How Do ChatGPT Ads Work? ChatGPT Ads vs Google Ads and Meta Ads

How do ChatGPT ads work?

ChatGPT ads are OpenAI's sponsored placements that appear below a ChatGPT answer for users on the Free and Go plans, whenever the conversation matches an advertiser's offer. Advertisers set context hints, locations and bids, and OpenAI's system then picks the winning ad in an auction that weighs relevance alongside the bid.

I have run search and social campaigns since 2012, and I have learned to meet every new ad channel with equal parts curiosity and caution. ChatGPT ads also fit that pattern: the channel is growing fast, yet OpenAI still describes it as a beta.

Because the channel changes fast, I took product details, dates and rules only from OpenAI's own announcements and Help Center. I also mark press figures clearly as press reports. So everything here reflects the status on 29 September 2026. If you want the user side of the story, my comparison of ChatGPT vs Google as a search tool covers it; here I focus only on advertising.

When did ChatGPT ads launch, and where do they stand now?

OpenAI announced its ad principles in January, started a US test in February and widened its buying tools in spring. Over the summer, the pilot then turned into an international ad network. These are the milestones from OpenAI's own posts:

• 16 January 2026: OpenAI published its advertising principles and said it would test ads for Free and Go users in the US. Go also launched in the US that day at 8 US dollars a month.
• 9 February 2026: The test then began for logged-in adult users on the Free and Go plans in the US.
• 26 March 2026: OpenAI reported no impact on consumer trust metrics and low ad dismissal rates, then announced pilots in Canada, Australia and New Zealand.
• 5 May 2026: A beta self-serve Ads Manager, cost-per-click (CPC) bidding, the Conversions API and pixel-based measurement arrived.
• 11 August 2026: Ads also went live in the United Kingdom, Mexico, Brazil, Japan and South Korea.
• 18 August 2026: OpenAI announced its expansion to 31 European markets, including Germany, France, Spain and Italy.
• 31 August 2026: OpenAI said ChatGPT Ads had reached a 1 billion dollar annualised revenue run rate in under 200 days, and it opened self-serve buying across India, Europe, the Middle East and North Africa.
• 23 September 2026: Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan joined. That step also took the total past 60 countries.

That pace is impressive; still, keep it in perspective. OpenAI's Help Center states plainly that the platform does not yet have performance benchmarks across advertisers, industries or campaign types. As a result, the most valuable data you can hold today is the data from your own small test.

Who sees ChatGPT ads, and who never does?

Only users on the Free and Go plans can see ads. According to OpenAI's Ads in ChatGPT FAQ, Plus, Pro, Business, Enterprise and Edu accounts carry no ads at all. In other words, most paying users sit outside this channel.

The age rule is also strict. OpenAI does not show ads to accounts it identifies as belonging to people under 18. To do that, it combines the stated age with signals such as account age, typical times of use, usage patterns and the general topics a person discusses. Ads never appear in these situations:

• Plus, Pro, Business, Enterprise and Edu accounts.
• Accounts that OpenAI identifies as belonging to under-18s.
• Temporary Chats.
• The ChatGPT Atlas browser during the test.
• Sensitive or regulated conversations such as personal health, mental health or politics.

Ads can also appear in logged-out ChatGPT, where OpenAI says every ad must suit all audiences. Meanwhile, Free users who dislike ads can switch to an ads-free option with lower message limits. That option also removes access to some tools, such as image generation and deep research.

Where can you buy ChatGPT ads today?

As of 29 September 2026, OpenAI's Ads Manager availability page lists more than 60 countries with self-serve access, including the United States, the United Kingdom, Canada and Australia. One condition matters here: the legal entity that runs and pays for the ads needs its home base in a listed country.

Some new self-serve accounts can advertise only in their home country at first. Identity verification and a required level of home spend then unlock other supported countries. Larger advertisers can also work with OpenAI's Ads Solutions team or through agency and technology partners.

On the user side, OpenAI rolls ads out gradually to eligible Free and Go users in selected regions. If you join the test, you then see clear in-product information and an Ad Controls section in Settings. So do not expect every user in a market to see ads on the same day.

Do ads change ChatGPT's answers?

OpenAI's answer is a clear no. Under its "answer independence" principle, ads do not influence ChatGPT's responses, and the model optimises answers for what is most helpful to you. The Help Center goes further: ads run on systems separate from the chat model, and advertisers have no ability to shape, rank or alter responses.

In practice, this means an ad budget does not buy your brand a mention inside the answer. Instead, your ad stays below the response, carries a sponsored label and sits in a visually separate area. OpenAI also stresses that seeing an ad does not mean it endorses or recommends the advertiser.

