Bing and Copilot: what noindex, noarchive, nocache, nosnippet and data-nosnippet do to AI answers

In Bing, noindex keeps a page out of Bing search, Copilot and grounding API results. noarchive keeps the page out of Copilot answers and grounding results, and nocache limits Copilot to the page's URL, title and snippet. nosnippet and data-nosnippet stop Bing from showing a caption, which may limit the quality of Copilot's citations. Section 10 of the Bing Webmaster Guidelines states all 5 effects.

What each directive does in Bing and Copilot

The Copilot column is §10 of the guidelines. The other 2 columns are Bing's list of the meta tags and attributes it supports.

When a page carries both noarchive and nocache, Bing's page says it treats the page as nocache. That page still appears in Copilot, with its URL, title and snippet.

Choosing between noarchive and nocache

The 2 directives answer different questions. noarchive takes the page out of Copilot answers and grounding results and leaves it in Bing search. nocache keeps the page in Copilot as a thin citation: a link, a title and a snippet, with no body text behind the answer. Bing's meta tag page also lists “Do not link in Chat and Copilot” under noarchive.

Pick by what you want an answer to do with the page:

• Use it fully: set neither. §10 says to avoid nocache on content meant for Copilot or grounding, because it reduces citation depth.
• Cite it by URL, title and snippet only: set nocache.
• Leave it out of Copilot, and keep it in Bing search: set noarchive.
• Leave it out of Bing entirely: set noindex.

Bing's meta tag page adds a training effect to both. noarchive keeps the content out of training Microsoft's generative AI foundation models, and nocache limits that training to the URL, title and snippet.

nosnippet, max-snippet and data-nosnippet

nosnippet removes the description and the preview thumbnail for the whole page. data-nosnippet does the same for one element, such as <div data-nosnippet>. max-snippet:[number] caps the snippet length in characters: 0 shows no text snippet, and -1 sets no limit. §10 warns that nosnippet and data-nosnippet may limit Copilot's citation quality.

§10 also says to “use the data-snippet attribute” to mark the text Bing may show or cite. Bing's list of supported tags does not include a data-snippet attribute, and none of the Bing pages cited here defines its syntax. Treat it as unspecified until Bing documents it.

How to write the directives

Put a meta tag in the document head, and separate several values with commas. Use name="bingbot" instead of name="robots" to address Bing alone. For a PDF or another file with no head, send the same values in an X-Robots-Tag response header. Bing accepts the header for each of these tags.

<!-- Option 1: out of Copilot answers, for every engine that reads the tag -->
<meta name="robots" content="noarchive">

<!-- Option 2: Bing only, and Copilot may cite the URL, title and snippet -->
<meta name="bingbot" content="nocache">

Mark a part of the page in its body:

<p>Text Bing may quote.</p>
<div data-nosnippet>Text Bing should not quote.</div>

X-Robots-Tag: noindex

Why robots.txt does not keep a page out of Copilot

§10 opens with the rule that robots.txt controls crawl access, not indexing. Bing's meta tag page adds that Bing must be able to crawl a page to see its noindex. A page blocked in robots.txt hides its own noindex from Bingbot.

To keep a page out of Copilot, let Bingbot fetch it and put noindex or noarchive in the page. §8 of the guidelines lists blocking Bingbot in robots.txt among the things to avoid. Crawl rules for AI crawlers are a separate decision, covered in AI crawler user agents in robots.txt.

What Google does with the same tags

Google's meta tag specification lists noarchive and nocache under historical rules that Google Search does not use. Those 2 are Bing levers only. nosnippet works in both engines. Google says it also stops the content being used as a direct input for AI Overviews and AI Mode.

There are 2 smaller differences. Google reads data-nosnippet only on span, div and section elements. Google treats max-snippet:0 as equivalent to nosnippet, while Bing's page says only that it shows no text snippet.

How to check a page with curl

Read the meta tags, the header and the data-nosnippet count:

URL=https://example.com/
curl -sS "$URL" | grep -Eio '<meta name="(robots|bingbot|googlebot)"[^>]*>'
curl -sS -o /dev/null -D - "$URL" | grep -i '^x-robots-tag'
curl -sS "$URL" | grep -o 'data-nosnippet' | wc -l

This site's homepage prints nothing for the first 2 commands and 0 for the third. Its 404 page carries a noindex:

SITE=https://agenticwebscore.com
curl -sS "$SITE/no-such-page" | grep -Eio '<meta name="(robots|bingbot|googlebot)"[^>]*>'

<meta name="robots" content="noindex, nofollow">

The grep counts text, so a data-nosnippet inside a script or a comment counts too. Read the matches before you trust the number.

What AgenticWebScore reports

Two checks read these directives on the homepage the scan already holds, so neither makes a request of its own.

Homepage indexable is a scored check in the essential tier of the web-standards part. It fails on noindex, or on none, in a robots, bingbot or googlebot meta tag or in X-Robots-Tag. A homepage with fewer than 25 words of visible text is not measured. A short page served with 200 may be a challenge page with its own noindex.

Snippet and AI grounding directives is an unscored check. It names each of noarchive, nocache, nosnippet, max-snippet:0 and data-nosnippet it finds, where it found it, and the effect above. It never fails a site, because each directive is the owner's choice.

In the scan records of 20 October 2026, none of the 5 measured homepages carries noindex, noarchive or nocache. Resend marks 5 elements with data-nosnippet. Vercel states index, max-image-preview:large, which asks for nothing to be left out.

The methodology explains how an unscored check is reported. To see what your homepage states to Bing and Google, scan your site.