For the first time, your brand can show up inside ChatGPT itself, in a slot right below the answer. Here’s who sees these ads, how to buy them, and how to treat them as an opportunity rather than a threat.
ChatGPT held out on advertising longer than most free products do. For years the answer came back clean, with nothing sponsored anywhere near it. That changed in early 2026, when OpenAI set out its approach to ads and switched them on for US users on February 9.
What began as a cautious pilot has moved fast since. By May, OpenAI had opened self-serve buying to US businesses of any size and announced expansion into the UK, Japan, South Korea, Brazil, and Mexico, on top of earlier pilots in Canada, Australia, and New Zealand. By July it had launched a full “Advertise in ChatGPT” ads manager, with Best Buy, Lowe’s, and VistaPrint among its first case studies.
For a brand, the news isn’t really that ads arrived. It’s that a new place to reach customers just opened, one that sits inside the tool a growing number of them now ask instead of a search engine. So the interesting questions are practical ones. Who sees these ads? How do you buy them, and what do they cost? Does the format actually work? And how do you make the most of it before everyone else does? Let’s take them in order.
Who sees ChatGPT ads?
Before we get into the nitty-gritty, it helps to be clear on who sees ads, what they see, and where. That’s what tells you where these ads make sense for your brand and where they’d just be throwing money at OpenAI.
Ads appear for adults on the Free and Go tiers, so anyone who pays for a subscription, or gets one through work or school on Plus, Pro, Business, Enterprise, or Edu, is exempt. The help documentation lays this out plainly, and free users can even opt out of ads in exchange for fewer daily messages. Geographically the pilot now covers nine markets, with the US the most built out. There, ads can reach logged-out users as well as logged-in ones, while other markets remain logged-in only for now.
The ad itself appears below a given answer, in a clearly visible “sponsored” slot set apart from ChatGPT’s response. That way the reader always knows what came from ChatGPT and what came from an advertiser. In practice today it usually looks like a shopping result, a product with a brand logo, a price, stock, and a delivery estimate, or a card you can tap to ask ChatGPT more about the item. If you’ve used the sponsored row on a retailer’s site, it will feel familiar.
One more thing worth knowing up front. ChatGPT won’t show ads near sensitive or regulated topics, including personal health, mental health, or politics. If your category lives there, the surface is limited by design, so it’s worth checking before you plan around it.
How do you buy them, and what do they cost?
This is where the story has changed most since launch. The pilot opened through the big agency groups, Omnicom, WPP, and Dentsu, with reported minimums near $200,000 and pricing on impressions. That gate is gone. Since self-serve buying opened in May, any US business can build a campaign, set a budget, and bid per click, and the July launch of “Advertise in ChatGPT” turned the pilot into a proper ads platform. If you want in, the entry point is OpenAI’s advertiser page, not a holding-company pitch meeting.
Targeting works on conversational context rather than keywords. A ChatGPT prompt often carries more information than a search query, “a standing desk for a small office on a tight budget” rather than “standing desk”, and the ad system matches against that context. As we’ll see below, what it never gets is the person behind the prompt.
But does it work?
This is the question every marketer should be asking, and the honest answer is that it’s still early, though no longer data-free.
Start with the case that it works. The placement has two things going for it that advertisers usually pay a premium for. It’s high-intent, arriving the moment someone asks about a product or a problem, and it’s high-trust, sitting right under an answer the reader asked for and tends to believe. That’s the same logic that made search ads one of the most effective formats ever built, matching a paid message to what someone just said they wanted. Adoption backs the interest up: the agency pilots filled, brands like Target, Ford, and Audible ran early placements, and OpenAI has since gone from gated test to open platform in five months, which is not the trajectory of a format its owner doubts.
Now the honest numbers. Early reporting puts click-through rates around 0.91 percent, well below Google Search benchmarks. That’s not damning for a brand-new format with unfamiliar placements, but it does mean nobody should walk in expecting search-ad economics on day one. Measurement is still maturing, benchmarks are thin, and formats are visibly still being tuned.
