On 21 October OpenAI said ads in ChatGPT were coming to 26 more countries, 89 in all, and that more than 100,000 advertisers had used them since the first test "just over eight months ago" (OpenAI, 21 October 2026). CNBC reported the same day that the ads business had passed a $1 billion annualised run rate in August (CNBC). So clients have started asking me the obvious question. Should we be in there too? Fair question.
The short answer
Probably not yet as your first ad channel, and quite possibly yes as a small, separate test. If people already search Google for what you sell, that's still where I'd put the first euro, because you choose the searches and you can see exactly which ones you paid for. ChatGPT ads make sense once your tracking works and your landing page converts, and when you can spend a few hundred euros a month on something you might switch off after six weeks.
That isn't a verdict on the platform. It's a new channel with a short history, and right now the businesses that learn the most from it are the ones that already know what a customer is worth to them.
What the ads look like and who sees them
An ad appears below ChatGPT's response, not inside it. It shows the advertiser's name and site icon, a title, a line of copy, an image and a link to your page (Ads in ChatGPT: the basics). OpenAI says ads are labelled, kept apart from the answer, and that advertisers can't influence what ChatGPT says.
Who sees them is narrower than the headline numbers suggest. OpenAI says more than 1.2 billion people use ChatGPT each week, but ads only go to logged-in adults on the Free and Go plans (Testing ads in ChatGPT). Plus, Pro, Business and Enterprise users don't see any, and neither does anyone under 18. OpenAI doesn't publish how many of those weekly users are on the free tiers, so treat any reach figure you're quoted with some suspicion.
What decides whether your ad turns up? OpenAI lists the conversation itself, your landing page and ad copy, and optional "context hints" you add to each ad group. Those hints aren't keywords, though. You don't bid on an exact phrase the way you would on Google, and that's the biggest difference for anyone used to search ads.
Which countries, and can you buy them yourself?
The US test started on 9 February. The UK, Mexico, Brazil, Japan and South Korea followed on 11 August, then 31 European markets later that month (OpenAI, 18 August 2026), and the October expansion brought the total to 89.
Seeing ads and buying them aren't the same thing, mind. To use the self-service Ads Manager, the legal entity paying for the ads has to be based in a country on OpenAI's list. When I checked it on 2 November, Bulgaria, the UK and the US were all marked "Available", along with most of the EU (Ads Manager availability). The list keeps changing, so check it on the day you sign up.
One catch if you sell abroad: OpenAI says some new self-serve accounts can at first advertise only in their home country, and open up other countries after identity checks and a certain amount of spend at home (Creating campaigns). If you're based in one country and sell mainly into another, plan for that.
What does a first test cost?
The minimum daily budget is €15 or £15, and $25 for accounts billed in dollars. A daily budget is an average over a Sunday-to-Saturday week, so a single day can spend up to twice that, but the week won't go over seven times it. At the minimum, that's roughly €450 a month.
You pick how you pay: per thousand views, per click, or "conversions", which still bills per click or per view while the system aims for your chosen action. You're never charged per conversion. OpenAI's own guide suggests a starting maximum bid of $3 to $5 a click and describes a relevance-weighted second-price auction. For comparison, I went through what a Google click costs in the US, UK and Bulgaria in my piece on Google Ads pricing. The short version is that a ChatGPT click isn't obviously cheaper.
Reporting covers impressions, clicks, spend, click-through rate, cost per click, cost per thousand views and conversions, and the UTM tags you add to your links carry through to your own analytics, so you can line the results up next to Google and Meta in one report. OpenAI warns that spend data can lag by seven or eight hours, so don't panic over a quiet morning.
How it compares with Google and Meta for a first test
Google Search answers a question someone has typed. You choose the searches, see a report of what you paid for, and exclude the junk. For a plumber, a dentist or an accountant, that's still the most direct line to a customer who's ready to buy.
Meta is better at showing something to people who weren't looking for it yet: a new product, an offer, a nice photo of your work. You're buying attention, and the creative does most of the work.
ChatGPT sits somewhere in between. Somebody is in the middle of thinking through a problem, maybe comparing three options or asking what a fair price is, and your ad appears under the answer. That's a promising moment for things people research before they buy: software, courses, travel, a kitchen renovation, an accounting package, a B2B service with a long sales cycle. It's a weaker moment for "I need an electrician in the next hour", where Google and Maps still win. I'd also expect the clicks to behave differently. Someone who has just read a long answer arrives with more context than a searcher does, so they may need fewer convincing paragraphs and more specifics: price, timing, what's included. Whether that turns into more enquiries per click is exactly what a test should tell you, because nobody outside OpenAI has published reliable numbers on it yet.
There's also less control. You can target countries, regions, cities and postcodes where available, and pick devices, but you can't pick exact phrases. Custom audiences need at least 25,000 matched people, and OpenAI says to avoid them altogether for campaigns in the EEA and Switzerland while ad personalisation isn't available there. For most European businesses that means context and location are the main levers you get.
Check the automatic edits setting
From 30 October, OpenAI switched on automatic creative edits by default for new and existing campaigns (OpenAI Help Center). The system can rewrite your headline and description, translate them, resize your images and change the destination URL, and you won't get a preview of every version. You can turn it off under Manage settings, and a campaign setting overrides the account default.
Meta does something very similar, so this may ring a bell. I wrote about which of Meta's AI ad settings to check, and the same rule applies here, because a machine that rewrites your headline to get more clicks has no idea that your price excludes VAT or that you can't legally promise a result. If you're in a regulated trade, quote prices, or promise anything specific, switch it off and write the copy yourself.
What to have ready first: a landing page worth the click
The ad is only the doorway. Someone who clicks has just been reading a detailed answer, so a thin page with a slogan and a contact form will feel like a step backwards. Before I'd spend a cent here, I'd want three things in place.
First, a page about exactly the thing you're advertising, with prices or at least a price range, who it's for, and what happens after they get in touch. Second, conversion tracking that you've seen working with a test enquiry, either through the OpenAI Pixel and Conversions API that OpenAI added in August or simply through UTM tags read in the analytics you already use, because a campaign judged on clicks alone will always look better or worse than it really is. Third, a number: what an enquiry is worth to you, so you can tell after a month whether €450 bought anything. Most people skip this one.
It's also worth separating the ads from the answers in your head. Paying for an ad doesn't make ChatGPT recommend you in its responses. That part comes from your site, your Google Business Profile and what other sites say about you, which is what I covered in how to get your business recommended by ChatGPT.
When to wait
I'd hold off if any of these are true:
- You aren't yet showing up on Google Search for your own service and town. Fix that first, it's cheaper.
- You have no conversion tracking, so you'd be judging the test by clicks.
- Your whole monthly ad budget is around €450, which is the ChatGPT minimum on its own.
- What you sell is urgent and local, like a locksmith or a taxi, where nobody stops to chat it through first.
- You'd have to rely on personal targeting in the EEA, which isn't on offer there yet.
And if none of those apply? Run one campaign, in one country, on the Clicks objective, for four to six weeks, with automatic edits off and a landing page built for that ad. Compare the cost per enquiry with your Google and Meta numbers, then decide. That's what we'd set up for a client, as part of our ad campaign management.
If you'd like a second opinion on whether your business is a fit, tell us what you sell or email contact@magnuszgr.com, and I'll tell you honestly whether I'd test it now or wait.

