On 21 October 2026, OpenAI added 26 markets to ChatGPT Ads, taking coverage to 89 countries. Argentina, Chile, Colombia, Peru, Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh and Pakistan were among the additions. Ads appear only on the Free and Go tiers; Plus, Pro and Enterprise stay ad-free.
For brands, the interesting part is not the map. It is that paid placement and organic product recommendations now occupy the same screen for the first time — and are allocated by completely unrelated mechanisms. Budget moves one. It does not move the other.
Key takeaways
- 89 countries as of 21 October 2026, up 26 markets in one step.
- Ads show on Free and Go only — which is also where most casual, signed-out shopping questions come from.
- Over 100,000 advertisers have run ads since launch on 9 February 2026 (cumulative, not currently active). Annualised revenue run rate reached $1bn on the 31 August 2026 figure, not updated in this announcement.
- CPM and CPC bidding; targeting works on conversation-topic hints, not keywords.
- Advertising cannot buy a product recommendation. Recommendations are decided by catalogue data and source citation.
Where things stand
Dimension | Current state |
|---|---|
Markets | 89 countries and regions (26 added 21 Oct 2026) |
Tiers shown on | Free and Go only; Plus, Pro and Enterprise have no ads |
Advertiser base | 100,000+ have run ads since 9 Feb 2026 — cumulative, not active |
Revenue scale | $1bn annualised run rate, on the 31 Aug 2026 figure |
Bidding | CPM and CPC |
Targeting | By platform (iOS app, Android app, web); relevance described through conversation-topic hints rather than keywords |
Newest format | Visual ads beside image-generation results, from 5 Oct 2026, US first |
Ecosystem | HubSpot and Shopify integrations, US-led |
Brand safety | IAS and DoubleVerify pilots in progress, US-led |
Two systems, one screen
This is the distinction that decides how you plan. Ads and product recommendations appear in the same interface and are allocated by separate machinery.
Ads go to whoever bids, on Free and Go, in 89 countries, against conversation-topic relevance.
Product recommendations and buy links go to whoever is in the merchant catalogue and whoever is cited among the answer's sources. They appear on every tier, including the paid ones where no ad will ever run. In our hands-on test of 97 buy links on 21 October, 17 carried utm_medium=feed — the signature of catalogue-sourced placement — and none of them had anything to do with ad spend.
The practical consequence: paid solves reach; catalogue and content solve being recommended. A team that funds only the first will buy impressions on the Free tier while losing the shelf on every tier. We measured what being absent from the sources costs: when a brand's product page is among the answer's sources, the buy button points at the brand's store 34% of the time; when it is not, 6%. No ad budget changes that number.
The tier split is a targeting fact, not a footnote
Ads show on Free and Go. Those are the tiers where signed-out and light users live — and signed-out, casual sessions are where the highest volume of shopping questions comes from. So the ad inventory is well aligned with purchase intent.
But it inverts for your most considered buyers. A Plus or Pro subscriber researching a significant purchase will never see your ad. For them, the only route into the answer is the organic one: catalogue presence and source citation.
Worth noting alongside this: in our testing, signed-out sessions showed only "Searching the web", while signed-in sessions displayed the rewritten search terms. The two populations are having measurably different experiences of the same product, and your plan needs to account for both.
Targeting by topic, not by keyword
Targeting is described in terms of conversation-topic hints rather than a keyword tree. That is a different unit of planning than paid search, and lifting a Google Ads structure across will not fit.
Three adjustments that follow:
- Organise creative and landing pages around conversational situations, not match types. "Gift for a father who has everything" and "earbuds that stay in for side sleepers" are the units, not head terms.
- Expect the ad and the answer to be read together. The user sees both in one scroll. An ad promising something the organic answer contradicts will underperform for reasons your ad platform will never show you.
- Keep measurement separate. Ad clicks arrive through ad tracking. Organic ChatGPT clicks arrive carrying
utm_source=chatgpt.com, with catalogue-sourced ones also carryingutm_medium=feed. Mixing them produces a channel number that cannot be acted on — see utm_source=chatgpt.com explained.
Where this sits in October
The ads expansion was one of three things that landed in October, and the only one of the three OpenAI documented fully. GPT-6 and the Intelligent UI shipped on 7–8 October and changed how shopping answers are rendered. The UTM outage ran 8–15 October and corrupted a week of everyone's attribution. The ads expansion landed on 21 October.
They get discussed as a single "October change". They are three separate systems, with three separate owners on your side, and conflating them produces bad diagnosis. The full breakdown is in how ChatGPT shopping results work after GPT-6.
What we would do with a budget
- Fund the organic layers first, because they are the ones that work on every tier and cannot be bought later. In order: product page data, catalogue feed, source presence. The sequence is in how to get listed in ChatGPT shopping.
- Use ads for reach into the Free and Go population, especially in the 26 newly added markets where competition is youngest.
- Instrument both before spending on either. If you cannot separate ad traffic, page-sourced organic traffic and feed-sourced organic traffic, you will not learn anything from either budget.
- Track competitors' ads as market intelligence. Who is bidding, on what topics, in which markets, is a usable read on where a category is heading — that is what GEOly's ChatGPT ads library is for.
- Keep a single scoreboard for the organic side. Share of Model across ChatGPT, Google AI Mode, Gemini, Google AI Overviews and Microsoft Copilot tells you whether the unbuyable half is moving.
FAQ
How many countries have ChatGPT Ads? 89 countries and regions as of 21 October 2026, after 26 markets were added in a single expansion.
Which ChatGPT plans show ads? Free and Go only. Plus, Pro and Enterprise do not show ads.
Can I pay to appear in ChatGPT's product recommendations? No. Ads and product recommendations are separate systems. Recommendations are determined by merchant catalogue data and by which pages are cited as sources.
How does ChatGPT Ads targeting work? Through platform selection — iOS app, Android app, web — and relevance expressed as conversation-topic hints rather than keyword matching. Bidding supports CPM and CPC.
How many advertisers use ChatGPT Ads? More than 100,000 have run ads since launch on 9 February 2026. That is cumulative, not the number currently active — the two are often reported interchangeably and should not be.
Are ChatGPT Ads worth it for a DTC brand? They buy reach on the tiers where casual shopping questions concentrate. They do not buy a place in the answer. For most DTC brands the catalogue and citation layers are the higher-return investment first — see ecommerce brand solutions.
Sources
Market count, tier restrictions and the expansion date: CNBC, "OpenAI expands ChatGPT ads to 26 new countries", 21 October 2026, and PPC Land's same-day reporting. Launch date, cumulative advertiser count, revenue run rate (31 August 2026 figure), bidding models, targeting description, integrations and brand-safety pilots: OpenAI announcements and the above reporting. Visual ads beside image generation: TechCrunch, 5 October 2026.
The buy-link parameter counts and the citation-versus-button percentages come from GEOly Research's 21 October 2026 hands-on test and from WISLR (4 brands, 250 prompts, 10–13 October 2026) respectively.
