When it comes to ChatGPT ads vs organic AI visibility, the short answer is that an ad gets you a labeled sponsored spot next to ChatGPT's answer, and nothing in the answer itself. OpenAI says in plain words that ads don't influence the responses, so the brands ChatGPT names, ranks and recommends are still decided the same way they were before anybody could pay (OpenAI, Our approach to advertising, Jan. 16, 2026).
I'm writing about this now because OpenAI announced on October 21 that ChatGPT Ads are expanding to 26 more countries, 14 of them in Latin America, and that more than 100,000 advertisers have run ads on the platform so far (AdExchanger, Oct. 21, 2026). That puts the total at 89 markets (PPC Land). If you run a brand, somebody on your team or at your agency is going to bring this up in a budget meeting soon if they haven't already, and I'd like you to walk into that meeting knowing what the money does and what it doesn't.
The short version
ChatGPT ads show up below or beside the answer, labeled as sponsored. By OpenAI's own rules they have no effect on which brands the answer recommends.
Early click rates are low in the US (0.73% all-time, per Similarweb panel data), and advertisers are reporting clicks their analytics can't find.
Ads appear more often in conversations that already name a brand, which means the organic answer is still doing the persuading next to your ad.
Fix what ChatGPT says about you first. Then test ads, with a measurement plan, if your category is crowded with them.
What OpenAI says ads can and can't do
OpenAI wrote down five ad principles back in January, and the one that applies here is called answer independence. The wording is "Ads do not influence the answers ChatGPT gives you," and the same page says responses should be "driven by what's objectively useful, never by advertising" (OpenAI).
The help page on testing ads fills in the mechanics. Ads are matched to "the topic of your conversation," along with your past chats and how you've interacted with ads before. They only run for logged-in adults on the Free and Go plans, so Plus, Pro, Business, Enterprise and Education users never see them, and they aren't eligible to appear near health, mental health or politics (OpenAI, Testing ads in ChatGPT, updated May 7, 2026).
So think about what that means for a brand. Your ad can show up when somebody is talking about your category. The paragraph above your ad, where ChatGPT lists the brands it thinks fit, is written without any knowledge that you paid. If you're not on that list, your ad is sitting under a recommendation for your competitors.
The look of the ads is changing, and that part I'd keep an eye on. AdExchanger reported that the ads started out in a shaded box with "sponsored" printed prominently at the top, and the newest format has no box and a smaller label off to the side. OpenAI still says people "shouldn't have to guess whether something is an ad" (AdExchanger). I take them at their word on the answer itself. The visual separation is getting thinner, though, which matters if you care about how your brand looks sitting next to it.
One ChatGPT reply, two different jobs
Tap each part to see who controls it and who sees it.
Sources: OpenAI, Our approach to advertising (Jan. 2026); OpenAI, Testing ads in ChatGPT (updated May 2026). Illustration only; brand names are placeholders.
What the early ChatGPT ad numbers say
The main public dataset right now is "The State of ChatGPT Ads," which Similarweb and Graphite put out at the end of September. A few caveats up front, because they matter. Similarweb sells the ChatGPT ad tracking tool that produced the panel data, the report gives percentages without raw counts, and the markets have anywhere from about five to 25 weeks of data each (Search Engine Journal, Sept. 29, 2026).
The headline numbers, via PPC Land's write-up (PPC Land, Oct. 4, 2026), were these.
In September, 43.4% of ad-eligible US users saw at least one ChatGPT ad.
The all-time click rate in the US was 0.73%. Canada and the UK were about the same. Brazil, Mexico, Japan and South Korea were all above 1%, though Brazil and Mexico only had about a month of data.
Worldwide click rate rose 112% between late March and mid-September.
61% of exchanges that carried an ad named a brand, compared with 36% of exchanges without an ad. The report labels that finding early and directional.
That last one is the number I'd look at twice. It suggests ads are landing in the conversations where people are already comparing brands by name, which makes sense given how the matching works. It also means your ad is most likely to appear right next to a paragraph where ChatGPT has already said which brands it thinks are a fit.
The measurement side is rough. Tech Times collected reports from agencies running small tests, including a Canadian agency that spent $415, got 60 clicks at around $7 each and saw none of them convert, and a B2B agency whose Ads Manager reported 57 clicks while Google Analytics showed fewer than 20 visits (Tech Times, Sept. 1, 2026). Those are anecdotes from small budgets, and there are counterexamples in the same piece, including OpenAI's own case studies. OpenAI's advertiser materials also say ChatGPT Ads don't have performance benchmarks yet, so nobody can tell you what good looks like. If you already struggle to see ChatGPT traffic in your analytics (most of it shows up as direct, which I wrote about in how to track ChatGPT traffic), paid clicks don't make that easier.
