What is ChatGPT Ads?
ChatGPT Ads is the name of the ad inventory OpenAI sells inside the ChatGPT interface. The ad appears where the answer to the user's question ends, as a block visually separated from the answer and marked with a sponsored label. OpenAI draws this distinction explicitly in its documentation: ads run on systems separate from the chat model, and advertisers have no ability to shape, rank or change ChatGPT's answers.
The unit is not visible to everyone. Ads are shown to users on the Free and Go plans; no ads appear on Plus, Pro, Business, Enterprise and Edu accounts. For Free plan users who do not want to see ads, there is also an ad-free mode in exchange for lower usage limits and reduced access to features. In other words, the audience an ad can reach is not all ChatGPT users but the segment on free and entry-level plans.
Geographic availability is still evolving. The ad unit was first tested in the US, then opened up to a number of markets including 31 countries in Europe, and self-serve advertiser access is expanding gradually. OpenAI has not explicitly listed Türkiye in its announcements. Anyone planning to run campaigns from Türkiye should verify access in their own ad account rather than assuming it.
How do ads appear in the chat flow?
Structurally, the placement differs from an ad in classic search results. The user asks a question, ChatGPT generates an answer, and once the answer is complete, if there is a sponsored product or service relevant to the context of the conversation, one or more ad units appear below the answer. A unit can contain several items from a single advertiser or items from different advertisers.
The trigger is not a list of keywords the user typed but the conversation itself at that moment. This changes the matching logic advertisers are used to. The user may have described a problem without ever typing a product name, and the ad is matched to the context of that problem.
There is an even more important structural difference. In search results, the ad appears at a moment when the user has not yet received an answer, and that is where its chance of being clicked comes from. In ChatGPT, the ad comes into play after the user has already received the answer. The ad unit does not fill in missing information; it suggests an action after a completed answer. That is the point at which the ad copy and landing page need to persuade the user. OpenAI also states that an ad appearing does not mean it endorses or recommends that advertiser or product.
Why are citations and ads separate channels?
This is the point most often confused when ChatGPT visibility is discussed. ChatGPT using a site as a source in an answer and the same site appearing in a sponsored block below the answer are the outputs of two separate systems. They can sit one above the other on the same screen and still be independent of each other.
An ad placement can be bought. You allocate budget, set up a campaign, and if the ad and landing page pass policy review, it goes live. A citation cannot be bought. Whether a page is selected as a source in ChatGPT's search features depends on the page being crawlable, genuinely containing the answer to the question and being judged worth using over other candidates at the selection stage. No link in this chain is managed from an ad account.
| Dimension | Ad placement | Organic citation |
|---|---|---|
| How it is obtained | A campaign is set up and budget is spent | A page suitable for selection as a source is produced |
| Relevant bot | OAI-AdsBot | OAI-SearchBot |
| Position on screen | A separate block below the answer | A source reference within the answer |
| Label | Marked as sponsored | Carries no sponsored label |
| Level of control | Copy, image and landing page are up to the advertiser | Selection is up to the system, page quality is up to the site |
| When the budget stops | Visibility ends | Visibility does not depend on budget |
To turn this table into an investment decision, one sentence helps: advertising is renting visibility, while citation is trying to own it. One does not replace the other. When the ad budget is cut, the unit disappears from the screen. A citation may continue, but no one can sell you a guarantee of it. Treating the two as alternatives within the same budget line misreads how the channels work. The same logic applies to the relationship between Google Ads management and organic search work on the search side, but in ChatGPT the separation is sharper, because the ad system and the model that generates the answer are technically separated.
What does OAI-AdsBot do?
OAI-AdsBot is a separate crawler OpenAI runs for the ads side, and it identifies itself as OAI-AdsBot/1.0 in its user agent. Its task is narrowly defined: it visits only pages submitted as ads in ChatGPT and verifies that those pages comply with OpenAI's advertising policies. It is not a general web crawler and does not roam the whole site.
The bot has a second function, and this is the point advertisers most often miss. OpenAI states that it may also use the content of the landing page to decide when, and to which users, the ad is relevant. In other words, the page does not only go through a safety and policy scan; it is also read. If the page does not state clearly what it sells, whom it is for and which problem it solves, the chances of the ad being matched with the right conversations are weaker.
The data the bot collects is not used to train generative AI foundation models. This is the clearest difference between it and GPTBot.
What happens if OAI-AdsBot is confused with the search bot?
OpenAI has four separate bots, and each does something different. Lumping them together leads to concrete mistakes.
| Bot | What it comes for | robots.txt behavior | Model training |
|---|---|---|---|
| GPTBot | Collects content to train foundation models | Stated to comply | Yes |
| OAI-SearchBot | Crawls the site to show it in ChatGPT search features | Stated to comply | No |
| ChatGPT-User | Carries out browsing and actions initiated by the user | Stated that rules may not apply | No |
| OAI-AdsBot | Verifies pages submitted as ads | Not specified in the documentation | No |
The last row of the table needs attention. OpenAI writes explicitly that robots.txt rules are respected for GPTBot and OAI-SearchBot. In the OAI-AdsBot entry, however, this behavior is not defined. In addition, there is no OAI-AdsBot equivalent of the IP range files published for the other bots, which means verifying on the server side via IP that a request really came from OpenAI is not as easy as with the other bots. These two gaps make any plan that reduces bot access to a robots.txt line fragile.
