OpenAI’s 2026 ChatGPT Search Guidance Explained

ChatGPT now answers more than 1 billion people a week, according to OpenAI’s own July 2026 disclosure — up from 900 million just five months earlier. A huge share of those conversations involve live web search, which makes OpenAI’s crawler and indexing guidance one of the more consequential technical documents a website owner can read this year, even though almost nobody has read it closely. This is the current ChatGPT search guidance broken down: what OpenAI’s documentation actually says today, what quietly changed in 2026, and what it means for whether your site shows up.

The short version: any public website can appear in ChatGPT search, as long as it isn’t blocking OAI-SearchBot — but “not blocking” and “fully opted out” turn out to be two different settings, and OpenAI added a fourth crawler this year that most site owners haven’t heard of. Here’s the complete, current picture.

What Changed in OpenAI’s Crawler Guidance in 2026

OpenAI’s crawler documentation has been updated more than once this year, and a few of the changes matter more than they’ve gotten credit for.

A Fourth Crawler Joined the Lineup: OAI-AdsBot

In April 2026, OpenAI’s crawler docs added a fourth bot: OAI-AdsBot. It’s separate from the other three in both purpose and scope.

• It only visits pages that have been submitted as ChatGPT ads — not general web content.
• It checks the landing page against OpenAI’s ad policies and may factor its content into ad-targeting decisions.
• Per OpenAI, data it collects is not used for training and doesn’t affect organic answers or citations.

One practical gap worth flagging: unlike the other three crawlers, OpenAI hasn’t yet published an IP address range file for OAI-AdsBot, which makes independently verifying a genuine OAI-AdsBot visit harder than it is for the other three. If your site doesn’t run ChatGPT ads, this bot has no reason to visit you at all.

The Noindex Catch-22 in OpenAI’s Publishers FAQ

This is the least-known part of OpenAI’s current guidance, and it directly contradicts a common assumption.

Most site owners assume that disallowing OAI-SearchBot in robots.txt fully removes a page from ChatGPT. OpenAI’s Publishers FAQ says otherwise: if OpenAI obtains a disallowed page’s URL from a third-party source, or by crawling other pages that link to it, and has signals the page is relevant to a query, it may still surface the bare link and page title inside ChatGPT Atlas — even though the full page was never crawled.

To fully opt out of that fallback behavior, OpenAI’s guidance says a noindex meta tag is required, not just a robots.txt disallow. And there’s a catch built into that fix: for OAI-SearchBot to read a noindex tag on a page at all, it has to be allowed to crawl that page in the first place. A page that’s fully disallowed can still have its bare URL and title surface elsewhere in the product — the noindex tag is the only setting that closes that gap completely.

Automatic UTM Tracking for ChatGPT Referrals

One quieter but genuinely useful change: OpenAI’s current guidance confirms that ChatGPT automatically appends utm_source=chatgpt.com to outbound referral links. Any site allowing OAI-SearchBot can already isolate ChatGPT-driven traffic in Google Analytics or similar tools without any extra tagging work — a detail that’s easy to miss if nobody’s checked the FAQ recently.

The Four OpenAI Crawlers, Compared

Conflating these four is still the most common configuration mistake site owners make. Each is controlled independently in robots.txt.

A site that wants search visibility but no training use should disallow GPTBot while explicitly allowing OAI-SearchBot — OpenAI’s documentation confirms these are two separate decisions, not one combined setting. For a deeper look at how these crawlers fit into the broader retrieval pipeline — including what happens after a page is crawled — see How Does ChatGPT Find and Cite Websites?

How to Configure robots.txt for OAI-SearchBot

Allow Search, Block Training

A minimal, current configuration that opts out of training while preserving search eligibility looks like this:

User-agent: GPTBot

Disallow: /

User-agent: OAI-SearchBot

Allow: /

User-agent: ChatGPT-User

Allow: /

OpenAI’s guidance notes that changes affecting search results can take roughly 24 hours to take effect after a robots.txt update — so don’t judge a fix by checking ChatGPT again five minutes later.

