ChatGPT Ads for B2B in Europe: what you need to know in 2026

ChatGPT Ads expanded to 31 European markets on 24 August 2026, with access going first through OpenAI's Ads Solutions team, agency partners and technology partners, and self-service through Ads Manager promised for "later this summer". It came a week later: on 31 August 2026 OpenAI updated its announcement to say self-service was available across the 31 markets, and its availability page, updated 25 September 2026, lists France, Germany and the other new European markets as open to it. Buying has changed since: impression-billed conversion campaigns (oCPM) became generally available in September, and OpenAI's documentation, updated on 1 and 2 October 2026, now tells advertisers to avoid custom audiences in the European Economic Area (EEA) and Switzerland. For a European B2B advertiser, the open questions are reach and measurement rather than access.

Four things tend to get collapsed into one, and this article keeps them apart: what OpenAI has officially launched in Europe, what its global Ads Manager documentation supports today, what independent observers have measured in earlier markets, and what nobody can yet verify for European B2B campaigns. I haven't run a ChatGPT Ads campaign in Europe, so read the performance discussion below as an assessment of other people's evidence, not as firsthand results or a European benchmark.

My sources are OpenAI's announcements, its Ads Manager and measurement documentation, its September 2026 ad policies, Digiday's reporting on the European rollout, consent and measurement, an OpenAI advertiser email published by Search Engine Roundtable, Search Engine Land's account of an early advertiser test, and the first published empirical audit of ChatGPT advertising. I rely on OpenAI for what the product can do, and on the independent material to check what anyone has actually observed and where the gaps are.

B2BLAST's paid-media offer still covers the five established platforms on the B2B advertising services page. What follows analyzes a launch; it doesn't claim ChatGPT Ads already has the evidence base of Google Ads or LinkedIn Ads.

Yes, as long as the legal entity that advertises and gets billed is based in one of the listed countries, but only since the end of August. OpenAI's European launch announcement, published 18 August 2026, said advertisers would initially enter through the OpenAI Ads Solutions team, agency partners and technology partners, with self-service to "follow later this summer". Digiday's report on the launch, published 19 August 2026, independently confirmed the same agency-led route and named Publicis, Omnicom, WPP, Havas, Dentsu and MediaPlus among the buying partners.

Self-service arrived a week after launch. OpenAI added an update to the top of its announcement on 31 August 2026 saying self-service through Ads Manager was now available across the 31 markets, and its Ads Manager availability page, updated 25 September 2026, marks France, Germany and the other new European markets as "Available". The same page still tells advertisers in those markets they can start through the Ads Solutions team or a partner instead, so managed buying and self-service now run side by side rather than one after the other.

In practice, access has moved like this:

Sources: OpenAI European launch announcement, 18 August 2026, updated 31 August 2026; OpenAI Ads Manager Availability, updated 25 September 2026; Digiday, 19 August 2026.

A smaller B2B advertiser can now open an account and buy directly, after checking two things: that the entity paying for the ads is based in an available country, and that the business and its landing page can pass OpenAI's ad review. OpenAI's campaign documentation, updated 2 October 2026, adds two practical limits: a minimum daily budget of €15 for accounts billed in euros, and a rule that some new self-serve accounts can at first advertise only in their home country, until identity verification is approved and the account has reached a required spend there. If you sell across Europe, run the first test where your billing entity sits.

OpenAI's advertiser basics, updated 24 September 2026, describe a unit placed below a ChatGPT response, and its user FAQ on ads, updated 28 September 2026, says ads are labeled as sponsored, visually separated from the response and without influence on the answers. Advertisers never receive the conversations.

The documented unit contains:

Sources: OpenAI, Ads in ChatGPT: The Basics, updated 24 September 2026; Ads in ChatGPT, updated 28 September 2026.

Reach is narrower than ChatGPT's usage figures suggest. OpenAI's user FAQ says ads may appear only on Free and Go plans and lists Plus, Pro, Business, Enterprise and Edu accounts as ad-free. In B2B that limitation weighs heavily: an advertiser can't assume that the professional audience behind ChatGPT's headline usage numbers is reachable through ads.

The first independent empirical paper, The Beginning of ChatGPT Ads, published in August 2026, collected more than 3,000 ads from 186 advertisers across 335 prompts, using 91 controlled US accounts. The early ads it recorded were clearly separated from answers and skewed toward consumer goods, which backs OpenAI's description of the format while saying nothing about European B2B reach.

OpenAI's announcement of 16 September 2026 also introduced opt-in text customization, which adapts headlines and descriptions to the conversation and translates them into the user's language. If your account offers it, test it on one campaign before you let it rework copy your legal team has approved.

The system picks the ad from the conversational context, not from an exact search term. OpenAI says it considers the intent of the current conversation, the landing page, title, copy and context hints supplied by the advertiser. Its ad-group documentation, updated 17 September 2026, asks for hints that describe what you offer, who it helps or when it is useful, and says outright that they are not exact-match keywords or targeting rules and don't guarantee delivery for particular conversations or audiences.

Account structure changes as a result. Where a Google Ads ad group can be built around query coverage and negatives, OpenAI's documentation builds a ChatGPT Ads ad group around a shared product, service or customer need, with several message variations behind it. The nearest familiar comparison is contextual matching fed by richer language; there is no hidden list of prompts to bid on one by one.

Europe adds a second distinction, between contextual and personalized delivery. Digiday's report on OpenAI's European privacy rules, published 8 June 2026, says personalized delivery requires explicit consent; without it, generic ads can draw on limited information such as the current conversation, location and time. In the EEA and Switzerland the question hasn't arisen yet: OpenAI's user FAQ says personalized ads are not initially available there, so ads rely on the current conversation and basic context such as general location and language.

