← All articles
AI search8 min read

ChatGPT ads: what they buy and what they don’t

ChatGPT ads are labeled and kept separate from the answer. What an ad buys, what the organic answer still decides, and how to choose with measured evidence.

FameScan EditorialResearch & practical guides
A coin pushed halfway into a coin slot, the coin in green: a ChatGPT ad pays for a placement below the answer, not for the answer itself.
The short answer

What matters first.

  • ChatGPT ads are paid, labeled units below ChatGPT’s answer, shown to users on the Free and Go plans. OpenAI says they do not influence answers (checked November 11, 2026).
  • An ad buys a place below the answer. A mention, a citation and a recommendation inside the answer are earned through the sources ChatGPT finds. They are four different wins.
  • Decide per buyer question, with its organic status in hand. FameScan measures that status in ChatGPT and four other engines, with intervals and a “real change versus noise” verdict.

ChatGPT ads are paid, labeled placements that OpenAI shows below ChatGPT’s answer and keeps separate from it. An ad buys a slot under the answer, not a mention, a citation or a recommendation inside it. So the useful question is not “ads or organic?” but which of your buyer questions needs which.

Set up your buyer questions

FameScan measures the organic side of those same questions: it asks ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews the buyer questions you choose, stores every answer with its sources, and shows whether your brand is mentioned, recommended and cited next to your competitors. Every fact about ChatGPT ads in this guide comes from OpenAI’s own help center and announcements, opened on November 11, 2026.

What are ChatGPT ads?

ChatGPT ads are paid, labeled placements that OpenAI shows below ChatGPT’s answer and keeps separate from it: they buy a slot below the answer, not a mention, citation or recommendation inside it. OpenAI began testing them in the US on February 9, 2026, and says plainly that seeing an ad does not mean OpenAI endorses or recommends the advertiser.

They appear for users on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu accounts stay ad-free, accounts identified as belonging to people under 18 get no ads, and Temporary Chats show none. Free users who prefer no ads can switch to an ads-free option with lower usage limits, according to OpenAI’s help article on ads in ChatGPT.

If you know Google Ads, the closest comparison is a search ad that reads the whole conversation instead of a keyword. Advertisers describe where their offer fits with context hints, and OpenAI’s advertiser guide says these hints are not exact-match keywords and do not guarantee delivery in specific conversations. You buy relevance to a conversation, not a fixed spot on a query.

How do ChatGPT ads work?

OpenAI matches one ad unit to the conversation, ranks eligible ads by relevance and bid, and charges per impression or click through its Ads Manager. The table collects what OpenAI’s own pages say. Every row reflects November 11, 2026, and the program is still expanding, so check the linked page before you plan a budget.

TopicWhat OpenAI says, as of November 11, 2026
Who sees adsUsers on the Free and Go plans. No ads for accounts identified as under 18 and none in Temporary Chats. Help center.
Ad-freePlus, Pro, Business, Enterprise and Edu. Free users can choose an ads-free option with lower limits.
WhereOne ad unit below a response, clearly labeled and shown separately. A multi-image format during image generation is in testing.
Parts of an adAdvertiser name, favicon, title, copy, landing page and an image. Advertiser guide.
How an ad is chosenContext and intent of the current conversation, the landing page, title and copy, context hints and targeting. With personalization on, select signals from the user’s wider ChatGPT use. Among several ads, relevance and bids decide.
BuyingAds Manager at ads.openai.com. Objectives Views, Clicks or Conversions. A maximum CPM or CPC bid per ad group; auctions set the price.
Starting bidFor CPC campaigns, OpenAI recommends starting with a maximum bid of $3 to $5 per click.
Advertiser reportingImpressions, clicks, spend, CTR, average CPC and CPM, conversions. Advertisers do not see chats.
ExcludedNo ads near sensitive topics such as personal health, mental health or politics. Political advertising is currently not allowed.
MarketsUS test since February 9, 2026, then a gradual rollout. Expansions announced by OpenAI include the UK, Mexico, Brazil, Japan and South Korea (August 11, 2026), Europe (August 18, 2026) and Southeast Asia and Taiwan (September 23, 2026). Testing ads in ChatGPT, ChatGPT Ads blog.
Self-serviceAds Manager is open to advertisers based in listed countries, among them the US, Canada and the UK (availability list). Personalized ads are not initially available in the EEA or Switzerland (help center).
Newer formatsSponsored Agents are in a limited alpha for selected advertisers. Since November 10, 2026, businesses can sponsor a Business Agent and use it as an ad destination, starting with eligible US ecommerce and online service businesses. Sponsored Agents, announcement.

Two rows matter most for the decision in this guide. The conversation picks the ad, so an ad can follow your buyer questions closely. And the ad sits outside the answer: it changes what the user sees below ChatGPT’s reply and, according to OpenAI, not what the reply says.

