AI Advertising

ChatGPT Ads: How They Work, What They Cost, and How to Win

By Alyssa Caridi · Last updated July 5, 2026

• ChatGPT Ads
• AI Advertising
• GEO
• Marketing

ChatGPT ads are sponsored placements that appear at the bottom of a ChatGPT answer when a product or service is relevant to the conversation. OpenAI began testing them in the United States on February 9, 2026, for logged-in adults on the free and Go tiers, and paying Plus, Pro, Business, and Enterprise users do not see them (OpenAI, 2026). Ads are clearly labeled, kept separate from the answer, and OpenAI says they do not influence what ChatGPT tells you. For brands, this is a new high-intent channel reaching a large share of ChatGPT's audience of more than 900 million weekly users. This piece covers how the ads work, who sees them, what they cost, what early performance shows, and how to advertise well before the platform opens to everyone.

What are ChatGPT ads and how do they work?

A ChatGPT ad is a single sponsored card that appears below the assistant's response when the conversation surfaces a relevant product or service. OpenAI describes the format as clearly labeled and visually separated from the organic answer, with a way for users to learn why they saw an ad or dismiss it and give feedback (OpenAI, 2026). OpenAI has committed to answer independence, meaning ads do not change the substance of a response, and the company says it does not optimize for time spent in the product. The design points toward conversational formats over time. OpenAI has said that a user might see an ad and then ask it questions directly before making a purchase decision, which is a capability no search or social placement offers today.

Who sees ChatGPT ads, and who does not?

During the test, ads reach logged-in adults in the United States on the free tier and the Go tier, which costs 8 dollars per month. Subscribers to Plus, Pro, Business, Enterprise, and Education plans see no ads (TechCrunch, 2026). OpenAI also withholds ads from accounts it predicts belong to users under 18, and it does not place ads near sensitive or regulated topics such as health, mental health, and politics. For advertisers, that means the reachable audience is the large, mostly consumer base on the lower tiers, and the content around your ad is filtered away from the categories most likely to create brand-safety problems. It also means your highest-value enterprise buyers, who tend to sit on paid seats, may never see the placement.

How does targeting and privacy work?

Targeting is contextual first. OpenAI matches ads to the subject of the current conversation, past chats, and previous ad interactions, so someone researching recipes might see a grocery delivery or meal-kit ad (TechCrunch, 2026). This is different from keyword targeting in Google or interest targeting in Meta, because the signal is a live, stated intent expressed in natural language. On privacy, OpenAI says conversations stay private from advertisers, data is not sold, and advertisers receive only aggregate performance such as views and clicks. Users can turn off personalization, view and clear their ad interaction history, and always have a paid path to an ad-free experience. For a brand, the practical takeaway is that you buy relevance to a moment of intent, and you will not get user-level conversation data back.

What do ChatGPT ads cost, and how do you buy them?

Access runs through the self-serve OpenAI Ads Manager, which as of May 2026 carried no minimum spend and supported both CPM and CPC bidding, with cost-per-click added after an initial impression-only launch (AdVenture Media, 2026). Reported rates during the early period put CPM roughly between 25 and 60 dollars per thousand impressions, and CPC between about 2.50 and 8 dollars, with business-to-business clicks reaching the top of that range (Launchcodex, 2026). Those numbers sit above typical Google Search costs. The premium reflects a smaller audience that arrives in research mode with a specific question, which is worth more per impression than a broad display view. Treat these figures as early and moving, because pricing on a channel this young changes quickly.

What does early ChatGPT ads performance actually show?

The early data is encouraging on intent and honest about its own limits. OpenAI Ads crossed 100 million dollars in annualized revenue within roughly six weeks of opening self-serve access, with more than 600 advertisers on board, which shows real demand (AdVenture Media, 2026). Reported pilot click-through rate landed near 0.91 percent, well under the Google Search benchmark, but click-through is a poor primary metric here, because users often keep talking after seeing an ad instead of clicking away. On conversion, Criteo data cited by early advertisers found that referrals from large language model platforms convert about 1.5 times better than other digital channels, and that visitors arriving from ChatGPT spent 60 to 80 percent more time on site than social traffic (Launchcodex, 2026).

The honest counterweight is measurement. In the managed pilot, several advertisers could not prove business results because the tools to attribute them did not yet exist. The ads may well have driven action that the reporting simply could not see. One enterprise advertiser reportedly spent only 3 percent of a 250,000 dollar pilot budget over several weeks and had no way to tell whether the impressions drove any action (The Keyword, 2026). The pattern for early movers is a channel that appears to attract high-intent, high-converting traffic while the reporting to prove it is still being built.

What are the best practices for advertising on ChatGPT?

The brands getting the most from the early phase treat it as a decision environment where people arrive with an active question, not as another slot to recycle display creative into (Forbes Agency Council, 2026). A few practices stand out.

• Write for the conversation. The context, pacing, and intent are different from search, so copy that helps the user move a decision forward will outperform a reworded Google ad.
• Match the landing page to the intent. Send clicks to a fast, mobile-first page that states the offer, the proof, and the next step near the top, because conversational traffic arrives curious and needs trust signals early.
• Build attribution for assisted conversions. Standardize UTM naming and use cross-channel dashboards, since ChatGPT often influences a purchase that closes on another channel later.
• Pair paid with organic answer-engine optimization. Brand visibility inside ChatGPT falls fast when a campaign pauses, so a paid placement works best on top of content the model already tends to cite. This is exactly what generative engine optimization builds.
• Start small and learn. With no minimum spend, a controlled test teaches you how your category behaves before you commit real budget.

What are the watch-outs and risks?

The same newness that makes this channel interesting also makes it risky, and a few issues deserve real attention before you spend (AdVenture Media, 2026).

• Brand safety works differently. A ChatGPT ad sits inside a living conversation that can go almost anywhere, so a keyword exclusion list built for display advertising will not protect you the same way.
• The platform is immature. Reporting, controls, and benchmarks are still forming, so anyone expecting a mature performance channel today is likely to be disappointed.
• Over-personalization repels. Relevance helps, but messaging that feels too personal crosses a line and pushes users away, which matters more on an intimate interface like chat.
• Last-click will mislead you. Chat-based discovery can influence revenue long before a purchase shows up in analytics, so a last-click view will undercount the channel.
• Some categories are restricted. Health, mental health, politics, adult content, financial trading, and certain regulated industries face limits or exclusions, so confirm eligibility before you plan.

How should brands prepare before the floodgates open?

You do not need to be in the test to get ready for it. The work that pays off now is the same work that makes a brand visible in AI answers generally. Invest in answer-engine optimization so ChatGPT and other models already understand and cite your brand, because early adopters of that discipline are reporting several times more brand mentions than late movers. Put a measurement plan in place that captures assisted conversions and cross-channel influence, so you can prove value the moment you start spending. Decide which of your products map to real, high-intent questions people ask an assistant, since those are the moments a sponsored card can help rather than interrupt. When access widens, the brands that prepared will be ready to buy with a plan instead of experimenting blind.

What does this mean for your marketing mix?

ChatGPT ads are early, unevenly measured, and already pulling real budget, which is exactly the profile of a channel worth understanding before your competitors do. The reachable audience is large and arrives with intent, the conversion signals look strong, and the reporting will catch up. Treat this as a place to learn now with small, well-instrumented tests, and build the answer-engine foundation that makes both paid and organic visibility hold.