Does noindex keep a page out of ChatGPT and Perplexity?

A noindex tag is a documented control on Google and Bing. It is not the documented control on ChatGPT or Perplexity.

Google states that a page must be "indexed and eligible to be shown in Google Search with a snippet" to appear as a supporting link in AI Overviews or AI Mode. So noindex removes eligibility, and nosnippet or max-snippet:0 removes the snippet. Bing documents the same idea at section level: content marked data-nosnippet is "still indexed normally, but it will be excluded from snippets and AI summaries." The differences between noindex, nosnippet and data-nosnippet matter when you only want part of a page hidden.

OpenAI and Perplexity describe their controls differently. Their crawler pages, as of October 2026, list robots.txt user agents and what each one does. Neither documents noindex as a way to stay out of answers. That doesn't prove they ignore the tag. It means you'd be relying on undocumented behavior.

The practical rule: use noindex for Google and Bing, and use per-bot robots.txt rules for ChatGPT and Perplexity.

Two details change how you write rules. First, there is no robots.txt token that blocks Google AI Overviews or AI Mode specifically. Googlebot governs all Google Search surfaces, so the on-page tags are your tool there. Second, OpenAI's bots are independent. GPTBot covers training, OAI-SearchBot covers search. OpenAI says "Each setting is independent of the others." If your goal is the reverse (stay in search, skip training), see how to block AI training but stay in AI search.

Say you run a SaaS company with an internal pricing draft at /pricing-draft/ that must not appear in AI answers. You'd keep a noindex tag on the page for Google and Bing, then add per-bot rules for the others:

User-agent: OAI-SearchBot
Disallow: /pricing-draft/

User-agent: PerplexityBot
Disallow: /pricing-draft/

User-agent: Claude-SearchBot
Disallow: /pricing-draft/

Don't add a Disallow for Googlebot on the same path. Googlebot has to fetch the page to read the noindex tag. Blocking it in robots.txt can leave the tag unseen.

For PDFs and other files that can't carry a meta tag, Google reads the same directives from an HTTP header. Our page on whether an X-Robots-Tag header keeps pages out of AI search covers that case.

To confirm the rules parse the way you expect for each bot, the robots.txt Checker for AI Crawlers resolves allow and disallow per user agent.

User-triggered fetchers

When a person pastes a URL or asks an assistant to open a page, a different agent may fetch it. OpenAI says robots.txt rules "may not apply" to ChatGPT-User. Perplexity says Perplexity-User "generally ignores robots.txt rules." The split between index-time crawlers and answer-time fetchers is why a blocked page can still show up in a single conversation.

Navigational links

OpenAI notes opted-out sites "can still appear as navigational links." Blocking OAI-SearchBot removes you from answers, not necessarily from every link. More on that in what happens to links when you block OAI-SearchBot.

What a model already knows

Robots.txt controls future crawling. It does not reach a model's parametric knowledge, which is frozen at a training cutoff. If a fact was public before that cutoff, a retrieval-off answer may still contain it.

Truly sensitive content

If a page must never be read by anyone outside your team, robots.txt and noindex are the wrong tools. Put it behind authentication.

1. List the pages or sections that must stay out of AI answers, and decide whether you need the whole page hidden or only part of it.
2. Add noindex (whole page) or nosnippet and data-nosnippet (parts) for Google and Bing. Keep Googlebot allowed.
3. Add per-path Disallow rules for OAI-SearchBot and PerplexityBot, plus Claude-SearchBot if Claude matters to you. Leave GPTBot rules as a separate decision about training.
4. Check the inverse: make sure a blanket AI block isn't hiding pages you do want in ChatGPT search. That's a common real-world error.
5. Move anything confidential behind a login, since user-triggered fetchers may not honor robots.txt.

For a fuller walkthrough by content type, follow the playbook for keeping specific content out of AI answers. Perplexity's agents are summarized in Perplexity's crawler documentation, digested.

Does blocking OAI-SearchBot also stop OpenAI from training on my page? No. OpenAI documents GPTBot for training and OAI-SearchBot for ChatGPT search, and says each setting is independent. Disallowing OAI-SearchBot removes the page from ChatGPT search answers but says nothing about training. If you want both, add a separate rule for GPTBot. If you only want to skip training, disallow GPTBot and leave OAI-SearchBot allowed.

Should I use noindex and robots.txt together on the same page? Yes, but aim them at different engines. Keep noindex on the page for Google and Bing, and keep Googlebot and Bingbot allowed so they can read it. Then add robots.txt Disallow rules only for the search bots that document robots.txt as their control, such as OAI-SearchBot and PerplexityBot. Blocking Googlebot in robots.txt can stop Google from seeing the noindex tag.

Does nosnippet or data-nosnippet work in ChatGPT or Perplexity? Neither OpenAI nor Perplexity documents these attributes as controls, as of October 2026. Google documents nosnippet, max-snippet and data-nosnippet for its AI features, and Bing documents data-nosnippet for its AI summaries. For ChatGPT and Perplexity, the documented route is a robots.txt rule for their search crawlers, which works per path rather than per section.

Can a firewall rule replace robots.txt for blocking AI search bots? It can block them, since a CDN or firewall can turn away a crawler that robots.txt allows. OpenAI's own eligibility guidance requires both robots.txt permission and network access from its published IP ranges. Robots.txt is still the clearer, documented signal of intent. Use firewall rules deliberately, and check they aren't also blocking bots you want.