Agent Experiences

Ads Arrive in ChatGPT: What OAI-AdsBot Means for Your robots.txt

By Alan WeibelPublished 8 min read

The change

OpenAI began testing ads in ChatGPT on January 16, 2026, for US users on the Free plan and the Go plan (OpenAI's lower-priced paid plan), rolled them out widely in the US on February 9, 2026, and on August 11, 2026, expanded to nine markets, including the UK, Mexico, Brazil, Japan, and South Korea, with self-serve buying (advertisers can set up and buy ads on their own). A new crawler (an automated program that visits web pages) came with the ad product. It's called OAI-AdsBot, and it visits advertisers' landing pages (the page an ad sends people to) to check them.

Source: OpenAI, August 2026

Most coverage was about the ads, but the new bot is the part that can cost you.

OpenAI now sends four different crawlers (bots that visit and read your pages) to your site. Each one announces itself with its own name (its "user-agent") and does a different job, so blocking each one costs you something different. Lots of sites still have old rules in their robots.txt (the file that tells bots which pages they may visit) that block every OpenAI bot by name, because "block OpenAI" was a decision people made once, in 2023, and never looked at again. If yours does, it's probably doing something you wouldn't choose on purpose today.

What OpenAI launched

US users on ChatGPT's Free and Go plans see ads at the bottom of some answers. OpenAI announced them as a test on January 16, 2026, and rolled them out widely on February 9. It also published a set of "answer independence" principles: the answer is written without regard to who advertises, and the ad is a separate box placed after the answer, not a change to the answer itself.

The August 11, 2026, expansion took ads to nine markets and added self-serve buying, which moved ChatGPT ads from a test toward a full ad business. For a site owner, though, the change that needs attention is in which bots visit your site and why.

Four OpenAI bots, four separate decisions

Bot name (user-agent)What it doesWhat blocking it costs you
GPTBotCollects pages that OpenAI uses to train its AI modelsNothing right away. Your content won't be used to train future models. If you only want to stop training, block this one.
OAI-SearchBotBuilds the index (ChatGPT's catalog of web pages) that ChatGPT search looks things up in and citesYour citations. Block it and ChatGPT search stops citing you, though OpenAI says your site can still show up as a plain link.
ChatGPT-UserFetches a specific page because someone asked ChatGPT about it just nowThe live visit for a person who has already asked about you. Somebody is waiting on this one. OpenAI says robots.txt may not apply to this bot, so a block may not hold.
OAI-AdsBotChecks advertisers' landing pagesThe check on pages you're paying to send people to, if you buy ChatGPT ads. If you don't advertise, it makes no difference.

Each bot needs its own decision, based on what it does and what your business wants. Many sites use one rule for all of them anyway, because the bot names look alike.

The mistake that costs the most

The usual setup looks like this, and it's normally left over from years ago:

User-agent: GPTBot
Disallow: /

User-agent: OAI-SearchBot
Disallow: /

User-agent: ChatGPT-User
Disallow: /

Sometimes it's worse: a rule that blocks anything with "OAI" or "GPT" in its name, set up in the WAF (web application firewall, the security layer that screens traffic before it reaches your site), where nobody will think to look.

The intention behind this is usually "don't train on my content." It also stops ChatGPT search from citing the site, which is a separate decision that got bundled in by accident.

Suppose what you want is: don't train on our content, but do cite us. Then blocking OAI-SearchBot works directly against you. That bot builds the index, and the index is where citations come from. If it can't see your site, ChatGPT search won't cite you, however good your content is. OpenAI says the site can still show up as a plain link, but not as a source in the answer. You'll notice it as "our AI visibility is bad" and probably never trace it back to one line in robots.txt written years earlier.

Advertisers have a similar choice to make. OpenAI's bot documentation says OAI-AdsBot visits landing pages to check that they comply with its ad policies, and may use their content to decide when an ad is most relevant to show. OpenAI's documentation doesn't say whether OAI-AdsBot follows robots.txt, or what happens to an ad whose landing page it can't reach. If you buy ChatGPT ads, the safe choice is to let it in.

