Where ChatGPT ads have got to
ChatGPT ads went live in the UK in June, after a pilot in the US. The ads appear underneath ChatGPT's answers, clearly labelled as sponsored, and only to people on the free and low-cost Go plans. Paying Plus, Pro and business users don't see them, and neither do accounts identified as under 18.
OpenAI says the ads don't influence what ChatGPT says, and advertisers don't get to read anyone's conversations. You don't buy keywords either. You give the system "context hints" about your business, and it decides which conversations your ad suits. You can buy through OpenAI's self-serve Ads Manager, which is still labelled beta, and pay by the click or by a thousand views.
The money is clearly there. Within six weeks of the US pilot, OpenAI said ChatGPT ads had passed $100 million in annualised revenue. Every agency I know has been asked about it by at least one client. So, are they any good?
Where ChatGPT ads sit in the funnel
My view is that ChatGPT ads aren't competing with Google Search ads yet. They sit further up the funnel.
Think about how people use each one. On Google, someone types "solar installer in Bristol" or "accountants near me" because they're ready to act. Search ads work because they catch people at the moment they're looking to buy, call or book.
People use ChatGPT earlier, while they're still working out the problem. They ask questions about installing solar power, or what to look for in an accountant, or which CRM suits a ten-person sales team. OpenAI's own study of how people use ChatGPT found that seeking information had become about a quarter of all messages, and was still growing. That's research and discovery. People are finding out which brands, products and services might solve their problem, often before they've decided to buy anything.
That makes ChatGPT a useful place to be seen. Your brand gets in front of someone while they're still forming a shortlist. But it's closer to a good paid social placement than to a Google Search ad, and it should be judged that way.
Don't expect Google levels of traffic or conversions
The early numbers back this up. Search Engine Journal's review of the first six months collected results from advertisers and agencies. Click-through rates sat around 0.7% to 1%, well below what a decent Search campaign gets. Costs per click ranged from about £2 to £22 depending on the advertiser and the market, and the UK was at the cheaper end.
The conversion stories are mixed. One high-value ecommerce brand saw a healthy return, but the figure moved by a factor of two depending on which attribution model they used. One B2B advertiser traced 336 paid clicks back to 146 organisations, and found only five of them matched the customers they wanted. None of that was visible in Ads Manager.
None of this means the clicks are worthless. It means someone researching in ChatGPT usually isn't ready to buy that day. If you judge ChatGPT ads on the same last-click cost per lead as your Search campaigns, they'll look poor almost every time.
Measurement is the biggest problem
This is where the platform feels most like a beta. Ads Manager reports impressions, clicks, spend, click-through rate, cost per click, cost per thousand impressions and conversions. That's about it.
There's no search terms report, because there are no search terms. You can't see the questions that led to your ad. There's no auction insights report, so you can't see who you're competing with or why your costs moved. OpenAI hasn't published benchmarks, so nobody can tell you what a good click-through rate or cost per click looks like for your industry. You can add a tracking pixel and send conversions back, but that only captures people who click and buy in the same visit.
That last part matters most for a channel that sits early in the funnel. Someone might see your ad in ChatGPT on Monday, think about it, then search for your business by name on Google on Thursday and fill in your form. Google gets the credit. ChatGPT gets nothing. Unless you look for that effect deliberately, you'll never see it.
So, are ChatGPT ads worth it?
For some brands, yes. Search is changing. More of the non-brand search queries that users use to research that used to happen on Google now happen in ChatGPT. If your customers research properly before they buy, being absent from those conversations is a risk. That applies to considered purchases, B2B services, software, higher-value ecommerce and brands that sell nationally. For them, I think ChatGPT is becoming a platform they need to be on.
For a local service business with a small budget, I'd wait. If most of your work comes from people searching for a trade or service in their area, your money will work harder on Google Search for now. I'd keep an eye on how Google's own AI answers change things too. I covered that in what Google AI Mode ads mean for UK campaigns.
How I'd test it
- Ring-fence the budget. Treat it as a test with a fixed spend and an end date, for example eight to twelve weeks. Don't take the money from Search campaigns that are already bringing in leads.
- Decide what success looks like before you start. For an upper-funnel channel, that might be more searches for your brand name, more direct visits or more enquiries that mention ChatGPT. Write it down before the first ad runs.
- Measure outside Ads Manager. Track ChatGPT clicks properly in GA4, watch your branded search volume in Google Ads and Search Console, and ask new enquiries how they found you. Add "ChatGPT" as an option on your enquiry form if you have a "how did you hear about us" field.
- Be specific in the ads. Advertisers with narrow, problem-focused messages have reported better results than those running broad brand messages. Write for the question someone is likely to be asking.
- Judge it against the right channels. Compare ChatGPT with your awareness and paid social activity, not your Search campaigns. If it's sending you cheaper new customers over the test, keep going. If it isn't, stop and try again in six months.
Where it's heading
ChatGPT ads are less than a year old, and it shows. OpenAI is still adding the basics. Target CPA bidding should arrive soon, it's testing a larger visual format and has hinted at ads that let people talk to a business directly. Google Ads looked very different in its first few years too.
I expect the platform, the reporting and the ads themselves to improve a lot over the next few years. For now, treat ChatGPT as a discovery channel. Test it with money you can afford to learn with, measure it properly and don't expect it to replace Google.
If you'd like a second opinion on whether it's worth testing for your business, or how it would fit with your Google Ads and paid social, send me a message below. I'll tell you what I'd do, including if the answer is not yet.