Are ChatGPT ads worth it for a small UK business?

Somebody will ring you about this in the next few months, and they will sound very confident. OpenAI has started selling advertising inside ChatGPT, it is open to UK businesses, and the agencies are already writing "ChatGPT Ads specialist" on their websites.

Nobody has a year of data on it. Not them, not us. So here is what is actually known, the sum you can do on the back of an envelope tonight, and the one thing that will decide whether your test tells you anything at all.

What has actually launched

Ads appear inside ChatGPT conversations for logged-in adults on the free and lower-paid tiers. They started in the United States in February 2026, a self-serve Ads Manager followed in May, and it has since opened to more countries including the UK.

The targeting works off the conversation you are having right now, not off following you around the internet. If somebody is asking ChatGPT how to fix a whistling extractor fan, that is the moment an ad from a ventilation firm can appear. Advertisers do not get to read the conversation.

That is the genuinely interesting part. Somebody describing a problem in their own words is further along than somebody typing three words into Google.

What it costs to find out

Reported click prices sit somewhere around $3 to $5. Treat that number carefully: it comes from advertisers talking publicly and from agency blogs, not from OpenAI, and early prices on any new platform move about. Check it in the Ads Manager yourself before you budget.

But run the sum with £3 a click, because the shape of the answer matters more than the exact figure:

• 100 clicks costs you £300.
• If 5 of those 100 people send an enquiry, that is £60 per enquiry.
• If 1 in 4 enquiries turns into a job, that is £240 per job won.

Now the only question that matters: what is a job worth to you?

If you install ventilation systems and a typical contract is £4,000, £240 is a good day. If you sell £45 haircuts, it is nonsense and no amount of clever targeting fixes that. The maths is brutal and it is the same maths as every other paid channel: your margin per sale decides whether the channel exists for you.

Worth saying plainly: a 5% enquiry rate and a 1-in-4 close rate are decent assumptions, not promises. Put your own numbers in. If you do not know your own numbers, that is the thing to fix before you spend anything on ads.

The part nobody mentions

Here is where most of these tests go wrong, and it has nothing to do with the ads.

Almost all conversion tracking runs in the visitor's browser. A small script fires when somebody reaches a thank-you page, and that is how the ad platform learns the click turned into an enquiry.

Except a good share of those never fire. Ad blockers stop them. Safari restricts them. Somebody fills in your form, hits send, and closes the tab before the page finishes loading. The enquiry is real, it is sitting in your inbox, and the ad platform never hears about it.

On an established channel you can live with that, because you already know roughly what good looks like. On a brand new one you cannot. You spend £500, you get six enquiries, the dashboard shows two, and you conclude the platform does not work for your trade. It might have worked fine. You just could not see it.

The fix is server-side tracking: your website reports the enquiry from your own server, at the moment your server accepts it, rather than hoping a browser script survives. OpenAI supports this properly through their Conversions API. Most small setups never bother, because it means touching the site rather than pasting a tag.

So before you put money into a test, ask one question: when somebody sends me an enquiry, does my ad platform definitely find out? If the answer is "I think so", you are about to buy an expensive opinion.

Who this makes sense for right now

Test it early if two things are true:

1. Your margin per sale is high enough to survive a few hundred pounds of learning. Contractors, clinics, jewellers, premium goods, B2B software: yes. Low-ticket, high-volume: not yet.
2. You can measure it honestly. See above. Without this you are not running a test, you are running a donation.

The argument for going early is simple and it is not hype: the auction is thin. Fewer advertisers bidding means cheaper clicks than the same audience costs you on Google. That advantage closes as everyone piles in, and it always does.

Who should wait

Wait if your website does not convert the traffic you already get. Paid clicks do not fix a site that people leave. You will just pay to watch more people leave, faster, and you will learn something you could have learned for free.

Wait, too, if somebody quotes you a cost per lead on this platform. Ask them where the number came from. On a channel this new, the honest answer is that there isn't one yet, and anyone who has one has invented it.

What to do this week

Do not start with the ads. Start with the measurement, because it is the half that stays true no matter what you spend, and it makes every other channel you run more honest at the same time.

1. Work out what one job is actually worth to you, and what you can afford to pay for one.
2. Check whether your site reports enquiries server-side or only from the browser.
3. Then, and only then, put a small budget behind a test and read the result.

Getting found in ChatGPT has a free half too, and it is worth more than the paid half for most businesses. We wrote about that separately in getting found in Google and ChatGPT.

If you want the measurement side handled properly, that is a fixed one-off job and we do it on sites we built and sites we didn't: ChatGPT Ads setup. We don't run campaigns, so we have nothing to gain from telling you the ads work.