Can `nosnippet` keep me out of AI Overviews
Google documents nosnippet blocks Search snippets and direct AI Overview input. Not a surgical Overview opt-out, and it still kills your blue-link preview.
William Spurlock Founder — Spurlock Studios 29 MIN
Yes — within the wording Google actually publishes. The robots meta spec says nosnippet applies to web search, Google Images, Discover, AI Overviews, and AI Mode, and that it prevents the content from being used as a direct input for Overviews and AI Mode. That is documented. It is not a surgical “Overview only” switch. Classic Search still loses the text snippet (a static thumbnail may remain). Google does not document that your brand can never appear in an Overview assembled from other sites, licensed content, or structured data it treats as a separate permission.
If you ranked and vanished from Overviews, you probably did this by accident. That diagnostic lives in ranked but missing AI Overviews. This spoke is the inverse: you asked for an opt-out, and you need the cost in writing. Strategy sits in the Answer Engine Optimization playbook.
The short answer
- Google documents
nosnippetas a snippet and direct-input control for Search surfaces that include AI Overviews and AI Mode. Quote the spec. Do not upgrade it to folklore. - You also pay classic Search: no text snippet, no video preview. Rankings can survive. The blue-link preview dies.
max-snippet:0is equivalent tonosnippet. A tinymax-snippetstarves usable quotation.data-nosnippethides listed HTML elements, not the whole URL.nositelinkssearchboxis unused. Google-Extended is Gemini training and listed grounding, not Search appearance. A GooglebotDisallowcan hide the meta tag so the opt-out never applies.- If you want out of generative units and you still want Search snippets, Search Console’s Search generative AI control is the documented property-level tool.
nosnippetis the blunt page-level tool.
What does Google actually document for nosnippet?
The robots meta page is the source, not a LinkedIn thread. nosnippet means: do not show a text snippet or video preview for this page in search results. A static image thumbnail may still be visible when Google thinks that helps the result. The same rule applies to all forms of search results at Google that the spec lists — web search, Images, Discover, AI Overviews, AI Mode — and will also prevent the content from being used as a direct input for Overviews and AI Mode.
The AI features page says the same family of tools in one sentence: to limit the information shown from your pages in Search, use nosnippet, data-nosnippet, max-snippet, or noindex.
| Claim you hear | What Google documents | What Google does not document |
|---|---|---|
nosnippet hides the Search snippet | Yes — no text snippet or video preview | That titles, URLs, or thumbnails also vanish |
| It applies to AI Overviews and AI Mode | Yes — listed with web search, Images, Discover | That Overviews are a separate robots vocabulary |
| It blocks “direct input” for Overviews / AI Mode | Yes — that phrase is in the spec | That your brand name can never appear via other sources |
| Rankings survive | Compatible with remaining indexed | A promise that position stays identical |
| It is an Overview-only opt-out | No | A switch that leaves classic snippets intact |
| It is instant | No — recrawl can take days to months | Same-day SERP cleanup as a SLA |
I have been SEO certified since 2021. I still open the spec before I tell a founder they “blocked AI.” If the sentence is not on Search Central or Search Console Help, it is a hypothesis. Put it in the notes column, not in the ticket title.
Where the rule lives, and which crawler it addresses:
| Placement | What it targets | Use when |
|---|---|---|
<meta name="robots" content="nosnippet"> | All crawlers that honor robots meta | Default page-level kill |
<meta name="googlebot" content="nosnippet"> | Google text results | You only want Google’s snippet off |
<meta name="googlebot-news" content="nosnippet"> | Google News | Full web snippet, no News snippet — not an Overview switch |
X-Robots-Tag: nosnippet | That HTTP response (HTML, PDF, media) | Non-HTML, or you cannot edit <head> |
Both robots and googlebot tags | Sum of negative rules for Googlebot | Avoid stacking unless you mean every restriction |
Google documents two user-agent tokens for the robots meta tag: googlebot and googlebot-news. Other name values are ignored. googlebot-news with nosnippet is a News-preview decision. It is not how you opt out of AI Overviews.
Does that guarantee I stay out of AI Overviews?
No. Google documented direct input from that page, not a brand-level invisibility cloak. Keep the hedge in the brief you send legal.
The AI-features page also says Overviews often do not trigger, that supporting links must be indexed and eligible to show in Search with a snippet, and that troubleshooting preview controls starts with URL Inspection plus waiting for recrawl. None of that is “type nosnippet and the feature can never mention you.”