Users can still pull an ad into the chat on purpose. If you tap the three-dot menu on an ad and choose Ask ChatGPT, you share that ad with ChatGPT, and it can answer questions about it. By default, ChatGPT cannot see the ads it shows you. Advertisers also have no influence over its answers about a shared ad.

What does a ChatGPT ad look like in the chat?

Ads appear below the end of a response. When there is a relevant match, you may see one or more ad units under the same answer. A unit can also feature several items from one advertiser or items from several advertisers. In longer conversations, the system also weighs the overall context and user experience.

For advertisers, the Help Center describes a standard ad unit with these parts:

• Advertiser name
• Favicon, the advertiser's logo
• Title, the headline
• Copy, the description
• Landing page
• Image

Beyond that, you can build campaigns from a product feed, and OpenAI listed product feeds among its key capabilities on 31 August. Users can also message an advertiser through an ad; in that case the advertiser sees only the messages the user sends directly.

One creative rule deserves attention. Specifically, OpenAI's ad policies ban ads that imitate the look, functionality or voice of ChatGPT. An ad that tries to pass as a ChatGPT answer will therefore struggle in review.

How does ChatGPT decide which ad to show?

OpenAI's ad system selects ads on expected relevance and expected outcomes. According to the Help Center, it weighs these signals together:

• The context and intent of the current conversation
• The ad's landing page, title and copy
• The advertiser's context hints and targeting choices
• Where personalisation is on, broader signals such as past chats, memory and ad interactions

When several ads qualify, relevance and the advertiser's bid decide the order together. OpenAI calls this a relevance-weighted, second-price auction. Put simply, the highest bidder cannot win with an irrelevant ad. Under second-price logic, the winner also tends to pay what it takes to beat the runner-up rather than its full bid.

For example, OpenAI notes that a user researching recipes may see ads for meal kits or grocery delivery. In short, the system looks at the topic of the conversation through a wider window than a single search query.

How do context hints differ from keywords?

Context hints sit at ad group level and describe the conversations, topics or keywords where your product may be relevant. However, OpenAI states that they are not exact-match keywords and do not guarantee delivery in specific conversations.

In Google Ads, match types let you draw the boundary around a query yourself. In ChatGPT, however, a hint works more like directions; the system that reads the whole conversation makes the final call. Therefore, I suggest you write hints as short descriptions of a customer's situation rather than one-word lists.

For example, a company selling bookkeeping software to small businesses might use hints like these:

• A small business hiring its first employee and setting up payroll and invoicing
• A freelancer looking for one place to track income and expenses
• A team deciding to move its bookkeeping from spreadsheets to software

Each hint should match a question your landing page answers clearly. The keyword suggestion tool on this site also helps with raw topic ideas; turning that list into situation-based hints is your job.

Which topics and industries are off-limits?

OpenAI's ad policies work on two levels. The first level decides which conversations can carry ads at all. It rules out topics such as child safety, hate and harassment, graphic violence, suicide and self-harm, political content, regulated goods like gambling or tobacco, and personal health conversations.

The second level defines which industries may advertise. Under the policy as of late September 2026, the early focus lies on household and consumer goods, local services, travel and entertainment, and digital products and education. On the other hand, these areas remain closed or tightly limited:

• Currently disallowed: gambling, alcohol and tobacco, adult content and dating apps, political ads and recreational drugs.
• US only, case by case: financial services, health services and legal services, often with proof of licensing; outside the US, OpenAI generally prohibits them.
• Individual listings: single job ads and single housing rentals or sales.

So for a UK bank, a private clinic or a law firm, the message today is simply "wait". Policies may change, so check your category against the official policy page before you set aside any budget.

Can advertisers see your conversations?

No. OpenAI says it does not share conversations with advertisers and never sells user data to them. Instead, advertisers receive only aggregated, non-identifying performance data, such as total views or clicks. Chats, chat history, memories, names, email addresses, precise location, IP addresses and sensitive information never reach them.

Meanwhile, users find their controls under Settings > Ad Controls. From there you can:

• Turn off ad personalisation
• Stop past chats and memory from shaping ads
• Review your ad history and topics
• Delete your ads data
• Hide or report an ad, or open About this ad to see why it appeared

Users who turn off personalisation still see ads, but those ads rely only on the current chat. Deleted ads data also leaves OpenAI's systems within 30 days. In the European Economic Area and Switzerland, personalised ads are not available at launch at all.

How do you start advertising on ChatGPT?

There are three routes today. First, large brands can work directly with OpenAI's Ads Solutions team. Agencies and technology partners form the second route; OpenAI names agency groups such as dentsu, Havas Media, Omnicom Media, Publicis Groupe and WPP, plus technology partners such as Adobe, Criteo, Kargo, Pacvue and StackAdapt. Finally, the third route is the self-serve Ads Manager.

