What are ChatGPT Ads and how do they work?

What are ChatGPT Ads?

OpenAI announced the plan on 16 January 2026 and began testing in the United States on 9 February 2026. By 31 August 2026 OpenAI said the product had reached “$1 billion in annualized revenue run rate” and was “used by tens of thousands of advertisers”. On 23 September 2026 OpenAI said “ChatGPT Ads are now available in over 60 countries”, linking that figure to its availability table, which lists the countries a business must be based in to buy ads itself through Ads Manager. Every country on it was marked available when we read it on 4 October 2026.

That growth is the reason this page exists. Seven months after the first test, a product that had never carried a paying advertiser takes money from tens of thousands of them, and most of what is written about it online is written by people selling services around it rather than by the company that runs it. Everything stated as fact below comes from OpenAI’s own pages, linked at the foot.

Where do the ads actually appear?

The match is made from the conversation. OpenAI describes it this way: “Our ads system uses the context of the current conversation to show an ad relevant to what someone is exploring. Depending on the country and a user’s settings, the ads system may also use context from the user’s broader ChatGPT experience.”

One thing worth knowing if you are picturing the screen: OpenAI describes the label the same way everywhere, “clearly labeled as sponsored”, but its own published pictures do not agree on the word. The screenshot in its current help article shows a card marked Ad. The advertiser guidance diagram on another page shows Sponsored. We are not going to tell you which one you will see, because OpenAI is showing both.

Who sees them?

The under-18 rule is worth reading precisely, because it is a strong protection stated in careful language. OpenAI says: “We do not show ads to accounts identified as belonging to users under 18, based on account-level age information and age prediction where available.” That is identification and prediction rather than verification, and it applies to accounts. Where someone is not signed in, OpenAI takes a different approach: “Ads shown in logged-out experiences are appropriate for all audiences, while mature-content ads are limited to eligible logged-in users.”

People can also turn things down. OpenAI’s settings offer an ad-free paid plan, a free option with fewer messages and no ads, a control to switch off ad personalisation, and a per-ad menu to hide an ad or ask why it was shown. Switching off personalisation is not the same as switching off ads: OpenAI says plainly that “it doesn’t remove ads”.

What does an ad look like?

More than one can appear at a time. OpenAI’s wording: “you may see one or more ad units below a ChatGPT response when there’s a relevant match to your chat. That unit may feature one or more items from an advertiser, or multiple advertisers.”

Shops can also build ads from a product feed, a file listing the items they sell. OpenAI calls this a beta, and says these ads use “formats designed specifically for showcasing items from your catalog, featuring details such as product images, titles, stars, prices, sales prices, and your brand”. The feed is for advertising only: OpenAI says the products “will not appear in organic ChatGPT conversations”, while adding that it may offer that in future.

OpenAI’s advertising terms reserve the presentation entirely to itself, “OpenAI may determine and modify the format, placement, size, labeling, and presentation of Ads”, and its ad policy forbids an advertiser from dressing an ad up as part of the product: “Ads must be clearly distinguishable from the ChatGPT product experience.”

What does it cost to run one?

The auction is, in OpenAI’s words, “a relevance-weighted, second-price auction to select among eligible ads”. Relevance-weighted means the highest bid does not simply win; how well the ad fits the conversation is part of the calculation.

The minimum daily budget is published as a table by currency, and it is the figure most often got wrong elsewhere. OpenAI lists 25 US dollars, 15 pounds sterling and 15 euros among others.

A daily budget is an average rather than a ceiling, and this catches people out: OpenAI states that “Your maximum daily spend is twice your daily budget” and “Your maximum seven-day spend is seven times your daily budget”. So a 15-pound daily budget can take 30 pounds on a single day, and 105 pounds across a week.

Billing happens after delivery rather than up front, OpenAI describes it as a “postpay billing model that charges advertisers after ads have been delivered”.

What has to be true of your website first?

This is the part with real money attached and it is easy to miss. When something is wrong, OpenAI’s own troubleshooting guidance describes the symptom an advertiser sees: an ad whose status is “Not serving”, where “Common causes include ad copy or landing pages that do not align with our Ad Policies, or landing pages that block OpenAI user agents or IP addresses.”

There is no manual override. OpenAI’s advice is explicit: “Do not rely on a manual bypass. Make the landing page crawlable for OAI-AdsBot by fixing robots.txt, WAF, CDN, bot mitigation, authentication, and rate-limiting blocks.”

The money side is worth reading twice, because it runs the opposite way to most people’s assumption. OpenAI’s advertising terms say: “OpenAI may suspend delivery of an affected Campaign if a Destination is inaccessible or restricted, and you remain responsible for all Fees incurred.” Your ads stop; the bill for what already ran does not.

In plain terms: nobody at OpenAI can wave your ad through. Whatever is turning those requests away on your site has to be switched off, and the ad may then need submitting again.

One common cause is not a decision anybody made. OpenAI names it: “Many websites use services such as Cloudflare, Akamai, or other web protection providers… These systems can mistakenly block legitimate crawlers, often returning 403 Forbidden errors.” In plain English, the security service in front of your website turns OpenAI away by mistake, and nothing on your screen tells you. OpenAI says on the same page that its ad checker “is officially verified and allowlisted by Cloudflare”, while still naming Cloudflare among the services that can block it by mistake.

What OpenAI has not published

Two other gaps are worth naming, because they get filled in confidently elsewhere and we could not confirm either. OpenAI does not publish the exact words shown on the sponsored label. And it publishes nothing about whether the same ad can be shown to you twice in one conversation, no frequency limit, no repeat rule, and no reporting figure that would reveal one. We went looking for both and found nothing, so this page says nothing about them.