In practice, the most common mistake is grouping the bots under a single rule. Blocking GPTBot to stay out of training data does not prevent being selected as a source in ChatGPT's search features, because that job belongs to OAI-SearchBot. Conversely, blocking OAI-SearchBot wholesale pushes the site out of the source pool but does not touch the ads side. Someone who writes a blanket OpenAI block can cut off both their chance of being selected as an organic source and the verification of their own ads at the same time. Starting with an AI crawl audit to see which bots your site is actually open to is safer than proceeding on guesswork.
What to prepare on the landing page
What an advertiser can control is largely the landing page. OpenAI's advertiser guidelines are clear on this: landing pages must use valid links and must not block OpenAI user agents, namely OAI-AdsBot and OAI-SearchBot. If the page cannot be verified, the ad should not be expected to go live.
Access barriers are not just robots.txt
Blocks do not come only from a robots.txt line. Among the things that break crawlability, OpenAI also lists firewalls, the CDN layer, bot mitigation solutions, authentication walls and rate limiting. Aggressive bot protection can reject a request before robots.txt ever comes into play. A page that requires login, is geo-restricted or is open only to certain user agents does not count as accessible for verification, even if it technically works.
This check is not a one-off. Monitoring continues after the ad is approved. If the page becomes inaccessible or inconsistent with the ad, the ad may be rejected or taken down. When access to the page is restored, the ad may need to be resubmitted and go through review again. An infrastructure change on the site side can stop an active campaign without anyone noticing.
Page content must be consistent with the ad
Review does not look only at accessibility. The page is expected to be consistent with the message in the ad, relevant to the topic and safe. That is why a campaign should point to the most relevant product, collection or content page rather than a generic home page. Every gap between what the ad says and what the user finds on the page creates risk, both in review and with users.
Ad copy fields are tight. There are upper limits of 50 characters for the headline and 100 characters for the description, and images are expected to be simple, consistent with the message and legible. When a promise is made in such a narrow space, the page needs to deliver on it visibly, above the fold.
The review process is mostly run by automated systems, supported by human oversight. Some verticals were closed at launch. In areas such as financial services, health and legal, approval may be subject to additional conditions; for example, proof of licensing may be required for financial services. Checking where your own sector stands on this list before setting up a campaign keeps the landing page you prepare from going to waste.
Finally, measurement should not be forgotten. Adding your own tracking parameters to destination URLs is the only practical way to tell which creative is driving traffic, because traffic that comes from a chat context is not easy to isolate by default.
Who this channel makes sense for
ChatGPT Ads is a channel worth considering for advertisers that sell in a market where ads are available, operate in a permitted category and have a landing page that passes bot verification cleanly. The initial scope has been kept broader for consumer-facing verticals, while the door is narrow in regulated areas.
There are also situations where the channel is premature. For an advertiser whose market does not yet have the unit, whose category is on the restricted list or whose landing page infrastructure sits behind strict bot protection, setting up a campaign is too early. In that case the job is not to wait for ads but to fix the page's accessibility and content clarity now, because the same preparation pays off on the organic side too.
The real decision here is this: an ad budget does not replace the work of being cited as a source in ChatGPT. An ad buys a placement at the end of the conversation. Being cited as a source means getting into the answer itself, and that is only possible when the page genuinely carries usable information. The two efforts support each other; one does not substitute for the other. AI visibility work on the organic side and an ad campaign should be tracked with separate goals and separate metrics. For concrete steps on the organic side, how GEO work is carried out for ChatGPT is a separate topic and proceeds independently of advertising.
Frequently Asked Questions
Does advertising on ChatGPT get my site shown as a source?
No. OpenAI states that ads run on systems separate from the chat model and that advertisers have no ability to shape, rank or change ChatGPT's answers. An ad budget does not add your page to the source pool.
What happens if I block OAI-AdsBot with robots.txt?
If you are running ads, your landing page cannot be verified and the ad should not be expected to go live. In addition, OpenAI's documentation does not specify how OAI-AdsBot handles robots.txt rules, so it would be wrong to assume this file works as an access control mechanism on its own. For a site that does not advertise, the bot does not come anyway, because it only visits pages submitted as ads.
How do I identify OAI-AdsBot visits in server logs?
The string OAI-AdsBot/1.0 appears in the user agent. Because there is no OAI-AdsBot counterpart to the IP range files published for other OpenAI bots, verifying the request via IP is not as easy as with the other bots. For that reason, a log rule that only looks at the user agent remains vulnerable to spoofed requests.
Which ChatGPT plans show ads?
Ads appear on the Free and Go plans. No ads are shown on Plus, Pro, Business, Enterprise and Edu accounts. On the Free plan, users can switch to an ad-free experience in exchange for lower usage limits.
Is the performance of ads and organic citations measured in the same place?
No. The ads side is measured through impressions and clicks in the campaign interface. Citations are tracked with query-based tests, that is, by repeating specific questions at different times and with different wording and monitoring which sources are used in the answer. The two measurements require separate methods and do not come together in a single dashboard.