In the self-serve route, the steps look roughly like this:

1. Sign up at ads.openai.com and enter your business details.
2. Complete the identity and business verification that OpenAI requests.
3. Add a payment method.
4. Create a campaign with an objective, budget, dates and countries.
5. Set up ad groups and context hints, then upload your ads.
6. Launch after review and watch your reports.

Every ad also passes a review before it can serve. If a campaign does not deliver, check account verification, billing status, campaign dates and the budget limit first. OpenAI's own FAQ recommends the same checklist.

Which campaign objectives and bidding models can you choose?

Every campaign in Ads Manager starts with an objective, and that objective sets both pricing and optimisation. You cannot switch the objective later, so a different goal needs a new campaign. According to OpenAI's campaign setup guide, there are three options today:

• Views (CPM): You pay per 1,000 impressions, which suits reach and awareness.
• Clicks (CPC): You pay per valid click, which suits driving traffic.
• Conversions: The system optimises for a conversion event you choose, while you pay per valid click (oCPC) or per 1,000 impressions (oCPM), not per conversion.

Next, you set a maximum CPM or CPC bid at ad group level. For CPC campaigns, OpenAI recommends starting with a maximum bid of 3 to 5 US dollars per click. Ads Manager may also show a bid strength indicator that tells you whether your bid is likely to compete.

According to OpenAI's 31 August update, CPC and outcome-optimised bidding now account for the majority of campaigns. In other words, the channel has moved from a brand showcase towards a performance channel.

How much do ChatGPT ads cost?

Let me start with the official numbers. The minimum daily campaign budget depends on your billing currency: 25 USD for US dollar accounts, 15 GBP for sterling, 15 EUR for euro and 20 CHF for Swiss francs, among others. In practice, a daily budget works as an average across seven days. Spend on a single day can reach twice the daily budget, but the seven-day total never exceeds seven times that amount.

Press reports show how quickly prices have come down. Adweek reported that OpenAI asked select advertisers to commit at least 200,000 dollars at the start of the beta. On 17 April 2026, Digiday reported buyers saying CPMs had fallen from about 60 dollars at launch to roughly 25 to 45 dollars, and that the minimum spend had dropped from 250,000 to 50,000 dollars.

However, none of these figures is an official rate card; they shift with auction dynamics and category. To run your own numbers, use the CPM calculator for impression scenarios and the ROAS calculator to see your break-even point before you spend.

What targeting options do you get?

Conversation context sits at the heart of ChatGPT targeting. Still, you get a few concrete settings at campaign level:

• Country and location: countries, plus states or regions, cities, markets and postal codes where available.
• Platform: Android app, Android web, desktop web, iOS app and iOS web.
• Custom audiences: include or exclude your own customer lists.

Custom audiences come with a threshold: inclusion targeting needs at least 25,000 matched users, while exclusions can use smaller lists. New product feed campaigns, meanwhile, support location targeting at country level only.

If you target Europe, also note one more limit. Because ad personalisation is not yet available in the EEA and Switzerland, OpenAI advises against custom audiences in campaigns aimed there. Instead of Meta-style interest lists, context hints and carefully written ad copy do the heavy lifting.

How do you measure ChatGPT ads?

Ads Manager reports currently include impressions, clicks, spend, click-through rate (CTR), average CPC, average CPM and conversions. Impressions, clicks and CTR refresh roughly every 15 minutes; spend data, however, may lag by 7 to 8 hours. So a spend figure of zero on day one does not mean you have no charges yet.

For conversions, OpenAI also offers the OpenAI Pixel, the Conversions API and third-party measurement integrations. You can also add static UTM parameters to landing page URLs; they persist on ad clicks and let you separate ChatGPT ad traffic in GA4 or any other analytics tool. The UTM builder makes that setup quick.

Separating organic ChatGPT visits from paid clicks is a task of its own, and I cover it in my guide to detecting AI traffic in GA4. One more note: install the pixel with proper consent, especially under UK GDPR and PECR in the UK or state privacy laws such as California's in the US.

When my team and I set up measurement, we pay the most attention to post-click behaviour. A visitor from ChatGPT has probably weighed the options in the chat already. Your first screen should therefore finish that conversation with a clear, specific answer.

ChatGPT ads vs Google Ads vs Meta Ads: what is the difference?

The core difference lies in the signal each channel relies on. Google Ads responds to the query a user types, Meta Ads to who the user is and how they behave, and ChatGPT ads to the context of the current conversation. The table below summarises the differences using the official help pages:

Read the table with one idea in mind: these channels do not replace one another; they catch different moments of the same buying journey. If you want to know what good Google Ads management involves, see what Google Ads consulting includes. Likewise, for Meta budgeting logic, my guide on calculating a social media advertising budget will help.

