What is OAI-AdsBot?

What is OAI-AdsBot?

It matters commercially because OpenAI requires that your page let it in. A site that turns OAI-AdsBot away is a site whose advertising can stop working while every page still loads perfectly for every human who visits.

Everything below is from OpenAI’s own crawler documentation, advertiser guidance and Ads Manager help articles, read on 4 October 2026.

What does OAI-AdsBot actually do?

That second job is the one people underestimate. The page is not only where the click lands, it is one of the inputs to whether the click ever happens.

OpenAI names some of what feeds the decision, it says “multiple signals, including”, and the list it gives runs: “the context and intent of the current conversation, the ad’s landing page, title, copy, advertiser-provided context hints…” Your page is in it, alongside your ad copy.

So a thin or unreadable page costs you twice. Once because fewer people who arrive turn into customers, and again because it is part of what decides how often anyone arrives at all.

OpenAI also states that data OAI-AdsBot collects “is not used to train generative AI foundation models”. It is an ad-review reader, not a training crawler.

How is OAI-AdsBot different from OpenAI’s other bots?

Three of the four are controlled separately in robots.txt, the file on your site that tells automated readers where they may go. OpenAI says those settings are independent: you can allow one and block another. ChatGPT-User is the exception, and the table below records why.

The distinction that costs businesses money is between OAI-SearchBot and GPTBot, and they are easy to confuse. A blanket block on “AI bots”, added to keep your writing out of training data, will also switch off the one that puts you in ChatGPT’s answers. The training problem gets solved and the visibility goes with it, and nothing tells you.

OpenAI publishes the full user-agent string for the ads reader, which is what will appear in your server logs:

Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; OAI-AdsBot/1.0; +https://openai.com/adsbot

Does OAI-AdsBot obey robots.txt?

OpenAI’s own example of what that looks like:

User-agent: OAI-SearchBot
Allow: /

User-agent: OAI-AdsBot
Allow: /

Its instruction is blunt about which one is not optional: “You must allow OAI-AdsBot. We recommend allowing both OAI-AdsBot and OAI-SearchBot.”

Further down the same page, OpenAI explains why the second one is only recommended: “OAI-AdsBot is required for ChatGPT Ads landing page validation and review. OAI-SearchBot is recommended because it may help OpenAI understand public web content, but OAI-AdsBot is the crawler advertisers should prioritize for ads readiness.”

There is one exception, for shops whose ads come from a product feed: “For product feeds, also allow OAI-SearchBot to crawl the product image URLs.” OpenAI adds that “a 403 response can block feed processing”. In plain terms, if the security service in front of your product images turns OAI-SearchBot away, OpenAI may not be able to process your product list.

If you would rather not read your own robots.txt, our free crawler check parses it for you and tries each agent for real. No email.

Why does OAI-AdsBot get blocked by accident?

This is rarely a decision anybody made. It is a default, and it is invisible from the outside, because a security service can turn away an automated reader while serving your homepage perfectly to every person who visits.

OpenAI describes three layers to check, in order: robots.txt first, then “Web protection and bot mitigation”, then “Human verification and anti-bot logic”. In plain terms, three things can turn a reader away, the small text file that lists who may visit, the security service sitting in front of your site, and any check that asks a visitor to prove they are human.

There is a detail worth knowing if you use Cloudflare, and it is not the reassurance it sounds like. OpenAI states that “OAI-AdsBot is officially verified and allowlisted by Cloudflare”, and publishes the mistaken-blocking guidance on the very same page. Verified does not mean it always gets through, because rules you or your host added sit on top of the defaults.

Rate limiting is the other quiet cause. OpenAI suggests checking for “429 Too Many Requests” responses around the time the crawler tried, particularly after a large batch of ads went up at once.

Can OpenAI let my ad through manually?

There is no support ticket that solves this, which is worth knowing before you spend a week waiting on one.

OpenAI adds that fixing the block may not be the end of it: “Ads may need to be re-uploaded or resubmitted for review after the landing page becomes accessible.” Once the page is fixed, the ad’s status may update by itself. If it does not, OpenAI’s advice is to re-upload or resubmit it.

Its own first-line checklist is a good one to hand an IT person: does the page return a successful response to OAI-AdsBot, does robots.txt allow that path, and do WAF, CDN, bot mitigation, JavaScript challenges, CAPTCHAs, authentication or geo rules block automated access.

One more piece of advice from the same guidance: send people to an ordinary web page wherever you can. OpenAI warns that “App-store links, deep links, documents, or destinations that require an app, login, region-specific access, or unsupported redirects may not provide enough crawlable content for validation or review.”

How do you know it is happening to you?

That is the whole trap in one sentence. The campaign is set up, the site loads, no error appears on the site itself, and the ad does not run, for a reason that shows in the ad’s status instead.

ChatGPT Ads currently charges you only after an ad has actually been shown or clicked. So a blocked page does not run up a bill, it runs up nothing at all. Your budget sits there unspent while the conversations you were bidding for go to somebody else, which is the more expensive mistake and the harder one to notice.

We check this because it is easy to miss, and because a tool pointed at Googlebot will not find it. Our free crawler check names all seventeen AI agents and asks your site for a page under sixteen of them. The seventeenth, Google-Extended, is only a setting in a robots.txt file, nothing ever visits under that name, so we read the file for it rather than pretend to knock. It needs no email, and it identifies itself honestly when it calls, see our own bot.