ChatGPT ads: lots of reach, few clicks
Brazil and Mexico went live Aug. 17, so their click rates rest on about a month of data. The US has about five and a half months.
Source: Similarweb and Graphite, "The State of ChatGPT Ads" (Sept. 2026), via PPC Land and Search Engine Journal. Similarweb panel data; percentages only, no raw counts.
Why the organic answer still does the heavy lifting
This is where my own research comes in. In the AI Verdict Study, I ran 432 search-on conversations across four engines, each ending with a buyer asking "should I choose it?" about a specific brand. When the engine named the brand in the first turn, the brand got rejected in 18% of conversations. When it didn't, rejection jumped to 44%. That's a correlation, and the study has real limits (one buyer per category, API calls rather than the consumer apps), so I wouldn't stretch it too far. Still, it lines up with the ad data. The thing that happens early in the conversation, before any ad has a chance to show, is the organic answer naming names.
ChatGPT builds that answer from what it finds about you. It runs searches, reads pages, pulls from review sites and comparison articles and forums, and the brands that show up consistently across those sources are the ones that get named (where AI engines go to decide about your brand). None of those sources care whether you bought an ad. Brands also get named without getting cited, which is its own problem and worth understanding before you start counting impressions (AI brand mentions vs citations).
So if you're weighing the two, I'd describe it this way. The organic work gets you onto the shortlist ChatGPT writes, and an ad gives you a second shot at somebody who's reading that shortlist whether you made it or not. Both can be worth money, they do very different jobs, and only the organic side reaches people on Plus, Pro, Business and Enterprise plans, since they never see ads at all.
Who should test ChatGPT ads right now
I'm not anti-ad here. If I were running a B2B SaaS brand I'd probably be testing, mostly because everyone around me is. B2B SaaS took 19.9% of US ChatGPT ad impressions between late March and early August, the biggest share of any category, and Monday.com, Jotform, HubSpot and Cursor all show up in the top 10 advertisers by worldwide impression share (PPC Land). If your competitors are in that sponsored spot and you're not in the answer either, you're invisible twice.
The Graphite team also found that 98% of ChatGPT ad copy was net new, with fewer than 2% of headlines matching the advertiser's Google search ads. People aren't recycling their paid search copy, and I'd follow their lead on that. A chat ad shows up in the middle of a conversation about a problem, and it should read like it.
Where I'd hold off is if you don't know what ChatGPT says about you today. If you can't answer that, any ad test is going to be hard to read, because you won't know whether a weak result came from the ad or from the paragraph above it.
What I'd do before spending on ChatGPT ads
Number one, find out what ChatGPT says about you. Run the 10 or 15 questions your buyers ask in your category, the "best X for Y" and "should I choose Z" questions, in ChatGPT with search on, and write down who gets named. The free Quick Check on my site walks you through a short version of this.
Two, fix the gaps that the answer is built on. If ChatGPT keeps naming three competitors, look at where it's getting that from. Usually it's comparison pages, review sites and a few forum threads, and those are things you can work on. I go into the bigger picture in the click that didn't happen, which is the post I point people to when they ask why any of this matters.
Three, if you test ads, set the measurement up first. Use OpenAI's pixel or Conversions API, tag everything, and compare Ads Manager clicks against what your analytics shows from day one so you're not surprised later. Decide ahead of time what you'd count as a result, because nobody has published a benchmark you can borrow.
Then last, keep the budgets separate in your head even if they come from the same line. You're renting the ad spot for as long as you keep paying, while the organic work changes what the answer says for every user on every plan, including all the ones who never see an ad. I spend a good chunk of Explainable on how those answers get assembled, because that's the part most brands can influence and the part they usually skip.
I'd love to hear from anybody who's run ChatGPT ads, especially if your numbers look different from what's in the Similarweb data. Send me a note through the site or find me on LinkedIn, and I'd be happy to compare notes.
Are you ready to spend on ChatGPT ads?
Six questions. Answer for where you are today and you'll get a score and the fixes I'd make first.
Answer all six to see your score.
Jarred Smith is the author of Explainable: Why AI Recommends Some Brands & Ignores Others, an Amazon bestseller on AEO, GEO, and SEO. He's a marketing leader with nearly 20 years of experience across healthcare, public media, retail, and environmental services. Find him at jarredsmith.com.