Still possible after nosnippet on your URL | Why the spec leaves room | What to do if that is unacceptable |
|---|---|---|
| Classic result with title + URL, no text snippet | Snippet is blocked; indexing is not | That is the intended Search cost |
| Static thumbnail on the result | Spec says a thumbnail may remain | max-image-preview:none if you also need that |
| Overview cites a different site that copied you | Direct input is per page Google fetched | Take down scrapers; do not expect robots meta to police the web |
| Overview uses licensed or partner content | max-snippet explicitly does not interrupt separately granted permission | Contract review, not a meta tag |
| Rich result built from structured data | Robots snippet limits generally do not govern schema.org except noted description fields | Remove or shrink the JSON-LD you do not want shown |
| Brand mentioned without your URL | Models and other pages still exist | That is not a robots-meta problem |
| You still see the old Overview for a while | Caching and recrawl, days to months | URL Inspection, then wait; do not stack a second control on day two |
If leadership wants “guaranteed Overview opt-out,” write back: Google documented snippet suppression plus direct input prevention for that URL. Guaranteed never-mentioned is not in the spec. I will not sign a sentence Google did not publish.
What does data-nosnippet hide that page-level nosnippet does not?
data-nosnippet is an HTML attribute on span, div, and section only. It marks those nodes as ineligible for the textual snippet. It is a boolean: true, false, and omitted all mean the same thing — the element is out. Any other tag is invalid. An unclosed div swallows everything after it.
Google typically renders pages to index them, and rendering is not guaranteed. Extraction of data-nosnippet may happen before or after render. Do not add or remove the attribute on existing nodes with JavaScript after the fact. If you create nodes in JS, set the attribute when you insert them.
Structured data declared inside a data-nosnippet element remains usable for search results. Hiding the visible paragraph does not hide the JSON-LD.
| Pattern | Effect | Typical accident |
|---|---|---|
<span data-nosnippet> around a price | That span stays out of the snippet | Fine if the answer lives elsewhere |
<div data-nosnippet> around <main> | Most of the page is ineligible for snippets | Cookie-banner wrapper reused on content |
Unclosed <div data-nosnippet> | Everything after it is treated as excluded | Template merge error |
<p data-nosnippet> or a custom element | Invalid — spec requires span, div, or section | Component library default |
| Toggle attribute after hydration | Race with pre-render extraction | “We hid the FAQ in JS” |
| FAQ JSON-LD inside the hidden node | Schema can still feed rich results | Thinking the attribute sanitizes the page |
Use data-nosnippet when one block is the problem (comments, PII, a legal footer) and the rest of the URL should remain snippet-eligible. Use page-level nosnippet when the whole URL should stop feeding snippets and Overview direct input. Do not mix them as a mood.
Checklist before you ship an element-level hide:
- Attribute sits on
span,div, orsection - Tags close; view-source is valid HTML
- The answer you still want quoted is outside the wrapper
- JSON-LD does not republish the hidden sentences you care about
- No post-render JS mutates the attribute on an existing node
What does max-snippet change, including max-snippet:0?
max-snippet:[number] caps how many characters Google may use as a textual snippet. It does not cap image or video previews. The spec says this applies to web search, Images, Discover, Assistant, AI Overviews, and AI Mode, and that it limits how much of the content may be used as a direct input for Overviews and AI Mode.
Special values matter more than the marketing copy on the SEO plugin:
0— no snippet. Equivalent tonosnippet.-1— Google chooses length. Same family as omitting the rule.- A small positive number — a starved preview, and a starved Overview input budget.
- Omitted — Google chooses. This is the default for citation work.
The cap does not apply when the publisher separately granted permission — in-page structured data, or a license agreement with Google. Recipe carousel content can still appear even when the text preview is limited. To manage schema, edit the schema. Do not assume max-snippet sanitizes JSON-LD.
| Setting | Snippet in classic Search | Direct input for Overviews / AI Mode (documented) | Use when |
|---|---|---|---|
omit / max-snippet:-1 | Google chooses | No extra cap | You want to be findable and citable |
max-snippet:160 or similar | Short preview | Limited input | Legal wants a cap and you still leave a usable span |
max-snippet:20 | Nearly useless preview | Starved input | Almost never, unless the page should not be quoted |
max-snippet:0 | None (same class as nosnippet) | Prevented as with nosnippet | Intentional full preview kill |
nosnippet plus any max-snippet | Combined negative rules win | Same | Redundant; pick one |
I see max-snippet:0 left in Rank Math / Yoast global defaults after a 2023 legal review. Per-post UI looks clean. The template still ships the kill switch. Audit crawled HTML, not the CMS screenshot.