Do ChatGPT ads change the answer?

No, according to OpenAI. Its help article states: “Ads do not influence ChatGPT’s answers.” OpenAI adds that ads run on systems separate from its chat model, that advertisers have no ability to shape, rank or alter responses, and that ads are clearly labeled. Its announcement describes them as labeled sponsored and visually set apart from the organic answer.

ChatGPT
Best help desk for small teams?
Earned
Brand
Mention
source
Citation
My pick: Brand
Recommendation
Paid
Brand
Sponsored
Example, not a real ChatGPT answer. Ad placement per OpenAI, checked November 11, 2026.

So an ad buys a slot below the answer. It does not buy your name in the text, a link to your page among the sources, or the sentence where ChatGPT says which option it would pick. All of that comes from the answer itself: from what ChatGPT finds when it searches and from how clearly pages answer the question. Why ChatGPT names one brand and not another is the subject of our guide to ChatGPT visibility.

The organic answer is still yours to earn, and it varies between runs. One answer is a snapshot; the pattern across many answers, engines and weeks is the status you plan a budget on. Measure that status per buyer question with intervals, so a real change stands apart from noise.

Paid slot versus earned answer: four different wins

A sponsored placement, a mention, a citation and a recommendation are four different wins, and only the first one is for sale. A mention is not a citation is not a recommendation, and none of them is an ad. Mixing them in one report hides where the real problem is.

WinHow you get it, and where it counts
Sponsored placementBought in OpenAI’s auction, per impression or click. Shown below the answer, labeled, in ChatGPT only.
MentionEarned in the answer text. Each engine decides on its own.
CitationEarned through the sources the engine finds, shown as a source link. Each engine decides on its own.
RecommendationEarned through comparison sources, in the text: “my pick …”. Each engine decides on its own.

“In ChatGPT only” is the part budgets forget. An ad in ChatGPT runs in ChatGPT. Perplexity, Gemini, Claude and Google AI Overviews build their own answers from their own sources, so the organic wins have to be earned and checked engine by engine. To compare your share against one rival across all of them, see our guide to share of voice in AI answers.

Worked example: the competitor is recommended, we are not

She opens the five weak questions one at a time. Each expanded prompt shows the latest run with its stored answers, and each answer opens with its engine, country, time and the sources ChatGPT returned. Two patterns repeat: the same comparison article and the same review site appear in the answers that pick Supportwell, and Ticketloom’s own pricing page never shows up among the sources.

Filtered to ChatGPT, the Insights page shows her Recommended rate next to her visibility, with the share of answers that pick her when she appears. The Gap analysis lists the sources where competitors are mentioned and Ticketloom is not, with a count of answers without her brand. That becomes her source list.

Then she clicks Ask the Copilot on the weakest question. The Copilot opens and sends a prepared question: who do the AI models recommend instead and why, which sources do they cite, and what content or outreach would get Ticketloom named. The Copilot works from the same stored answers and sources and is asked for a short plan.

Decide per buyer question, not per channel

Decide between an ad and organic work per buyer question, using that question’s organic status in ChatGPT. A channel-level decision (“we do ChatGPT ads now”) spends money on questions you already win and leaves the hard ones to chance.

Status
Next step
Recommended
Hold, keep measuring
Named, not picked
Win the comparison sources
Absent, rival picked
Test an ad + build sources
Decision rule per buyer question, not measured data.
  • Where ChatGPT already recommends you, hold: an ad mostly buys a second appearance next to an answer that already picks you. Keep measuring, because answers vary between runs.
  • Where you are named but not picked, you are in the answer, but the pages ChatGPT leans on make a rival the pick. The Action Center ranks the source and content work, each item with its evidence.
  • Where you are absent and a rival is picked, test an ad and build sources: an ad can buy presence in those conversations now. The source work is what can change the answer itself.

In the example, Ticketloom tests an ad on the five questions Supportwell wins, starts source work on the four middle ones, and holds the three it already wins. Choosing the questions matters more than either tactic. Our guide to prompt tracking explains how to build a question set that reflects real buyers.

Keep the organic side measured in FameScan

Measure the organic answer on exactly the questions you advertise on and the ones you work on, so you know where each one stands in the answer itself. In FameScan that takes four steps:

  1. On the Prompts page, give the questions you advertise on a tag, such as chatgpt-ads, and give the source-work questions another. Topics group them by theme as well.
  2. Filter the Overview and Insights by that tag and by the ChatGPT model, then pick the date range: Last 7 days, Last 30 days, Last 90 days, or any range in the calendar.
  3. Read the panel Visibility with confidence. It compares the selected period with the period of equal length before it and shows visibility, recommendation rate and citation rate with a 95% interval and a verdict: Real change, Within noise or Too few answers.
  4. Add your main rival as a competitor. Alerts are raised after every scored run and only for statistically real movements, for example “Supportwell is gaining visibility” or “Supportwell overtook you”. The weekly digest email sums up the real changes every Monday.