A worked example robots.txt

Here's a setup for a site that wants four things: don't train on us, do cite us, answer people who ask about us, and check the pages we pay to promote. The comments explain each decision so the next person who opens the file knows why each line is there.

# ---------------------------------------------------------------
# OpenAI agents. Four distinct crawlers, four distinct decisions.
# Reviewed: 2026-08-17
# ---------------------------------------------------------------

# 1. Training corpus collection.
# DECISION: block. We do not want our content used to train models,
# and blocking this has no effect on whether we can be cited.
User-agent: GPTBot
Disallow: /

# 2. The search index ChatGPT retrieves and cites from.
# DECISION: allow. This is the one that produces citations.
# Blocking it keeps us out of ChatGPT search answers.
# We still exclude internal and non-public paths.
User-agent: OAI-SearchBot
Allow: /
Disallow: /admin/
Disallow: /internal/
Disallow: /cart/

# 3. User-triggered fetch. A person asked ChatGPT about this page
# and it is retrieving it live, right now.
# DECISION: allow. There is a human waiting on this response.
User-agent: ChatGPT-User
Allow: /
Disallow: /admin/
Disallow: /internal/

# 4. Ad landing page validation.
# DECISION: allow, because we buy ChatGPT ads and OpenAI uses this bot
# to check the landing pages we are paying to send traffic to.
# If you do not advertise on ChatGPT, it only visits pages submitted as ads.
User-agent: OAI-AdsBot
Allow: /
Disallow: /admin/
Disallow: /internal/

Two things to watch when you put this live.

A named section replaces the general one. Each bot follows only the most specific section of robots.txt that names it. Once a bot has its own User-agent section, it ignores your general User-agent: * section completely, including any Disallow lines in it. So every path you block for all bots has to be repeated inside each named section. That's why the Disallow lines above appear more than once.

robots.txt is a request, not a lock. Bots that read the web in bulk, like GPTBot, are documented as following it. Bots that fetch a page because a person asked are less clear-cut. OpenAI says the same of ChatGPT-User: because a person started the visit, robots.txt rules may not apply. Google says Google-Agent, the bot its AI agents use when they browse for a person, generally ignores robots.txt for the same reason. Check each bot's current documentation before you rely on a rule. If you need to actually stop a bot rather than ask it to stay away, do that in your WAF or on your own server. Our AI opt-out guide covers how to do that, and lists the bots from every major AI company.

What to do this week

  • Read your robots.txt and your WAF rules with fresh eyes. Many sites still have old rules that block every OpenAI bot by name. Find out what yours actually says.
  • Check whether OAI-SearchBot is blocked. It's the most expensive mistake on this list, and nothing will warn you about it.
  • Make a deliberate decision about OAI-AdsBot instead of letting a catch-all rule decide for you. Allow it if you advertise on ChatGPT. If you don't, blocking it does no harm.
  • Write the reason and a review date next to every rule. The people who read this file later won't have been in the room when the decision was made.
  • Check your server logs, not just your settings. Search the logs for each of the four bot names and make sure each one is getting the treatment you meant it to get.

Do ChatGPT ads change who gets cited?

It's a fair question: does buying ChatGPT ads affect whether you get cited in ChatGPT answers?

We've seen no evidence that it does. OpenAI's answer independence principle says the answer is written separately from the ad, which is placed after it. Taken at face value, that means citations and ads come from separate systems that run on separate information.

We're only about four months into a self-serve ad product, though, in a market where one company runs both the answers and the advertising, and nobody can say yet how the two will affect each other over time. The principle is clear, and there's no sign it's being broken. But OpenAI earns from both sides, so keep watching as the ad business grows.

What we won't do is claim a connection nobody has shown. If you want to know for your own business, measure it instead of guessing. If you run ChatGPT ads, track how often you're cited in ordinary, unpaid answers before and during the campaign, and see whether anything changes. Our measurement guide shows how to track your citation rate closely enough to spot a change.

For more on why rules for each bot are getting harder to enforce at all, read Atlas is dead, agentic browsing isn't and the court just ruled your agent is you.

References

Get every bot decision right, not just OpenAI's

Our opt-out guide covers the main bots from OpenAI, Anthropic, Google, Perplexity, and others: what each one does, and what blocking it costs you.