Quick reference: ChatGPT ads in 2026
- Launched: February 9, 2026 in the United States
- Where they appear: Below the response in ChatGPT, clearly labeled as sponsored, visually separated from the answer
- Who sees them: Free and Go tier users; Plus, Pro, Business, Enterprise, and Education stay ad-free, as do under-18 accounts
- Audience scale: 900M+ weekly active users; 50M paying subscribers (late February 2026)
- Pricing (April 2026): ~$25 CPM and $3–$5 CPC, ~$50K minimum spend — down from $60 CPM and $200K–$250K minimum at launch
- Reporting: Aggregate impressions and clicks; no chat logs, no user-level data
- Vs. Google: AI Overviews already serves ads above, below, and within AI-generated answers, using both query and Overview content for relevance
- Anthropic position: Claude has publicly committed to staying ad-free
On February 9, 2026, OpenAI did the thing it spent two years saying it didn't want to do: it put ads inside ChatGPT.
The pilot is small. The format is conservative. Sponsored placements appear below the answer, clearly labeled, on the Free and Go tiers only. Plus, Pro, Business, and Enterprise stay clean. By every public measure, OpenAI has gone out of its way to make the launch feel measured, principled, and reversible.
Don't let the calmness of the rollout fool you. This is the most consequential shift in digital advertising since mobile, and most agencies haven't even started thinking about it.
I don't say that lightly. The agencies that figured out paid social in 2012 spent the next decade outpacing the ones that kept treating Facebook like a banner ad network. The ones that figured out programmatic in 2015 did the same thing to the agencies still trafficking IO-based display. There's now a third inflection point on the table, and the framing — "we'll wait until it matures" — is the exact wrong instinct.
What ChatGPT Ads Actually Are: Three Shifts Converging
Three things are converging, fast.
The first is that the answer surface is now a media surface. In its January 16, 2026 announcement, OpenAI described where ads would appear: "at the bottom of answers in ChatGPT when there's a relevant sponsored product or service based on your current conversation." That is a new kind of inventory: not a search results page, not a feed, not a banner — a placement that fires after the user has already articulated their problem in natural language and received a tailored response. The targeting signal isn't a keyword. It's the entire reasoning context.
The second is that Google has already crossed this line. Per Google's own Ads help documentation, ads can serve above, below, and within AI Overviews, and "both the user query as well as the content of the AI Overview are considered when serving these ads." That's not a future capability. That's a live ad surface inside the answer most of your clients are competing to be cited in. We covered the operational implications of that shift in The Google Ads Job Changed in 2026.
The third is that trust is volatile but not absent. Gartner's 2025 consumer survey (377 U.S. consumers, fielded June–July 2025) found that 53% don't trust the reliability and impartiality of AI search summaries, and 61% wish they could turn AI summaries off entirely. But that same population is the one increasingly using AI to make purchase research decisions. The audience is skeptical, present, and forming habits at the exact moment the platforms are deciding what advertising in this format looks like.
That combination — massive new inventory, contextual targeting that goes deeper than keywords, and a skeptical audience whose trust patterns are still being set — is the entire opportunity. And the entire risk.
The "Wait and See" Position Is More Expensive Than It Looks
I've heard the argument from other agency owners over the last three months. It goes: ChatGPT ads are limited, the categories are restricted, the reporting is thin, the CPMs are inflated, and there's no scalable buying interface yet. Why bother?
Here's why.
The pricing tells a sharper version of the same story. The Feb 9 launch was CPM-only at roughly $60 per thousand impressions with a $200K–$250K minimum spend — an enterprise-tier pilot. Ten weeks later, by April, CPMs had eroded to around $25, OpenAI had added CPC bidding in the $3–$5 range, and the minimum had dropped to about $50K. The reporting advertisers receive is aggregate views and clicks — no chat logs, no user-level data, no rich attribution. By every traditional metric an agency uses to evaluate a media channel, this looks unfavorable.