Worked combinations (comma list or two meta tags — same idea):
<meta name="robots" content="max-snippet:20, max-image-preview:large">
<meta name="robots" content="nosnippet">
<meta name="robots" content="max-snippet:0">| You intended | You shipped | What Google’s combining rule does |
|---|---|---|
| Short preview, large image | max-snippet:20, max-image-preview:large | 20-character text cap; image preview may still be large |
| No text snippet | nosnippet | No text snippet / video preview; thumbnail may remain |
| “Tone it down” | max-snippet:0 | Same class as nosnippet |
| News silent, web loud | googlebot-news nosnippet only | News snippet off; web snippet still allowed |
| Cap plus kill | max-snippet:50 and nosnippet | No snippet — negatives add |
If the number is missing or not readable as a number, Google says max-snippet is ignored. A plugin that writes max-snippet: with a blank value is not a cap. It is a no-op plus a false sense of control.
Does nositelinkssearchbox keep me out of Overviews?
No. Google lists nositelinkssearchbox under historical and other unused rules. Search no longer uses it to control a sitelink search box, because that feature no longer exists. The rule is ignored.
If a plugin still exposes the checkbox, it is a placebo. It will not keep you out of AI Overviews. It will not restore a sitelinks search box either.
| Rule people confuse with preview control | Status in the robots spec | Overview effect |
|---|---|---|
nosnippet | Current | Documented: no snippet; no direct Overview / AI Mode input |
max-snippet | Current | Documented length cap, including Overview / AI Mode input |
data-nosnippet | Current | Element-level snippet exclusion |
noindex | Current | Page not shown in Search — Overviews cannot use it as a supporting Search result |
nositelinkssearchbox | Unused / ignored | None |
noarchive | Unused (cached-link feature gone) | None |
nocache | Unused | None |
Delete the unused checkbox from the runbook. Leaving it “on” teaches the next hire that folklore is a control.
How is this different from noindex, robots.txt, and Google-Extended?
Four different jobs. Collapsing them into “block AI” is how teams either nuke Search or do nothing to Overviews.
noindex means do not show the page in search results. The AI-features page includes it in the list of ways to limit information shown from your pages. You also lose the blue link. That is a Search exit, not a snippet trim.
robots.txt Disallow means Googlebot must not crawl. Preview controls live in the HTML or X-Robots-Tag, which are discovered when the URL is crawled. If you Disallow the URL, Google never sees nosnippet. The spec is explicit: URLs that must obey indexing or serving rules cannot be disallowed from crawling.
Google-Extended is a robots.txt product token for listed Gemini training and grounding uses. Google documents that it is not a method for managing how content appears in Google Search. The AI-features page points you at Google-Extended only when the question is training and grounding in other Google systems, not Search snippets.
| Control | What it is for | Search snippets | AI Overviews / AI Mode (Search) | Gemini training / listed grounding |
|---|---|---|---|---|
nosnippet / max-snippet:0 | Page-level preview kill | Text snippet off | Documented: no direct input from that page | Not this token |
data-nosnippet | Hide specific HTML | Those nodes out of the snippet | Those nodes out of snippet-style extraction | Not this token |
noindex | Remove from Search | No result | Not eligible as a Search supporting link | Not this token |
robots.txt Disallow for Googlebot | Do not crawl | Google may lack a fresh rule | You also hid the meta tag | Different from Google-Extended |
Google-Extended Disallow | Gemini product uses | No Search-appearance change documented | Not a Search Overview switch | Yes — listed uses |
| Search Console generative AI control | Property-level generative features | Classic Search not the target | Documented exclude for listed generative features | Training still wants Google-Extended |
ChatGPT and Perplexity do not read your Google robots meta as a platform-wide kill switch. If buyers ask there, the comparison is ChatGPT vs Perplexity vs AI Overviews, not another nosnippet debate.
Should I use Search Console’s generative AI control instead?
If the actual goal is “stay in classic Search, leave listed generative features,” yes — that is the control Google documented for that split. nosnippet is the wrong tool for that sentence because it also kills the snippet.
Search Console Help for the Search generative AI control (checked 2026-09-05; Help notes a worldwide rollout as of August 31, 2026) lists AI Overviews, AI Mode, and generative AI features in Discover. Two choices:
- Include (default) — links and content can appear in those features and help ground responses; you can receive impressions and traffic from them.
