Bing Grounding Preference: DisallowExtended Explained

The switch you went looking for#

I read Microsoft's documentation looking for a way to limit Copilot's use of long-form articles while keeping those pages in Bing, and the value that does that is DisallowExtended. The easy mistake is reaching for the crawling control when you only wanted to change grounding, because if you block Bingbot in robots.txt you do not just leave Copilot, you leave Bing.

What grounding means for Bing#

A language model on its own only knows what it absorbed in training. Grounding is the extra step where the assistant runs a search, pulls a handful of current pages, and writes its answer from them. Copilot is one of the Microsoft-operated experiences that does this, and Microsoft also provides grounding services to third parties, such as Web IQ and the Foundry grounding API.

That matters because any page Bing has indexed is a candidate for being quoted, whether or not the reader ever visits it. It is also why grounding is a different question from training. Training is a one-off ingestion of content into a model. Grounding happens again on every query, so it decides whether your page gets used, and cited, today.

The controls Microsoft documents#

In October 2026 Bing published Your content, your choice: clearer preferences for training and grounding, and it changes what this page is about. Bing now has two separate preferences. AI Grounding takes the values Allow, DisallowExtended and Disallow, and it is page-level only, set with a meta tag or an X-Robots-Tag header. AI Training is the other one, and it is the only one that can also go in robots.txt, using User-Agent: __usage_ai-training with Disallow. Bing says the restrictive grounding values reduce AI visibility but do not change eligibility for Bing search results, so neither one removes you from Bing search. Bing Webmaster's help page, New AI Preferences, covers the same settings.

Setting	Bing search results	Microsoft-run experiences	Third-party grounding	How to set it
AI Grounding: Allow	Eligible	Can use the page	Can use the page	Default, nothing to add
AI Grounding: DisallowExtended	Eligible	Can use the page	Blocked (Web IQ, Foundry grounding API)	<meta name="Bingbot" content="ai-grounding:disallowextended"> or the header form X-Robots-Tag: Bingbot: ai-grounding:disallowextended
AI Grounding: Disallow	Eligible	Not used (content is not used to inform generated responses)	Not used (content is not used to inform generated responses)	The same meta tag with the value disallow, or the header X-Robots-Tag: Bingbot: ai-grounding:disallow
Disallow for bingbot	Not crawled, so it drops out	Not crawled	Not crawled	robots.txt, a crawl block and not a preference

What changes on Mar 1, 2027#

If you rely on noarchive or nocache today, act before Mar 1, 2027, either by moving to the ai-grounding meta tag or by recording a temporary preference in Bing Webmaster Tools. After Mar 1, 2027, noarchive and nocache are no longer read as preferences, and a site that uses them but has recorded no temporary grounding preference defaults to Allow. The temporary preferences expire on Sep 30, 2027, so I'd put the ai-grounding tag on the pages themselves well before then instead of depending on the temporary setting. The full timeline is below.

Choosing between the values is a trade-off, and Bing is plain about it: DisallowExtended and Disallow significantly reduce visibility in AI experiences. DisallowExtended is the narrower one, because Microsoft's own products can still use the page while third-party services cannot. If you want to stay in Microsoft's own answers but keep your content out of grounding services that other companies build on, I'd start there. Disallow is the stronger choice, and I'd keep it for pages where you accept losing that visibility.

DisallowExtended is a value of the page-level grounding preference, so you write it in a meta tag or an X-Robots-Tag header, and there is no robots.txt rule or site-wide switch for it. The preference also does not override a robots.txt block on Bingbot, because a page Bing cannot crawl never gets as far as having a preference read. As for reporting, I did not find any documentation of Bing Webmaster Tools showing that a preference was read, so I will not claim it does. What Bing Webmaster Tools does hold is the temporary grounding preferences for sites that use the old tags, which open on Dec 15, 2026.

Legacy until Mar 1, 2027#

In September 2023 the Bing Webmaster blog announced new options for webmasters to control usage of their content in Bing Chat, the product now called Copilot. It used two existing robots meta tags. These are being replaced by the preferences above, and after Mar 1, 2027 they are no longer read.

Legacy tag	Where it goes	Bing search results	Grounded answers in Copilot	Training Microsoft's models
nocache	Robots meta tag or X-Robots-Tag header	Eligible	URL, title and snippet only	URL, snippet and title may still be used
noarchive	Robots meta tag or X-Robots-Tag header	Eligible	Not included	Kept out of training

A page that carries both tags is treated as nocache. nocache does not keep a page out of training, because Bing's 2023 post says the URL, snippet and title may still be used. Only noarchive keeps a page out of training.

How is this different from Google-Extended?#

Google's version of this choice has a different shape. Google-Extended is a robots.txt user agent token, so it works at the site level, where Bing's grounding preference works page by page. We go through what it does and does not affect, including controls like nosnippet, in Does blocking AI crawlers hurt your Google rankings?.

Bing has no robots.txt token for grounding, and only its AI training preference can go there.

Setting it up without leaving Bing#

This is the order I would follow on a real site, because it changes the least and is easiest to undo.

1. Decide which sections should limit grounding. Product docs you want cited everywhere might stay on the default Allow, while long-form articles you depend on for clicks might get DisallowExtended.
2. Add the tag to those pages, either in the HTML head or as an HTTP header (examples below).
3. Leave Bingbot allowed in robots.txt, because the preference does not override a robots.txt block on Bingbot.
4. Submit the changed URLs through IndexNow. It prompts a recrawl but does not guarantee one, and the preference applies from Feb 1, 2027 as pages are crawled.
5. If your pages use noarchive or nocache, record a temporary grounding preference in Bing Webmaster Tools once it opens on Dec 15, 2026, and do it before Mar 1, 2027.

<!-- in the <head> of a page -->
<meta name="Bingbot" content="ai-grounding:disallowextended">

<!-- or, as an HTTP response header -->
X-Robots-Tag: Bingbot: ai-grounding:disallowextended

# robots.txt: keep Bing crawling
User-agent: bingbot
Allow: /

If you also want to see what other AI bots are reading your site, our AI crawlers list has every user agent name and what it is used for.

How do you know it worked?#

Start with what you can verify today. Run curl -I https://yoursite.com/page to confirm the header is being sent, or view the page source to check the meta tag made it past your CMS or CDN. That shows you what Bing will receive when it crawls the page.

What you cannot test yet is Copilot's behaviour. The new preferences are not live until Feb 1, 2027 and only apply after Bing recrawls the page, so a question to Copilot today would tell you nothing about the new tag. I also would not expect Bing Webmaster Tools to confirm that a preference was read, because no source I found shows that.

The free AI readiness check tests your robots.txt against AI crawlers such as OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot. It does not name Bingbot or check Bing's meta preferences, so curl and view-source remain the way to check this one.