Here's the thing that changed, and most people missed it: ChatGPT is now an ad platform you can buy into yourself, with no minimum spend.
The timeline is quick. OpenAI started testing ads inside ChatGPT in February, for free and lower-tier users in the US. At first, buying one was a walled garden — a managed-service deal with a reported $200,000 minimum and a flat $60 CPM, the kind of thing only a big brand or its agency could touch. Then the floor dropped to $50,000. Then, on May 5, OpenAI opened a self-serve Ads Manager to every US advertiser and removed the minimum entirely. It added cost-per-click bidding, a measurement pixel, and a conversions API — the same plumbing you already know from Meta and Google.
To put the demand in perspective: the ad business reportedly pulled in about $100 million in its first six weeks. This is not a tentative experiment. It's a new channel being built at speed.
What the ad actually is
This matters, because it's easy to picture something worse than what's shipping. The ads appear as clearly labeled, lightly tinted boxes at the bottom of a response — not woven into the answer itself. OpenAI's stated line is that ads don't influence what the model tells you. Placement is contextual: the system matches an ad to what you're currently talking about, plus your past chats and earlier ad interactions, rather than to a keyword you typed into a search box.
So if someone asks ChatGPT to help them pick a standing desk, a desk brand can appear underneath the answer. The model still gives its honest take above. The ad is a separate, marked unit below it. That's the deal as it stands today.
For fifteen years there were two places to buy intent: Google and Meta. Quietly, in one quarter, a third one opened — and the line to get in is now a credit card.
Should a small business touch it yet?
Honest answer: probably not as your first move, and here's the math. The reported costs run roughly $25–$60 per thousand impressions, with cost-per-click bid floors around $3–$5. A $3–$5 click floor is not cheap — it's in the range of competitive Google Search clicks, on a platform where nobody has years of data telling them what converts. Early reports also point to low click-through rates, which makes sense: people in ChatGPT are mid-task, not shopping.
That combination — real click prices, thin performance history, distracted users — means this is a channel to learn, not to bet the month's budget on. If you have a working funnel on Meta or Google and spare test money, a small, tightly-themed campaign is a reasonable experiment. If you're still figuring out who your customer is, your money does more on the channels you can already measure.
The move that costs nothing is to go run the searches your customers would run. Open ChatGPT, ask the questions a buyer in your category asks, and watch what comes back — both the answer and the ad underneath it. You'll learn two things at once: whether your category already has advertisers fighting for that box, and whether the model recommends businesses like yours when nobody's paying. That second one is the bigger story.
The lesson under the headline
The ad box is the visible part. The quieter shift is that more buying decisions now start as a question typed into an AI, and the AI's unpaid answer carries more weight than the sponsored box beneath it. Paying to sit under the answer is one lever. Being the kind of business the model names in the answer — because your site is clear, your reviews are real, your category fit is obvious — is the one that compounds.
Watch the ads. But spend more energy being the recommendation than buying the slot beside it.