TL;DR

Ads inside ChatGPT have been live in the UK since June, and this week OpenAI switches them on across 31 more European markets. They sit below the answer, they're labelled "sponsored," and only people on the free and low-cost Go tiers see them. The tools to buy one are already open to UK businesses. The temptation for a small business is to read this as "a new place to buy ads" and go shopping. I think that's the wrong first move. The bigger shift is that the tool a growing share of your customers now ask for a recommendation has just become a paid surface. The businesses that win there won't be the ones who bid first. They'll be the ones an AI can already understand well enough to recommend on merit. Here's the order I'd do things in.

For about a year now, a quiet change has been happening in how people find things to buy. Instead of typing a phrase into a search box and scanning ten blue links, a good chunk of them open ChatGPT and ask a question in full sentences: "who's a reliable independent jeweller near me for a resize," "which local firm can help a small company with its year-end," "what's a good hair salon in town that does curly hair." The assistant reads the web, weighs it up, and hands back a short answer with a couple of names in it. No browser. No scrolling. Just an answer.

That answer already has a neighbour. OpenAI launched ChatGPT Ads in the US back in February, brought them to the UK in June, and this week switches them on across 31 more European markets. So for a UK business, this is already live. Your customers can see these ads today, and the self-serve tools to buy one have been open here since the summer. The mechanics are worth knowing precisely, because the detail is where the useful thinking lives.

How the ads work

Three facts matter. First, the ads appear below ChatGPT's answer, they're marked "sponsored," and they're visually fenced off from the reply itself. This is not the model being quietly bribed to name a paying business inside its recommendation. The organic answer and the advert are separate things on the page. Second, only people on the free and low-cost "Go" tiers see ads at all; anyone paying for Plus, Pro or Enterprise gets a clean, ad-free view. Third, the ads are built around the conversation rather than a search keyword. The system reads what the person is trying to do and places against that intent.

For advertisers, buying is already straightforward. You can go through OpenAI's own ads team, an agency partner, or the self-serve Ads Manager that opened to UK businesses over the summer. Campaigns can run on the usual levers (pay per thousand views, pay per click, or click campaigns optimised for a conversion) with country targeting and custom audiences. In other words, it already looks and feels a lot like running ads on any other big platform. That familiarity is exactly what makes me want to slow business owners down.

Why "buy the ad" is the wrong reflex

Every time a large platform opens an ad channel, the same thing happens: a wave of small businesses pile in early, bid against each other for attention they don't yet understand, and spend three months paying to learn what they could have worked out for free. I watched it happen with search ads, then with social, and the pattern never changes. Early ad inventory on a new surface is a laboratory that charges you rent.

There's a deeper reason to wait, though, and it's specific to this surface. On a conversational assistant, the ad sits next to an organic recommendation the person arguably trusts more, because it came from the assistant's own reasoning rather than from someone's wallet. If your business isn't the sort of thing ChatGPT already understands and mentions on merit, an ad won't rescue you; it'll just put your name beside a recommendation that isn't you. The unglamorous work of being legible to the machine pays off in both columns at once. The ad-buying work only pays off in one, and only after the first has been done.

The test: open ChatGPT today and ask it, as a customer would, to recommend a business like yours in your area. Ask three or four ways. If your name doesn't come up, or comes up with the wrong details, that's your real problem this week, and no amount of ad spend fixes it. It just makes it more expensive.

The order I'd work in

If a client came to me on Monday worried about this, I wouldn't open an ads account. I'd do three cheaper things first, in this order.

  1. Find out what the assistant already says about you. Ask ChatGPT (and Gemini, and whatever else your customers use) to recommend a business like yours, in your words, several times. Write down what it returns. This is your baseline, and most owners have never once looked at it.
  2. Fix the facts it's getting wrong. Usually the gaps are dull and correctable: an out-of-date address, no clear description of what you do, reviews and details scattered in places a machine can't tie together. The assistant is only as good as the public trail you've left. Tidy the trail.
  3. Only then consider paying. Once the assistant can describe you correctly and recommend you on merit, an ad becomes a way to amplify something that already works, instead of a plaster over something that doesn't. That's when the spend is defensible.

A retail client of mine, an independent shop that lives and dies on local custom, did exactly the first step with me a while back, before any of this ad news. We simply asked the big assistants to recommend "a shop like theirs" nearby. The result was quietly alarming: the assistant knew the shop existed but described it using details that were two years stale. We spent an afternoon fixing the public facts. Nothing clever, nothing paid. The next month the recommendations came back accurate. If ads had existed then, buying one on top of the stale version would have been money spent pointing people at the wrong picture.

What this means if you sell to other businesses

If your customers are companies rather than consumers, don't tune out, but read the tiering carefully. Ads only show to free and Go users; paid tiers stay clean. A lot of professional buyers doing serious research are on paid plans, which means the person evaluating you properly may never see an ad at all. For B2B, the organic recommendation matters even more than the ad, because your best prospects are precisely the people the ads can't reach. That makes the "is the assistant describing me correctly" question the whole game.

The one-line version

Ads arriving inside ChatGPT is a real event, and over the next year it will change how attention gets bought online. But the move it should trigger for a small business this week is simple. Find out whether the assistant your customers already ask can describe you correctly, and fix it if it can't. Do that, and you're in a strong position whether or not you ever buy a single sponsored slot. Skip it, and the ad just buys you a louder version of a wrong first impression.