- Exclude — content is prevented from being visible in those features, including being linked and helping ground responses. You will not receive traffic or impressions from them.
Google says exclude is not a ranking or inclusion signal for the rest of Search, does not override Merchant Center or Ads, does not replace noindex if you want out of Search entirely, and does not affect AI training — training still points at Google-Extended. Content crawled from the site is not eligible as an input to generate an AI response or preview in those Search generative features once exclude is live. Other sites can still publish similar copy. Changes generally take a few days; Help says exclusion within 1–2 days after the control goes live, with some cached content taking longer.
| Goal | First control | Do not reach for |
|---|---|---|
| Keep blue-link snippets; leave Overviews / AI Mode / generative Discover | Search Console Search generative AI → Exclude | Sitewide nosnippet |
| Kill snippets and documented Overview direct input on one URL | nosnippet or max-snippet:0 on that URL | Google-Extended |
| Hide one paragraph from snippets | data-nosnippet on span/div/section | Property-level Exclude |
| Remove the URL from Google Search | noindex (and keep it crawlable) | A plugin labeled “AI” |
| Limit listed Gemini training / grounding | Google-Extended in robots.txt | nosnippet |
| Opt out of ChatGPT or Perplexity | Those vendors’ crawler tokens | Google robots meta |
Child properties can inherit or override a parent. Check which Search Console property you are editing before you congratulate yourself.
What Exclude does not do, in Help’s own limits:
- It is not a ranking signal for the rest of Search.
- It does not override Merchant Center or Google Ads participation.
- It does not stop your content from helping Search as a whole understand queries and pages.
- It does not replace Google-Extended for training of the models used to generate responses in those features.
- It does not stop other sites from publishing copy that looks like yours.
- It does not
noindexyou.
| Property you opened | Risk | Check |
|---|---|---|
Domain property example.com | Applies broadly, including subdomains that inherit | List child properties before Exclude |
URL-prefix https://www.example.com/ | http, apex, and other hosts may not follow | Verify the cited host |
URL-prefix /blog/ | Marketing homepage still included | Path is not “the brand” |
| Staging property | You excluded a host Google barely serves | Production property, then recrawl |
| Inherit from parent | A parent Exclude you did not notice | Settings screen says inherit or not |
Procedure when Exclude is the real request:
- Confirm the verified property is the host you think it is (www vs apex, http vs https).
- Settings → Search generative AI → Exclude.
- Screenshot the setting into the ticket.
- Log a frozen query panel the same day (day 0).
- Re-check at day 3 and day 14. Do not add
nosnippeton day 1 “to be sure.”
When is nosnippet the right tool versus an accident?
Right tool: a URL that should remain indexed as a title-and-URL result, must not donate body copy into snippets or Overview direct input, and you accept that cost on that URL. Think: a gated excerpt you accidentally published in full, a legal notice, a thin thank-you page that still needs to resolve, a PDF you only want as a filename in Search.
Accident: a “prevent AI scraping” plugin, a security review that copied nosnippet onto every template, or a max-snippet:0 default that nobody reopened. Those sites keep ranking and then file a visibility ticket. That is ranked but missing, not this post.
| URL class | Default | nosnippet only if | Prefer instead |
|---|---|---|---|
| Money page, offer, pillar, comparison | Snippet-eligible | You have a written reason to hide the body | Fix extractability |
| Blog you want cited | Snippet-eligible | Never, unless the post is a takedown | Answer-first lead |
| Thank-you / cart / account shell | Often noindex | Indexed but body must stay out of snippets | noindex if it should not rank |
| Press embargo HTML | Temporary noindex | Embargo lifted but body still sensitive | Take the page down |
| PDF spec sheet | X-Robots-Tag chosen per file | Body must not preview | Header on that file only |
| Entire marketing site | Almost never | Legal instructed a Search-preview kill | GSC generative Exclude, or stop publishing |
Decision list:
- Want citations → do not ship
nosnippeton money templates. Full stop. - Want Search, not generative units → GSC Exclude. Measure, then stop.
- Want this one URL silent in snippets →
nosnippetormax-snippet:0on that URL, crawlable, logged. - Want off Google →
noindex, not a snippet toy.
Bravery is not a robots strategy. Write the goal in one sentence before you touch meta.
How do you ship preview controls so Googlebot actually sees them?