Does Google's search intent still beat ChatGPT?

My short answer: yes, at least for now. On Google, people type what they want in their own words, so you bid on that query directly. No other system yet matches that clarity for capturing existing demand. Google's inventory, measurement stack and automated bidding have also matured over many years.

That said, Google is bringing ads into AI answers too. According to the Google Ads Help Center, ads can appear above or below AI Overviews in more than 200 markets. Ads inside AI Overviews, however, currently run only in English in 12 countries, including the US, Canada, Australia and India. The UK is not on that list, and advertisers also can neither target nor opt out of that placement.

ChatGPT's difference is that it reads intent from a whole conversation rather than a single query. In practice, users who explain their budget, constraints and priorities write their own brief. As a result, for considered purchases with several steps, that context can be richer than any single search. If you want stronger search campaigns in the meantime, my team offers Google Ads management for exactly that.

Where does Meta's audience model differ from ChatGPT?

On Meta, an ad reaches people before they have asked anything. The system then predicts who might care about your offer from demographics, interests, behaviour and lookalike audiences. Meta's auction then ranks ads by a total value that combines the bid, the estimated action rate and ad quality.

That is why Meta remains strong at creating demand and at visual storytelling. If you launch a new brand or want to bring back cart abandoners, Meta's in-feed formats and remarketing audiences do the job well. In ChatGPT, by contrast, the user has already put the need into words; your ad does not plant a new desire but offers an option for an existing question.

In practice, I recommend giving each channel a role instead of pitting them against each other. For example, Meta can carry awareness and reminders, while ChatGPT covers visibility at the moment of comparison. Both channels then reach the same person at different moments, and your budgets stop competing.

Which businesses should test ChatGPT ads now?

OpenAI describes ChatGPT as a place where people explore options, compare choices and make decisions. That description makes the channel relevant right away for some businesses:

• Travel, accommodation and experience brands.
• Consumer products that people compare before buying, and e-commerce shops with a product feed.
• Digital products such as software, apps and online courses.
• Local services where choosing a provider takes some research.

On the other hand, some businesses should wait for now. Companies in closed categories, for instance, cannot advertise anyway. Very small budgets also struggle with the minimum spend and the learning period, so those businesses should first strengthen their Google and Meta foundations.

Also remember that the channel is young. OpenAI has shared an e-commerce advertiser that achieved a 3x return on ad spend over 28 days, and a technology partner that reported more than 80% of ad-driven ChatGPT traffic coming from new customers. However, OpenAI chose these examples itself, so they do not guarantee the same result in your industry.

What should you do before you spend?

Instead of only watching the channel evolve, you can prepare at low risk. Here is the order my team and I recommend to brands:

1. Check eligibility: Is your company's country on the Ads Manager list, and does the ad policy allow your category?
2. Build measurement first: Prepare a UTM template, a GA4 channel split and, if you use the OpenAI Pixel, consent management.
3. Write decision-stage content: Create landing pages that answer comparison, pricing and "which one fits me" questions.
4. Phrase hints as situations: Write a handful of context hints that describe what your customer is going through.
5. Run a small, time-boxed test: Start with the Clicks objective, a budget you can afford to lose and a fixed end date.
6. Keep existing channels running: Fund the test from extra budget, not from Google Ads or Meta campaigns that already work.

Above all, the goal of the first months is learning, not profit. Seeing which contexts earn impressions, which messages win clicks and what happens after the click gives you the basis for your next budget decision.

How do organic AI visibility and paid ads work together?

Because ads do not influence ChatGPT's answers, getting your brand named inside the answer is a separate job. Whether ChatGPT lists you among the options depends on your reputation across the web, the clarity of your content and your footprint in independent sources.

Ideally, the two layers support each other: the answer mentions you organically, and your ad below it offers a concrete next step. I also explain the organic side in my article on how your brand shows up in ChatGPT and Gemini. In our AI SEO and GEO services, my team tracks both layers in the same measurement framework.

One warning, though: an ad cannot hide a weak organic presence. So if ChatGPT recommends your competitors in the answer while you appear only in the ad slot, users may trust you less. Take a snapshot of your organic AI visibility before your first test.

What should you realistically expect from ChatGPT ads?

ChatGPT ads bring a new model that reads intent from conversation, and they reached more than 60 countries within months. In the US and the UK, advertisers can buy them today; users, however, see them through a gradual rollout. The rules are clear: ads do not change answers, chat data stays away from advertisers and many regulated industries remain outside for now.

My advice, therefore, is neither to overhype the channel nor to ignore it. Google Ads keeps capturing existing demand and Meta keeps creating it, while ChatGPT opens a new window at the moment of decision. A small, measured and well-tracked test will give you the most reliable data on where your budget should go next.