Nosnippet and Data-Nosnippet: What AI Can Quote

By Muntazir · Published 11 October 2026

On 13 February 2024, Bing’s Fabrice Canel wrote that there were “no plans to provide support for this in the near future” for the data-nosnippet attribute. Twenty months later, on 15 October 2025, Bing’s webmaster blog announced it had shipped exactly that — and tied the attribute not just to search snippets but to Copilot’s AI-generated answers. A feature Bing had no interest in became, within two years, a control Bing specifically sells as reaching an AI product.

That reversal is a useful way into a pair of directives most site owners have never had to think about, because until recently neither one did much that mattered.

What do nosnippet and data-nosnippet actually control?

nosnippet is a whole-page switch. Set it as a <meta name="robots" content="nosnippet"> tag or as an X-Robots-Tag HTTP header, and Google withholds the text snippet and any video preview for that page across web search, Images, Discover, AI Overviews and AI Mode. A static thumbnail can still show. The page keeps its ranking position — it just appears with nothing underneath the link.

data-nosnippet is narrower and newer. It’s an HTML attribute, not a meta tag: <span data-nosnippet> or <div data-nosnippet> around one passage, and only that passage is excluded. The rest of the page can still generate a normal snippet. Google’s robots meta tag reference notes the attribute’s value is ignored — its presence alone is the instruction — and warns against toggling it with JavaScript after the page loads, since a crawler’s extraction can happen before or after rendering.

Neither directive is new. Google published the whole-page version years ago. What’s new is how deliberately both have been pulled into the AI-answer conversation since. A SERP preview tool shows what a page’s snippet looks like today — it won’t show what an AI Overview does with the same page, because that’s a separate pipeline with its own rules, covered next.

Why would a page want to hide a snippet and stay indexed?

Ranking and snippet are two separate wins, and most SEO advice conflates them. A business with a paywalled analysis, a page of legal disclaimers sitting above the useful content, or product copy still running an A/B test has a reason to stay indexed and rankable while keeping that specific text out of a public preview. nosnippet is the blunt version of that choice — the whole page loses its description. data-nosnippet is the version that only costs the marked paragraph.

Bing’s own announcement names the same use cases: subscriber-only tables, cookie notices, volatile reviews a business doesn’t want frozen into a snippet, and copy that’s mid-experiment. None of that is new strategy. What changed is that the tool for it now reaches further than a visible search result.

Does nosnippet reach AI Overviews — and does data-nosnippet reach them the same way?

For nosnippet, Google says so directly. Its robots meta tag reference lists AI Overviews and AI Mode alongside web search, Images and Discover as surfaces the directive covers, and adds that it also stops the page’s content from being used “as a direct input” to those two features — a stronger claim than just hiding a snippet. A page marked nosnippet isn’t just invisible in the preview; Google states it won’t feed that page’s text into the generated answer either.

The data-nosnippet section of the same documentation makes no equivalent statement. It describes excluding the marked passage from search result snippets. It says nothing about whether an AI Overview drawing on the rest of the page would still pull from the excluded span, or whether the same input restriction that applies to nosnippet applies here too. That’s not a minor omission — it’s the exact question a site owner marking out a paywalled section would want answered, and the documentation that explicitly closes this gap for the whole-page tag simply doesn’t address the partial one. Nothing published since closes it either.

What does Bing’s support actually add — and where’s the daylight with Google?

Bing’s attribute support, as described in its own announcement, excludes marked content from both Bing Search snippets and Copilot answers while leaving the page indexed and ranking-eligible — a clean, stated answer to the question Google leaves open. Bing frames the AI-answer coverage as part of the feature from day one, not a later clarification bolted onto an older search-only control.

That’s one engine answering the question the other hasn’t. It’s also one more reason not to treat “an AI tool respects this directive” as a single fact about the web — Google and Bing can, and currently do, document different scopes for what look like parallel features. The same per-crawler gap shows up elsewhere: Google-Extended governs AI training, not search ranking, and it took a separate, explicit line in Google’s docs to settle that — nothing about one AI-related control’s scope can be assumed from another’s.

What does a site’s own markup actually show, and what doesn’t it show?

Reading a page’s robots meta tag for a noindex directive is a narrow, mechanical check — the kind that site auditing tools, this one included, run routinely. What a check like that typically reports is whether a page is excluded from the index at all. A nosnippet value sitting in the same meta tag, right next to noindex, usually isn’t the question that check was built to answer, because withholding a snippet while staying indexed is a different outcome than being excluded altogether — and a scan built to answer “can this page be indexed” has no particular reason to also flag “will this page show a description.” A site owner relying on that kind of indexability result alone would see nothing wrong with a page that renders in search with no snippet at all.

data-nosnippet sits even further outside what a typical crawl-based audit looks for: it’s an attribute on an arbitrary element deep in the page body, not a document-level signal like a meta tag or header, so a check built around page-level directives has no reason to go looking for it there.

None of that means the directives aren’t working. It means the layer most audits check — can this page be found at all — isn’t the layer where a missing or misapplied snippet control shows up. Those are two different questions, the same split that runs through most AI search optimization work: being findable and being quoted correctly are not the same project, and a checklist built for one rarely covers the other. Conflating them is exactly how a nosnippet tag sits unnoticed on a page for months.