Verifying Real OpenAI Traffic by IP

User-agent strings can be spoofed by anyone. OpenAI publishes IP address ranges for its three older crawlers as JSON files:

• openai.com/gptbot.json
• openai.com/searchbot.json
• openai.com/chatgpt-user.json

Matching incoming requests against these published ranges — not just the user-agent string — is the only reliable way to confirm a request genuinely came from OpenAI rather than a scraper impersonating it.

Does Being Indexed Guarantee Website Visibility in ChatGPT?

No — and this is worth stating plainly, because OpenAI’s own guidance doesn’t promise it either. Allowing OAI-SearchBot removes a technical barrier to appearing in ChatGPT search. It doesn’t guarantee a citation for any specific query, and OpenAI has not published a ranking formula describing how it selects among eligible, indexed pages.

What indexing does guarantee: eligibility to be considered. What it doesn’t guarantee: that you’ll actually be selected, cited, or ranked above another eligible source for a given question.

If your site still isn’t appearing despite a correct robots.txt setup, the issue is usually further down the pipeline — see Why Is My Website Not Showing in ChatGPT? 10 Things to Check for the full diagnostic walkthrough.

What This Means for Website Visibility Right Now

E-Commerce: The Product Feed, Not llms.txt

For product-based businesses specifically, OpenAI’s current guidance points toward a structured product feed submission as the closest thing to an official, direct listing channel — distinct from general content indexing. Worth noting clearly: OpenAI has made no public commitment to supporting llms.txt, and its current documentation doesn’t endorse any special schema requirement beyond standard structured data practices. For product visibility specifically, the feed spec is the one OpenAI actually documents.

A May 2026 Referral Traffic Shift

Independent measurement from several SEO trackers reported a sharp jump in ChatGPT referral traffic to publisher sites around May 2026, coinciding with a change in how prominently ChatGPT displays outbound source links in its answers. Industry observation: OpenAI did not formally announce this change — it was identified through third-party traffic measurement rather than official release notes, which is a pattern worth expecting more of, not less, as OpenAI continues shipping changes to search behavior without always documenting them in the Publishers FAQ.

Frequently Asked Questions

Does OpenAI have an official llms.txt policy?

No. OpenAI’s current documentation makes no commitment to supporting or requiring an llms.txt file. For structured, official visibility channels, its guidance points toward the product feed spec for e-commerce rather than llms.txt.

If I block OAI-SearchBot, is my site completely invisible in ChatGPT?

Not necessarily. OpenAI’s Publishers FAQ states that a disallowed page’s bare URL and title may still surface in ChatGPT Atlas if OpenAI obtains that URL from another source and judges it relevant to a query. A noindex meta tag — not just a robots.txt disallow — is required to fully prevent this.

What’s the difference between GPTBot and OAI-SearchBot?

GPTBot collects content for potential model training. OAI-SearchBot builds the index behind ChatGPT’s search results and citations. OpenAI’s documentation confirms they’re controlled independently in robots.txt — blocking one doesn’t affect the other.

What is OAI-AdsBot?

A crawler OpenAI added to its documentation in April 2026 that visits landing pages submitted for ChatGPT ads, checking policy compliance. Per OpenAI, it does not affect organic search results or citations, and it’s unrelated to general content crawling.

How long does a robots.txt change take to affect ChatGPT search visibility?

OpenAI’s guidance states that changes can take approximately 24 hours to be reflected in ChatGPT’s search systems after a robots.txt update.

Conclusion

OpenAI’s 2026 guidance clarifies more than it changes: OAI-SearchBot still governs search eligibility, GPTBot still governs training, and the two remain independent settings. What’s new is the noindex catch-22, the fourth ad-focused crawler, and confirmation that llms.txt isn’t part of OpenAI’s official plans. Start by auditing your robots.txt against the table above — LLMrush’s GPTBot Checker and AI Crawlability Checker confirm your current configuration in seconds.