Google only follows rules it can fetch. The AI-features troubleshooting list is short: make the preview control correct and visible to Googlebot, then wait for recrawl (several days to several months). Use URL Inspection on the HTML Googlebot received, not DevTools on your laptop.
Page-level HTML:
<meta name="robots" content="nosnippet">Google-only snippet kill (other engines ignored for this token):
<meta name="googlebot" content="nosnippet">HTTP header, including non-HTML:
X-Robots-Tag: nosnippetname and content are case-insensitive. Multiple robots meta tags combine; Google uses the sum of the negative rules. robots plus googlebot is not “the second one wins.” A page with nofollow on robots and noindex on googlebot is treated as noindex, nofollow for Googlebot.
Implementation order:
- Decide URL vs template vs property. Do not start in the CDN WAF.
- Put the rule in HTML
<head>orX-Robots-Tag, not robots.txt. - Confirm the URL is allowed in robots.txt for Googlebot.
curl -sIthe live URL; look forX-Robots-Tag.curl -sthe HTML; look for<meta name="robots"andgooglebot.- URL Inspection → crawled HTML. Search for the same strings.
- Request indexing only after the crawled copy matches. Then wait.
| Place you “set it” | What Googlebot might actually get | Fix |
|---|---|---|
| SEO plugin UI | Template still emits max-snippet:0 | View crawled HTML |
| Tag manager | Race / not in first HTML | Put meta in the document head |
CDN header on /blog/* | Header on HTML and on images | Scope the header |
| Staging only | Production missing the tag | Compare hosts |
| robots.txt Disallow | Meta never read | Allow crawl; keep noindex if needed |
SPA client-only <head> | First response has no meta | SSR or prerender the tag |
PDFs and images cannot carry a robots meta tag in HTML. Use X-Robots-Tag on those responses. Do not invent a <meta> inside a PDF.
Google’s spec shows the header pattern for a class of files. Adapt the rule; do not copy noindex if you only meant nosnippet.
Apache (site .htaccess or vhost):
<Files ~ "\.pdf$">
Header set X-Robots-Tag "nosnippet"
</Files>NGINX:
location ~* \.pdf$ {
add_header X-Robots-Tag "nosnippet";
}A second header can add unavailable_after or noimageindex. Multiple X-Robots-Tag lines combine. A header that says googlebot: nosnippet addresses Googlebot; a header without a user agent applies to all crawlers that honor it.
| File type | HTML robots meta possible? | Typical header |
|---|---|---|
| HTML page | Yes | Meta in <head> or header, not both fighting |
| PDF spec / white paper | No | X-Robots-Tag: nosnippet (or noindex if it should not rank) |
| PNG / JPEG / GIF | No | noimageindex if images must stay out of Image search |
| One sensitive PDF | No | location = /secrets/file.pdf (nginx) or <Files "file.pdf"> |
If HTML and the header both speak, Google still sums negatives. A “safe” template meta plus a CDN nosnippet is a kill. Log both layers in the ticket.
What still shows after you set nosnippet?
Expect a Search result that can still show a title and URL. Expect that a thumbnail may remain. Expect that schema.org you left in the page can still feed rich results, with the spec’s noted exception around some description fields and max-snippet. Expect other people’s pages to remain citable.
Do not expect ChatGPT, Perplexity, or a Gemini chat surface that is not Google Search to honor this tag as if it were their robots.txt. Different products, different tokens.
| Surface | nosnippet on your Google-crawled URL | Honest expectation |
|---|---|---|
| Google web result snippet | Text snippet / video preview off | Title + URL; thumbnail maybe |
| Google Images / Discover (listed with the rule) | Spec says the rule applies | Do not assume every Discover layout is identical |
| AI Overviews / AI Mode | No direct input from that page | Not a brand-mention warranty |
| Google Search generative features via GSC Exclude | Separate property control | Use it if snippets must stay |
| Gemini Apps / Vertex listed grounding | Google-Extended, not this tag | Different ticket |
| ChatGPT search / Perplexity | Not this spec | Vendor crawlers and retrieval |
| Your competitor’s rewrite of your post | Unaffected | Their URL, their rules |
If you still see your copy inside an Overview after a correct, crawlable nosnippet, follow Google’s troubleshooting: inspect crawled HTML, wait for recrawl, then post in the Search Central Help Community. Do not add noindex in a panic on day two unless you actually want off Search.