The source work runs through the Action Center. It ranks recommended actions with the evidence behind them and later shows the measured impact of an action: the rate in the days before against the days after, with the same verdict. Ask the Copilot on any item and it drafts the concrete steps, the content to create and the sources to target, from your own data.

Keep the same questions running in Perplexity, Gemini, Claude and Google AI Overviews too. An ad budget in ChatGPT says nothing about those answers, and a rival can win there while you only watch ChatGPT. Why one answer is only a snapshot and the pattern is what counts is covered in our guide to measuring AI visibility.

Agencies pull the same numbers into Claude, Cursor or their own scripts through the FameScan MCP server and REST API, included with every paid plan. Lite costs $12 a month, or $9 a month billed yearly, with 20 Answer Credits a month; Starter schedules daily runs within 200 Answer Credits and costs $49 a month, or $39 a month billed yearly, as of November 11, 2026. Details are on the pricing page.

Track the organic side in ChatGPT

Before the first ad runs, write the stop rule

What you can show your boss or client on Monday is not a forecast but a stop rule: four lines, written before the first ad runs, that say what the test covers and how it ends.

  1. Scope. The ad runs only on the questions tagged chatgpt-ads, the ones where a rival is picked and you are absent.
  2. Ad line. Judged in Ads Manager by impressions, clicks and conversions against the budget you set.
  3. Organic line. Judged in FameScan by the recommendation rate on the source-work questions, with its 95% interval and verdict.
  4. End date. A date you write down today, when you extend or stop the ad on the evidence of both lines.

Report the two lines side by side and never add them into one number. A click on an ad and a recommendation in the answer are different wins, and once they are added together, neither can be judged.

Ten minutes, today: list the five buyer questions where losing a comparison to a competitor would hurt most, and put each one to ChatGPT once. One answer is only a snapshot, so the result is your scope list, not a status. Then add those five in FameScan, or start with the first check, which is free and one time.

Frequently asked

Does ChatGPT have ads?

Yes. OpenAI shows labeled ads to users on the Free and Go plans in regions where ads are available; testing began in the US on February 9, 2026. Plus, Pro, Business, Enterprise and Edu accounts have no ads, accounts identified as belonging to people under 18 see none, and Temporary Chats show none. Availability keeps expanding, so check OpenAI’s help article “Ads in ChatGPT” for the current state.

Do ChatGPT ads influence the answers?

OpenAI says no (checked November 11, 2026): ads run on systems separate from its chat model, advertisers cannot shape, rank or alter responses, and ads are clearly labeled and shown separately below the answer. The organic answer still varies between runs, so judge your status in it over many answers, with intervals that separate a real change from noise.

Can I pay to be mentioned or recommended in ChatGPT’s answer?

No, according to OpenAI (checked November 11, 2026): advertisers cannot shape, rank or alter ChatGPT’s responses, and an ad is a labeled placement below the answer. A mention, a citation and a recommendation inside the answer are earned: through the sources ChatGPT finds when it searches and through how clearly your pages, and the pages about you, answer the buyer’s question. OpenAI also says an ad is not an endorsement.

How much do ChatGPT ads cost?

There is no fixed price. Ads are sold in auctions in OpenAI’s Ads Manager, billed per impression or per valid click, and you set a maximum CPM or CPC bid per ad group. The winning bid sets the price you pay. For click campaigns, OpenAI recommends starting with a maximum bid of $3 to $5 per click, according to its advertiser guide checked on November 11, 2026.

Should we run ChatGPT ads or work on organic visibility?

Decide per buyer question. Where ChatGPT already recommends you, an ad mostly buys a second appearance next to an answer that picks you. Where you are named but not picked, work on the comparison sources. Where a competitor is recommended and you are absent, an ad can buy presence while source work builds the answer. You need the organic status of each question first, measured over many answers rather than one.

Do ChatGPT ads help in Perplexity, Gemini or Google AI Overviews?

No. A ChatGPT ad runs in ChatGPT. Perplexity, Gemini, Claude and Google AI Overviews build their own answers from their own sources, so visibility there is earned and measured engine by engine. FameScan asks the same buyer questions in each engine you track and stores every answer, so you can compare them side by side.

Sources & further reading

Sources checked on 11 November 2026.

Keep reading

From understanding to measurement

What answers do your buyers get?

Start with your brand and your buyers’ questions. FameScan helps you understand mentions, sources and gaps.

Start brand discovery How FameScan worksDiscovery included · dashboard with subscription or active trial