OpenAI's own data on user base and conversation volume: Europe up 10.8×, a third of usage from 35+, and 2 billion weekly prompts about professional services. With the caveats that actually matter.
Free analysis
Curious how your marketing is really performing?
Get a free analysis of your digital channels — no fluff, just clear findings and next steps.
Book free analysisOpenAI has shared a data pack with advertising partners covering who uses ChatGPT and what they ask about. The numbers are interesting in themselves, but what makes them useful is that they say something about where ads can plausibly work — and where they probably cannot.
Here is the walkthrough, with the caveats that actually matter to a Swedish advertiser.
Europe has grown 10.8× in three years
Indexed to July 2023, weekly active ChatGPT users in Europe grew 10.8 times by June 2026. Fast, but not the fastest:
| Region | Growth, July 2023 → June 2026 |
|---|---|
| Africa | 24.1× |
| Asia | 15.9× |
| South America | 11.0× |
| Europe | 10.8× |
| North America | 9.2× |
Note what this does not say. It is relative growth, not absolute user counts and not ad reach. North America grows slowest because its base was already largest. For a Swedish advertiser the relevant conclusion is not "Europe is growing fast" but that the curve still points steeply upward in a market where ads now exist.
The audience is getting older — which matters for B2B
In June 2026, people over 35 accounted for 33.6% of user messages globally, up 5.3 percentage points year on year. The 18–34 group still accounts for 66.6%.
The direction matters more than the level. A platform where a third of usage comes from people over 35 — and the share is climbing steadily — is a different platform for B2B than one where that figure is a fifth. Decision-makers at Swedish companies sit in the growing part of the curve.
But two things have to be held in mind at once. Ads are shown only to users on Free and Go plans. The more senior and buying-capable the segment, the more likely they pay for Plus — and therefore see no ads at all. The age data describes the whole user base, not the addressable part of it.
What people actually ask about
OpenAI has categorised global user messages by primary topic. Volumes are average weekly messages in July 2026; growth compares July with January 2026.
| Category | Weekly prompts | Growth since Jan 2026 |
|---|---|---|
| Media & entertainment | 4.4B | +106% |
| Healthcare | 2.4B | +31% |
| Professional services | 2.0B | +23% |
| Jobs & recruiting | 700M | +22% |
| Financial services | ~600M | +33% |
| Travel | 330M | — |
| Cooking & recipes | 275M | +15% |
| Business operations (small business) | 90M | — |
The professional services category covers legal, financial and tax, business research, and sales and marketing. Two billion messages a week, growing 23% in six months. That is the single most relevant line in the whole pack for a Swedish B2B company: the conversations exist, at volume, and they are about precisely the areas where consulting services get sold.
Note the bottom row. OpenAI qualifies it themselves: 90 million business operations messages are "core small-business operations" and explicitly not a B2B buyer-intent measure. That is an unusually honest caveat in a sales deck, and it deserves respect. None of these figures are ad inventory or verified purchase intent — they measure conversation volume.
The seasonal patterns are real, but American
The pack shows clear peaks in product-search activity around US retail holidays:
- Cyber 5 (27 Nov–1 Dec): +47% on average, peaking at +94% on Black Friday. Computers/electronics and appliances +94%, apparel +63%.
- Back-to-school (August): +26% on average, personal care +46%, apparel +42%.
- Labor Day week: +24%, personal care +31%.
All of this is based on US product-search activity. Black Friday and Cyber Monday have a foothold in Sweden and the pattern is probably transferable to some degree; Labor Day and the American school year are not. Treat the figures as evidence that seasonal patterns exist in the channel rather than as a Swedish calendar.
For Swedish e-commerce the practical conclusion is that purchase-decision activity concentrates predictably, and budget should be planned accordingly. For Swedish B2B it is less useful — our seasonality follows fiscal years and holidays, not sale weekends.
What advertisers report
The pack also contains advertiser accounts from the pilot. Two are worth highlighting.
HubSpot joined to understand how AI-driven discovery is reshaping customer journeys. They report competitive CPAs across a number of campaigns and — more interestingly — downstream revenue signals suggesting that traffic from ChatGPT Ads actually produces business outcomes. Their Lead Marketer in Paid Advertising describes it as "another way to diversify our paid media strategy while reaching customers in a more conversational environment".
Figma reports that ChatGPT Ads quickly became a top performer for their Make campaigns, with higher click-through rates than traditional search and stronger-than-average onsite engagement after the click.
That last point deserves attention. The early third-party figure of 0.91% click-through rate, which circulated widely when the pilot was new, is in direct conflict with what Figma reports. We leaned on 0.91% ourselves when building our own campaign and derived bids from it. We are leaving the calculation in place because it is conservative — but treat the figure as a pessimistic planning assumption, not a verdict.
What the data should actually change in your plan
Four conclusions we draw for Swedish advertisers.
Volume is not the problem
Two billion weekly messages in professional services means the conversations exist. The question is not whether your audience is there, but whether you can reach them profitably given that only Free and Go users see ads.
Category decides whether the channel is right
Media, healthcare and finance are growing fastest. If you are in a category with high conversation volume and high order values, the channel is worth testing now. If you are in a niche B2B category with low volume, GEO is probably the better first investment.
Measure on conversion, not clicks
Both HubSpot and Figma report on downstream signals rather than CTR. That is the right instinct in a channel where click behaviour differs from search.
None of these numbers are reach
OpenAI repeats the caveat on nearly every slide: volume is not ad inventory and not verified purchase intent. Use the data to choose category and timing — not to forecast impressions.
Want to know how your company shows up in AI-generated answers today? Book a free analysis, or read more about GEO optimization.