But evaluating that as a wait-and-see proposition is exactly the mistake. Anyone who waited "until pricing comes down" already missed the window where it did — the channel got 5× more accessible in ten weeks while their RFP was still circulating internally.
Three years ago, programmatic display was cheap, granular, and generated mountains of data. It also generated almost nothing in terms of buyer confidence at the bottom of a high-consideration funnel. Cookie loss, iOS signal degradation, and walled-garden attribution reform have made every dollar I spend on display feel less and less like a real read on customer intent.
Conversational ads invert that trade-off. The data is thin, but the moment is qualitatively different. An ad that arrives inside a research conversation — after the user has explained their actual problem in their actual words — is closer to a referral than a banner. The right question isn't "Is the CPC higher than display?" The right question is "Does this channel produce more confident downstream behavior — deeper research engagement, faster internal forwarding, earlier technical validation — than the channels I'm comparing it to?"
Most agencies aren't built to answer that question. That's the bigger problem than the channel itself.
The Four Things Every Agency Should Be Doing Right Now
I'll be direct about what we're doing at Choice OMG and what I think the rest of the industry should be doing too.
One: stop treating AI search and AI ads as separate problems. They're the same problem. If a user asks ChatGPT about your client's category, the system pulls from public web content, structured information, brand mentions, and citation patterns to compose an answer. Then — increasingly — it serves an ad next to that answer. SEO content strategy and paid AI strategy are now operationally linked. If your SEO team is optimizing for AI Overview citations and your paid team has never opened the ChatGPT ad pilot documentation, you have a strategy gap, not a tactical one. We track citation patterns across ten AI platforms for exactly this reason — see We Track Your Business Across 10 AI Platforms.
Two: rewrite the copy rules. A ChatGPT ad is not a Google search ad. The user has already been answered. Your ad lives in the cognitive moment after the answer, when the user is deciding what to actually do. "Book a Demo" doesn't work in that context. What works is a continuation of the reasoning: state the problem you solve, show one concrete piece of proof, offer a low-friction next step that doesn't feel like a sales handoff. OpenAI's published ad policies are explicit on this: ads must be clearly distinguishable from the ChatGPT product (no interface imitation), claims must be truthful (no exaggerated outcomes or false endorsements), and creative must be consistent end-to-end with the landing page. Vague superlatives and "looks like a citation" creative are wrong both strategically and from a policy standpoint. The agencies that win this format will write copy that reads like a useful next step, not a banner squeezed into a chatbot.
Three: redesign your landing pages for proof, not capture. The user clicking from a ChatGPT ad is further along in their thinking than a user clicking a paid search ad. They don't need a hero image and a form. They need to see whether you can actually do the thing you said you do. That means architecture pages, sample reports, technical walkthroughs, comparison documents — proof objects that reward the higher cognitive investment of the click. The "continue the reasoning" landing page is the design pattern that matches this channel, and it sits squarely inside conversion optimization work, not paid-search creative production. Most landing pages built for paid search will underperform here, badly.
Four: build first-party measurement, because the platform won't give it to you. OpenAI gives advertisers aggregate views and clicks. That's it. If you want to know whether ChatGPT ads are producing higher-quality downstream behavior than display, you have to measure it yourself. That means matched-cell experiments — same offer, same audience definition, different traffic source — and a scorecard that captures things like proof-asset engagement, multi-stakeholder visit patterns, demo show rates, and time-to-technical-review. Click-through rate is the wrong KPI. Confidence-building behavior is the right one. If your agency can't run that experiment cleanly, you don't have a measurement function — you have a reporting function.
The Honest Part Most Agencies Won't Say
Anthropic — the company behind Claude, the AI assistant our agency uses internally for production work — has publicly committed to keeping Claude ad-free, including a Super Bowl spot on February 4, 2026 with the tagline "Ads are coming to AI. But not to Claude." Their argument is that even clearly separated conversational ads create incentive misalignment in a product whose value depends on independent reasoning. That's a real philosophical position, and it's worth taking seriously rather than dismissing as marketing differentiation.