Checklist for “it is still in the Overview”:
- Crawled HTML contains
nosnippetormax-snippet:0 - No robots.txt Disallow on that URL
- Canonical is the URL you edited
- You waited through at least one recrawl window (days, not hours)
- The Overview citation is your host, not a scraper
- Structured data is not republishing the same claims
How do you measure whether the opt-out is working?
Treat this as a control change, not a content launch. Freeze the query list before you touch meta.
Search Console still rolls AI Overviews and AI Mode into the Performance report under the Web search type. Exclude via the generative AI control should show up as lost impressions and clicks from those features; Google says you will not receive traffic or impressions from them after exclude. nosnippet will not give you a dedicated “Overview opt-out” scoreboard. You measure the SERP and the Overview unit yourself.
Day-0 panel (write it down):
- Ten queries that currently fire an Overview and cite you, if you have them.
- Ten queries that fire an Overview and cite someone else.
- Five queries with no Overview (negative control).
- For each: screenshot, cited hosts, whether your snippet text is visible, date.
- URL Inspection dump for the money URLs (crawled HTML hash or a saved copy).
Re-run on day 7, 14, and 30. Log four fields: Overview present (Y/N), your URL cited (Y/N), your text quoted (Y/N), classic snippet present (Y/N). Citation without quotation is a different failure than quotation with a snippet.
| Signal | Suggests the control landed | Suggests you measured the wrong thing |
|---|---|---|
| Classic snippet gone; title remains | nosnippet / max-snippet:0 seen | You noindexed by mistake |
| Your URL drops out of Overview citations after recrawl | Compatible with documented direct-input block | Feature simply did not fire that day |
| Snippet gone, Overview still quotes a paragraph | Wrong URL, scraper, or schema | “Google ignored us” on day one |
| Nothing changed in 24 hours | Recrawl not done | Control is “broken” |
| ChatGPT still names you | Expected | You thought Google meta governs OpenAI |
| GSC generative impressions fall after Exclude | Compatible with Help | You also shipped nosnippet and confused the two |
Do not use one mobile SERP from headquarters Wi-Fi as the KPI. Location, login, and whether the Overview fires are noisy. The panel exists to make the noise visible.
Sample log (copy this; do not invent extra columns in week one):
| Date | Query | Overview? | Your host cited? | Your sentences quoted? | Classic snippet? | Crawled HTML has nosnippet? | Notes |
|---|---|---|---|---|---|---|---|
| Day 0 | [money query] | Y | Y | Y | Y | N | Baseline |
| Day 7 | same | Y | N | N | N | Y | Compatible with recrawl |
| Day 7 | same | Y | Y | Y | N | Y | Wrong URL, scraper, or schema |
| Day 14 | no-Overview control | N | — | — | N | Y | Feature never fired — not proof |
Pass/fail for a page-level nosnippet ticket:
- Crawled HTML shows the rule on the canonical you edited.
- Classic text snippet is gone on queries that still show the blue link.
- Your host stops appearing as an Overview supporting link or stops supplying quoted spans — after recrawl, on queries that still fire the unit.
- Negative-control queries without Overviews stay without Overviews. You did not “fix” a feature that was never there.
Fail: snippet still shows your first paragraph. The tag is not in crawled HTML, or a CDN is serving a different cache to Googlebot.
Fail: snippet is gone, Overview still quotes you word-for-word from your URL, and Inspection is clean. Then you are in Google’s troubleshooting branch (wait, then Help Community) — not in a “add noindex” branch.
What usually fails first when teams try this?
They pick the wrong switch, apply it sitewide, and then cannot tell whether Search or Overviews moved.
Failure mode I keep seeing on audits: a plugin labeled something like “discourage AI” writes nosnippet or max-snippet:0 on every public template. Organic titles still rank. Classic CTR falls because snippets disappeared. Overviews stop using the site as direct input. The channel report says “AI killed us.” The HTML says you opted out of the preview. Cost is the clicks you used to get from the blue link, plus every citation you just refused. Fix: remove the sitewide rule, keep nosnippet only on URLs with a written reason, request recrawl, restore snippets, then decide Overviews separately in Search Console.