Put those together and the picture is neither hype nor a trap. The format has real reasons to perform, especially for products people research before they buy, and the barriers to a small test have mostly fallen. The brands testing now are buying a cheap education in a surface their competitors haven’t touched, and that head start is worth something even if the first campaigns only break even.
The line between ads and answers
The thing that makes this surface safe to stand on is a wall OpenAI built between the ads and the answers, and it’s worth understanding because it works in a brand’s favor.
The commitment, stated plainly, is that “ads do not influence the answers ChatGPT gives you.” The help center puts the same thing in engineering terms: ads run on separate systems from the chat model, and advertisers have no ability to shape, rank, or alter ChatGPT’s responses. On the data side, advertisers get aggregated, non-identifying information about views and clicks, and never reach your customers’ chats, memories, names, emails, precise location, or IP addresses. Personalization draws on the conversation and, if the user allows it, their history, but that stays inside OpenAI. As an advertiser you buy exposure to a context, not a look at the person.
That matters for two reasons. Your customers’ private conversations stay private, which keeps a brand well clear of an awkward headline. And because the answer itself can’t be bought, showing up in the sponsored slot doesn’t make you look like you gamed the system. The ad is clearly an ad, and the answer above it is clearly independent.
Not everyone welcomed ads landing this close to a private conversation, and OpenAI took some criticism when it launched. These guardrails are its answer, and they’re the reason the slot is a reasonable place for a brand to be. One honest caveat belongs here, though. The wall is a commitment, not a law of physics, and OpenAI is now projecting billions in ad revenue against it. It has held so far. A sharp brand keeps watching whether it holds as the targets grow, which is one more reason to build the part of your visibility that no policy change can touch.
Show up in both places
Here’s the useful way to think about visibility in ChatGPT. There are two ways to appear, and they don’t compete. One you can pay for, one you earn, and the smart move is to work on both.
The one you pay for is the sponsored slot we’ve been describing. The one you earn is the recommendation inside the answer itself. When ChatGPT names a product, a tool, or a supplier in the body of its reply, that isn’t a media placement and no budget moves it. It comes from what the model has learned about your category from sources it trusts. Getting your product into that answer is the new version of search visibility, and the field already has a clumsy name for it, answer engine optimization, or AEO. Under the jargon it’s an old question. Does the model have clean, current, quotable information about what you sell, from places it trusts, or is it working off a stale third-party page?
The reason to do both is simple. The ad puts you in front of a buyer today. The organic work makes the model reach for you tomorrow, and it keeps paying off after any single campaign ends. Brands that treat them as either-or leave half the surface on the table.
What to keep an eye on
None of this is finished, and a sharp brand watches a few things as the format matures rather than assuming today’s version is the final one.
Performance is the big open question: early click-through rates trail search, so treat first campaigns as buying a read on the format as much as buying results, and push for whatever view, click, and conversion data the platform gives you. Formats and rules are still shifting quarter to quarter, so what works now may look different by winter. And market coverage, while now nine countries deep, is still uneven, with the US far ahead of everywhere else on capabilities. None of these is a reason to stay out. They’re reasons to start small, measure honestly, and scale as the picture fills in.
Rent the box, earn the answer
There are two ways to show up in ChatGPT now. You can rent the box below the answer, and you can earn the recommendation inside it, which no budget will ever buy. Ads arriving in ChatGPT rewards being early and thoughtful rather than panicked or absent: test the slot while it’s cheap to learn, keep your privacy and brand-safety position clear, and put in the organic work so the model recommends you on its own. Do both, and you show up whether the customer is reading the answer or the ad beneath it.
That’s close to the work we do, helping brands get their products, content, and data into the shape an AI will surface, and thinking through where paid placement fits. If you’re working out what visibility in ChatGPT looks like for your category, Hire Digital can help you move before your competitors do.