The honest read is that both positions can be partially right. OpenAI's case — that ads enable broader access to AI for users who won't or can't pay — is defensible. Anthropic's case — that mixing ads and conversational reasoning is a long-term trust hazard — is also defensible. The market is going to test both theories simultaneously, and the answer probably won't be clean.
But that uncertainty is exactly the wrong reason for an agency to do nothing. The reason to build capability now is that we don't know which version of this becomes the default — and the agencies that have run real experiments, written real copy for the format, and built real measurement frameworks will be the ones whose clients win regardless of which platform model dominates.
The agencies that wait will be paying senior strategists to do basic catch-up work in 2027 that should have been done in 2026.
Where ChatGPT Ads and the Wider AI Advertising Shift Are Heading
The framing I keep coming back to: in 2010, Google and Facebook were the only platforms most agencies had to think seriously about. That was a luxury. The next five years are going to fragment the answer surface across ChatGPT, Google AI Overviews, Perplexity, Claude, and whatever comes next — and each of those platforms will resolve the trust-vs-monetization tension differently. Some will go ad-heavy and lose user confidence. Some will stay clean and become premium media. Some will try a middle path and fail. We don't know which is which yet.
What we do know is that the agencies pretending this isn't happening will, in eighteen months, be having uncomfortable conversations with clients who are watching competitors show up in conversational answers they don't appear in.
I'd rather be the agency that learned this format while it was still shaped by user feedback than the one that learned it after it had hardened around someone else's playbook.
The age of AI answers is here. The age of AI advertising arrived in February 2026. The agencies that conflate the two are going to be expensive to catch up to, and the agencies that ignore both are going to be replaced.
Decide which you want to be.
Frequently Asked Questions
When did ChatGPT ads launch?
OpenAI introduced sponsored placements inside ChatGPT on February 9, 2026, in the United States. Ads appear below the response, are clearly labeled as sponsored, and are visually separated from ChatGPT's answer.
Which ChatGPT subscription tiers see ads?
Ads appear only on the Free and Go tiers. Plus, Pro, Business, Enterprise, and Education subscriptions stay ad-free. OpenAI also blocks ads for accounts where users have identified as under 18 or where the system predicts an underage user.
How much do ChatGPT ads cost in 2026?
Pricing has moved fast. The February launch was CPM-only at roughly $60 per thousand impressions with a $200,000–$250,000 minimum advertiser commitment — effectively an enterprise-tier pilot. By April 2026, CPMs had dropped to around $25, OpenAI had introduced CPC bidding in the $3–$5 range, and the minimum spend had fallen to about $50,000.
Can my business advertise on ChatGPT?
If your category is permitted under OpenAI's ad policies and you can meet the current minimum spend, yes. The lowered $50K threshold opens the format to mid-market advertisers, not just the launch-partner tier (Target, Adobe, Williams-Sonoma, Albertsons). Creative must comply with OpenAI's published ad policies on sponsored disclosure, accuracy, interface separation, and landing-page consistency.
Are AI Overviews ads the same as ChatGPT ads?
No. Google AI Overviews ads have been live longer and serve above, below, or within Google's AI-generated overview answers, using both the user query and the AI Overview content for relevance. ChatGPT ads are a separate inventory inside OpenAI's product, with different creative requirements, a different bid model, and different reporting. Both are conversational placements, but the buying interfaces and audiences are distinct.
Will Claude (Anthropic) ever have ads?
Anthropic has publicly committed to keeping Claude ad-free, including a February 4, 2026 Super Bowl spot with the tagline "Ads are coming to AI. But not to Claude." Anthropic's stated rationale is that ad-supported business models create incentive misalignment in a product whose value depends on independent reasoning. That position could change, but it has been stated explicitly and recently.