| Failure | How it shows up | Repair |
|---|---|---|
Sitewide plugin nosnippet | Rank holds; snippets empty; Overviews go quiet | Remove global rule; recrawl money URLs |
max-snippet:0 leftover | Same as nosnippet, hidden in “advanced” | Set omit or -1 unless you mean it |
data-nosnippet on <main> | Page looks indexed; nothing quotable | Unwrap the content tree |
| robots.txt Disallow + meta | Rule never read | Allow crawl |
| Google-Extended treated as Overview opt-out | Search unchanged; Gemini policy changed | Re-read the token page |
nositelinkssearchbox as a plan | No effect | Stop |
GSC Exclude plus sitewide nosnippet | You cannot attribute the drop | Reverse one change at a time |
| Client-only meta tag | Googlebot never saw it | Put it in first HTML |
Day-two noindex panic | You left Search | Remove noindex; wait |
The expensive part is not the tag. It is a quarter of missing snippet CTR plus a Slack thread that rewrites the blog instead of viewing crawled HTML. I have shipped hundreds of production sites. The ones that “mysteriously” left Overviews almost always had a preview control someone considered harmless.
What should you skip if you only have a week?
Skip a content rewrite. Skip a new schema package. Skip blocking Googlebot. Skip Google-Extended unless training policy is the actual ticket.
Do this instead:
- Write one sentence: Search snippets, generative Search features, Gemini training, or off Google entirely.
- View crawled HTML for the top 10 money URLs. Search for
nosnippet,max-snippet,data-nosnippet,noindex. - If you want citations and you find those tags, delete them from the templates. That is the week.
- If you want out of Overviews and you want to keep snippets, use Search Console Exclude. Do not add
nosnippet. - If you want one URL silent, add
nosnippeton that URL only, keep it crawlable, inspect, request indexing. - Freeze a 15-query panel and screenshot day 0. Re-check once at the end of the week even if recrawl is incomplete.
- Do not change ChatGPT or Perplexity crawlers as a side effect of a Google ticket.
When this is not worth doing yet: you have no indexed public pages, or nobody in the company can say whether they want citations. Then the week is a policy meeting, not a meta tag. An unsigned “block AI” ticket will still ship. It will just ship the blunt tool.
If you only remember one row: Google documented nosnippet as snippet suppression plus direct input prevention for Overviews and AI Mode. That is a real, costly control. It is not a guaranteed brand-level Overview erasure, and it is a terrible default on pages you still want quoted.
Pages to inspect first if the ticket is “are we even snippet-eligible?” — the same money set the playbook already uses, not a random blog:
- Home, About, and the offer URL that ranks for the category query.
- The comparison or pricing URL Overviews already quote (or that you wish they would).
- One PDF or press HTML the CDN wraps with a global header.
- The template that ships every blog post (global SEO plugin defaults).
- The canonical www vs apex pair, in case only one host has the meta.
| If Inspection shows | This week’s job | Not this week’s job |
|---|---|---|
nosnippet on the offer URL, you want cites | Remove it; request recrawl | New cluster |
| Clean HTML, you want out of generative Search | GSC Exclude | Sitewide nosnippet |
max-snippet:0 in the blog layout | Fix the layout default | Per-post hero copy |
data-nosnippet wrapping <main> | Unwrap | Schema sprint |
Disallow in robots.txt on /resources/ | Allow crawl; keep headers honest | Another plugin |
FAQ
Can nosnippet keep me out of AI Overviews?
Google documents that nosnippet applies to AI Overviews and AI Mode and prevents that page’s content from being used as a direct input, while also blocking the Search text snippet. That is an opt-out of snippet-style use of that URL, not a warranty that your brand never appears. Other sites, licensed uses, and structured data Google treats as separately permitted can still show up.
How do I measure whether can nosnippet keep me out of AI Overviews is working?
Save crawled HTML that shows the tag, then re-run a frozen query panel and log Overview present, your URL cited, your text quoted, and whether the classic snippet still renders. Recrawl can take days to months, so a 24-hour screenshot proves almost nothing. If ChatGPT or Perplexity still names you, that is a different product — Google robots meta did not claim those surfaces.
What usually fails first when teams try this?
They apply nosnippet or max-snippet:0 sitewide from a plugin, or they Disallow Googlebot so the meta tag is never read. The next failure is using Google-Extended or nositelinkssearchbox as if they were Overview switches. Repair is: crawled HTML, one control per goal, recrawl, then measure.
How long does this take to show results?
Google’s AI-features page says recrawl after a preview-control change can take several days to several months. Search Console’s generative AI Exclude is documented on a shorter clock — generally a few days, with Help citing 1–2 days after the control is live and longer for some cached content. Do not declare failure the next morning, and do not stack a second kill switch while you wait.
What should I skip if I only have a week?
