ChatGPT ads are paid placements shown inside ChatGPT conversations, bought through OpenAI's Ads Manager, which is currently in beta. They only appear to logged-in adults on the Free and $8/month Go plans, they're matched to conversations using broad "context hints" rather than keywords, and they use a short chat card format
Whenever a new ad platform launches, clients ask us the same things: who will see it, how much control do I have, how will we measure it, and does it deserve budget? ChatGPT ads don't answer those as neatly as search or social do. So rather than listing features, the sections below tackle the questions we'd expect to hear in a client call, starting with who actually sees the ads.
Ads only show to logged-in adults on the Free plan and the $8/month Go plan. Anyone on Plus, Pro, Team or Enterprise sees no ads at all.
That skews the audience toward casual and budget-conscious users, so check your customer fits the group you can reach before you put money behind it.
Device matters too. In our own testing, desktop users stayed on site almost five times longer than mobile users. Desktop sessions averaged 2 minutes 51 seconds, against just 35 seconds on mobile. We feel the reason for this will be desktop users are more likely to be working, planning or comparing options, which is the research behaviour where these ads seem to do best. Mobile use tends to be quicker and more exploratory. Thats why we're weighting more of our budget to desktop to help target b2b audiences looking for new makreting agencies.

Detailed location targeting is only available in the US, where you can target by state, DMA and ZIP code. In the UK, you're limited to targeting the whole country, or at best a large region.
That rules out a lot of what UK advertisers take for granted. You can't target a city like Leeds or Manchester, a postcode area, or a radius around a shop or office. If you're a local business, such as a solicitor, a trades company or a restaurant, you can't yet focus spend on the people who could actually become customers.
What we're doing. Because of these limits, our campaigns are targeting the UK as a whole. It's the only practical option for UK advertisers right now, and it shapes how we approach the platform:
Privacy rules, including UK GDPR and the EU's version, are part of why international targeting is rolling out slowly. We'd expect finer UK targeting to arrive eventually, but we wouldn't plan a campaign around it.
No. Advertisers supply "context hints", which are broad themes, topics or intents, and OpenAI's models match your ad to a conversation based on what the user is asking.
You can't add negative keywords, and OpenAI doesn't currently give you query insights, so you can't see what people typed before clicking. If you're used to search term reports, this is the biggest adjustment.
The main format is the chat card, which has a headline, a description, a single 1:1 square image and a small brand favicon.
| Element | Recommended | Hard limit |
|---|---|---|
| Headline | 16–24 characters | 50 characters |
| Description | 32–48 characters | 100 characters |
| Image | 1:1 square | 1 image |
Testers report that text gets truncated on mobile if you go beyond the recommended lengths. Write short, say one thing, and treat the hard limits as a ceiling you rarely need.
Yes, but not from one master account. There's no umbrella dashboard yet, so each client has to create their own brand account, complete business verification and invite agency staff as individual users.
It's not complicated, but it adds admin and waiting time. Build it into launch timelines.
For most brands, as a small, controlled test rather than a replacement for search or social. The platform is in beta and still has rough edges, but OpenAI is shipping quickly. It has already added a Conversions API, bulk product feed uploads via spreadsheet and a programmatic DSP integration with partners including Amazon.
Brands in research-led categories with a free-tier-friendly audience have the strongest case to try it now.
No, and it's maddening. OpenAI gives you no query data at all, so you've got no idea what someone typed into ChatGPT before they clicked your ad. No search term report, no prompt data, nothing. You're told your ad matched "the conversation" and left to take that on trust. Until OpenAI changes that, the only workaround is reading the post-click data: landing pages, device split, engagement and conversions. It's a poor substitute for knowing what people actually asked.
For most UK businesses, yes, but as a small, capped test and not a shift of your main budget.
ChatGPT ads are early, and it shows. You can't target below country level in the UK. You can't see what people searched. Your audience is limited to free and Go users. In our own testing, mobile visitors spent 35 seconds on site against 2 minutes 51 seconds for desktop, so where you see the clicks matters as much as how many you get.
The upside is that people are using ChatGPT to research and plan, which is exactly when a useful ad can earn a click. The brands best placed to try it are national or online businesses in research-led categories like travel, finance and ecommerce, with customers who sit on free plans.
Categories:
General |