setting is not the same as activating it. Microsoft has to crawl the page to discover it. IndexNow can signal that a page changed and raise the chance of a recrawl, but it does not guarantee a recrawl date.
- The current Bing search syndication API returns links for traditional search scenarios and is not intended for AI-generated summaries, so the grounding preference does not affect it.
- Existing NoArchive and NoCache users should not assume their old restrictions carry over automatically.
The transition timeline
The dates below apply in full only to sites that use NoArchive or NoCache. Microsoft says all other sites need no transition action.
| Date | What happens, per Microsoft |
|---|---|
| 15 December 2026 | Temporary grounding transition preferences open in Bing Webmaster Tools, only for sites using NoArchive or NoCache on that date |
| 1 February 2027 | Microsoft starts honouring AI Training and AI Grounding preferences as pages are crawled |
| 1 March 2027 | Deadline. After this date NoArchive and NoCache are no longer read as training or grounding preferences |
| 30 September 2027 | Temporary grounding preferences expire, or earlier for any page Microsoft recrawls and finds a new AI Grounding preference on |
Training and grounding move differently through this transition. For training, a site publishes its preference in robots.txt or on pages by 1 March 2027. There is no temporary training setting in Bing Webmaster Tools. For grounding, a NoArchive or NoCache site has to do two things before 1 March 2027: publish the AI Grounding preference on its pages, and record a temporary grounding preference in Bing Webmaster Tools.
Microsoft explains why the second step exists. It may not retrieve new on-page preferences until it recrawls the site, and that recrawl may not happen between February and 1 March. The temporary preference holds the chosen grounding treatment for each page until Microsoft finds the new on-page setting or until 30 September 2027, whichever comes first.
Two fallback rules follow. If a site does not record a temporary grounding preference in Bing Webmaster Tools by 1 March, pages using NoArchive or NoCache default to Allow for grounding until Microsoft recrawls them and discovers the new preference. And once Microsoft detects any new AI Training or AI Grounding preference on a site, it disregards whatever NoArchive and NoCache tags remain there. Page-level settings published after 1 March 2027 can still take effect on recrawl, but Microsoft says a recrawl before the temporary preference expires is not guaranteed.
What the Bing post did not say
- The post only says NoArchive and NoCache stop being read as training or grounding preferences. The companion Bing Webmaster Tools FAQ (Q20) adds that Microsoft will go back to the original definitions of these terms in traditional search.
- It names Web IQ and the Foundry grounding API as examples of third-party grounding services but does not list the companies that use them.
- It gives no figures for how much visibility a restrictive grounding setting removes, only that the reduction will be significant.
- It says nothing about other search engines' grounding controls and nothing about any particular country, Thailand included.
Microsoft says a companion FAQ is published in the Bing Webmaster Tools help documentation.
What to check on your own site
This section is RA analysis based on the Microsoft post.
- Search templates and server config for noarchive and nocache, in both meta robots tags and X-Robots-Tag headers. If neither appears anywhere, Microsoft says no transition action is needed.
- If they do appear, check what they were added for. If the intent really is to keep content out of Copilot and Bing AI answers, the replacement is an explicit ai-grounding setting on each page.
- Choose between Disallow and DisallowExtended deliberately. DisallowExtended keeps Microsoft-operated AI experiences open, while Disallow closes all grounding use, at the visibility cost Microsoft warns about.
- Confirm the site is verified in Bing Webmaster Tools before 15 December 2026, since the temporary grounding preference is recorded there.
- Do not use a Bingbot crawl block to control AI use. It also removes pages from Bing search, and an Allow preference will not undo it.
- Set up IndexNow if it is not already running, keeping in mind Microsoft's caveat that it raises the chance of a recrawl but does not guarantee one.
Sites that want to stay citable in AI answers while limiting training can use the training and grounding split directly. RA's generative engine optimisation (GEO) work covers this kind of AI visibility setting.
What this means for Thai marketers
Microsoft's post does not mention Thailand. The controls live on each site's own pages and headers, so they apply to Thai sites the same way they apply anywhere. Thai publishers and ecommerce sites that added noarchive in the past, for any reason, should check those pages before 1 March 2027. After that date the tag no longer works as a training or grounding preference, and, unless a temporary grounding preference was recorded in Bing Webmaster Tools by then, grounding falls back to Allow until Microsoft recrawls and finds a new setting. Teams already working on AI SEO for Copilot and Bing answers will want the grounding setting decided page by page rather than inherited from an old template.
FAQ
What is the deadline for replacing NoArchive and NoCache in Bing?
1 March 2027. Microsoft says that after that date NoArchive and NoCache will no longer be read as preferences for training or grounding, and that grounding restrictions on NoArchive or NoCache pages need both an on-page AI Grounding preference and a temporary preference in Bing Webmaster Tools before then.
What does DisallowExtended do?
It limits grounding to Microsoft-operated experiences. Microsoft lists AI Mode, AI Overview, Copilot and Microsoft 365 as examples where the content can still inform responses, and Web IQ and grounding through API access in Foundry as third-party services where it cannot.
Will blocking AI grounding remove my pages from Bing search?
No, according to Microsoft. The post says the preferences do not change eligibility for traditional Bing search results, subject to normal crawl and indexing requirements, although restrictive grounding settings will significantly reduce visibility in AI-generated experiences.
Can I set an AI grounding preference in robots.txt?
No. Microsoft says AI Grounding is a page-level preference only, set with a meta tag or an X-Robots-Tag header, and that there is no site-wide robots.txt grounding preference. Only AI Training can be set in robots.txt.
My site never used NoArchive or NoCache. Do I need to do anything?
Not for the transition. Microsoft says sites that do not use NoArchive or NoCache need no transition action, and both preferences default to Allow. Publishing a Disallow or DisallowExtended setting is optional and takes effect as Microsoft crawls, from 1 February 2027.
Microsoft has split one old tag into two clear choices, with a firm date for sites still using it. If you want a check of which pages carry noarchive or nocache today and what each one should become, RA's GEO team can run that review with you.