Skip rewriting the cluster and skip blocking Googlebot. Inspect crawled HTML for preview tags on money URLs, pick one control that matches the written goal, and freeze a small query panel. If you want citations, the week is removing accidental nosnippet. If you want out of generative Search with snippets intact, the week is Search Console Exclude.
When is this not worth doing yet?
When you cannot say whether you want to be quoted. nosnippet is a real Search cost. If the pages are not indexed, or the only sensitive copy is already behind login, you do not have an Overview problem — you have a publishing problem. Hold the tag until the goal is a sentence a lawyer and an SEO would both sign.
CTA
Pick the control Google documented for the cost you will actually pay — snippets, generative Search, training, or off the index — and stop treating nosnippet as a magic Overview toggle.
Lane overview: /visibility. Next step: a visibility audit.
What questions does this article answer?
- Can `nosnippet` keep me out of AI Overviews?
- Google documents that `nosnippet` applies to AI Overviews and AI Mode and prevents that page’s content from being used as a **direct input**, while also blocking the Search text snippet. That is an opt-out of snippet-style use of **that URL**, not a warranty that your brand never appears. Other sites, licensed uses, and structured data Google treats as separately permitted can still show up.
- How do I measure whether can `nosnippet` keep me out of AI Overviews is working?
- Save crawled HTML that shows the tag, then re-run a frozen query panel and log Overview present, your URL cited, your text quoted, and whether the classic snippet still renders. Recrawl can take days to months, so a 24-hour screenshot proves almost nothing. If ChatGPT or Perplexity still names you, that is a different product — Google robots meta did not claim those surfaces.
- What usually fails first when teams try this?
- They apply `nosnippet` or `max-snippet:0` sitewide from a plugin, or they `Disallow` Googlebot so the meta tag is never read. The next failure is using Google-Extended or `nositelinkssearchbox` as if they were Overview switches. Repair is: crawled HTML, one control per goal, recrawl, then measure.
- How long does this take to show results?
- Google’s AI-features page says recrawl after a preview-control change can take several days to several months. Search Console’s generative AI Exclude is documented on a shorter clock — generally a few days, with Help citing 1–2 days after the control is live and longer for some cached content. Do not declare failure the next morning, and do not stack a second kill switch while you wait.
- What should I skip if I only have a week?
- Skip rewriting the cluster and skip blocking Googlebot. Inspect crawled HTML for preview tags on money URLs, pick one control that matches the written goal, and freeze a small query panel. If you want citations, the week is removing accidental `nosnippet`. If you want out of generative Search with snippets intact, the week is Search Console Exclude.
- When is this not worth doing yet?
- When you cannot say whether you want to be quoted. `nosnippet` is a real Search cost. If the pages are not indexed, or the only sensitive copy is already behind login, you do not have an Overview problem — you have a publishing problem. Hold the tag until the goal is a sentence a lawyer and an SEO would both sign.
Last reviewed — Google robots meta spec, AI features page, Search Console Search generative AI control Help, URL Inspection Help, and Google-Extended common-crawler docs checked 2026-09-05.
Keep reading.
AI Visibility Do I need FAQ schema to show up in AI Overviews
No. FAQ schema is not a ticket into AI Overviews. Google's floor is indexed and snippet-eligible; visible Q&A helps extraction. FAQ JSON-LD does not buy a cite.
AI Visibility Should I optimize for zero-click search or fight it
Optimize for citation and brand capture on informational queries. Fight only when a protectable commercial URL is being stolen, not every zero-click result.
AI Visibility How do I get products recommended in AI answers and AI Overviews
To get products recommended in AI answers, ship unique specs in HTML, matching Product schema, consistent reviews, and a Merchant Center feed that agrees.
AI Visibility
AI Visibility Cannabis visibility when the ad accounts are banned
Google and Meta will not take the usual spend. The models still answer dispensary, cultivator, and brand questions — if the site can be read and the cart can clear a 21+ order.
AI Visibility When ChatGPT names the franchise, not your shop
Run the best-HVAC-near-me prompt panel. If the model names a national franchise, fix corroboration and entity facts — not another blog calendar.
AI Visibility How do I get cited by Perplexity specifically
Allow PerplexityBot, put a liftable answer and unique numbers in HTML, then log numbered sources on a frozen prompt panel. There is no bought citation rate.
AI Visibility What belongs in an AI visibility monthly retainer vs a one-time audit
A one-time audit is the baseline plus prioritized fixes. A monthly retainer is prompt-panel tracking, entity hygiene, page jobs, and citation recovery.