The same limit reaches custom audiences. OpenAI's custom audiences documentation, updated 1 October 2026, lets advertisers upload customer or prospect lists (emails, phone numbers or Google Advertising IDs) to include or exclude people or to adjust bids, with at least 25,000 matched users needed for inclusion, but tells them to avoid custom audiences in campaigns targeting the EEA or Switzerland while ads personalization is unavailable there. The CRM-list targeting B2B teams run on LinkedIn or Google has no equivalent here for now.

In B2B, and more so in Europe where context does most of the work, that turns landing-page clarity and intent design into part of the media buying. A campaign about pipeline reporting, say, needs a page and context hints that name the business problem and the buyer who owns it. That fits the revenue-first measurement approach of the BLAST methodology, but the matching itself stays a black box: no advertiser can see the prompt behind each impression.

More than you'd expect from an awareness pilot. OpenAI's campaign documentation, updated 2 October 2026, lists three objectives, Views (CPM), Clicks (CPC) and Conversions, and the last can now be billed per 1,000 impressions (oCPM) as well as per valid click (oCPC). None of them charges per conversion: on a conversion campaign the bid cap is the most you'll bid for one, and the system turns predicted conversion likelihood into the auction bid.

Impression billing is the change to watch. OpenAI called it generally available, and its "recommended model for maximizing performance", in an advertiser email of 16 September 2026 that Search Engine Roundtable published the same day. OpenAI's conversion documentation, updated 1 October 2026, says this option optimizes toward conversions after ad views as well as clicks, and that choosing click billing will lower your conversions.

For B2B, I see that as a shift of risk toward the advertiser. On oCPM you pay for every impression, so a frequent, low-value event such as a newsletter sign-up can make a campaign look efficient without producing pipeline, while the event that matters, a qualified lead, is rare enough to fall under OpenAI's own warning that rare events make performance harder to evaluate. If you test it, optimize toward the standard event closest to a qualified lead and judge the result in the CRM against a click-billed campaign.

OpenAI's documentation describes this buying and measurement stack:

Sources: OpenAI Create Campaigns for ChatGPT Ads, updated 2 October 2026; Conversion-optimized Campaigns and Set up Custom Audiences, updated 1 October 2026; Measure Results, updated 29 September 2026.

OpenAI's advertiser basics still recommend a starting maximum CPC bid of $3-$5, and its conversion documentation gives no recommended bid at all. Read the CPC figure as a platform recommendation in US dollars, not a European clearing price or the cost per click to expect. No public source publishes a reliable European B2B distribution for CPC, CPM, cost per lead or conversion rate.

OpenAI's measurement documentation, updated 29 September 2026, says attributed conversions can take 24-48 hours to appear, and that with the one-day view window on, the Conversions total includes eligible view-through conversions (someone converting within one day of seeing an ad, with no click taking the credit). The attribution windows change reporting only, not optimization or billing. A total that can blend clicks with view-throughs is one more reason a serious B2B test needs its own analytics and CRM outcomes, tagged with UTMs, alongside the platform report. HubSpot users get part of that from OpenAI's HubSpot integration (help page updated 26 September 2026), which shows contacts and customers next to cost.

Very little about Europe, and nothing that deserves to be called a European B2B benchmark. The most concrete public comparison is Search Engine Land's early-campaign guide, published 10 August 2026, which reports GA4 data from about 1,500 users who reached the same website through Google Ads and ChatGPT Ads. It discloses no multi-advertiser sample and no European segment, so what you get is one test, not a market average.

Even so, it is a useful warning against assuming that conversational intent automatically brings better traffic:

Source: Search Engine Land, 10 August 2026; one advertiser comparison covering approximately 1,500 users.

The academic audit answers a different question: its 3,000-plus observed ads show that real delivery existed across 186 advertisers and that early inventory leaned consumer rather than B2B.

The newest reporting describes friction, not benchmarks. Digiday's report on OpenAI's measurement gaps, published 1 October 2026, quotes agencies running live campaigns: in one UK lead-gen campaign, form fills reached the advertiser while the platform still showed zero conversions two weeks later, and ChatGPT counted about 100 clicks against 20 in the agency's own tracking, though other agencies saw no click discrepancies. It also reports conversions arriving 24 to 36 hours late and legal teams refusing OpenAI's tracking terms. OpenAI's advertiser FAQ, updated 26 September 2026, says as much: the platform does not yet have performance benchmarks across advertisers, industries or campaign types.

Taken together, the sources describe a channel that exists and changes fast, and none of them measures how it performs for European pipeline. Until someone publishes a European dataset with its method disclosed, judge the channel on your own qualified-lead and opportunity economics; the comparison framework in our Google Ads B2B benchmarks article explains why a platform average can't stand in for account-level revenue data.

What makes it interesting is when the ad appears. In a single conversation, someone can lay out a business problem, their constraints, their current stack and how they will evaluate options, which is richer context than a two- or three-word query, and OpenAI designed its matching system around that conversational intent. For complex software and services, this opens a plausible route to being discovered before the shortlist is fixed.

Addressability is the first thing that could sink it. Because ads are limited to Free and Go plans, the professional and enterprise segments a B2B advertiser wants most may be underrepresented in the ad inventory. No public source breaks down the European split between free and paid users by job role, seniority or company size, so any claim that ChatGPT Ads reaches a given number of European decision-makers would be invented.

Then there is how little advertisers can see. They can supply context hints but can't look at the conversations or prompt themes that triggered their ads, and Search Engine Land compares the experience to the early days of Performance Max, when advertisers had to trust the matching with little diagnostic data. OpenAI offers country, device and platform breakdowns, conversion events and CSV exports, but no equivalent of a search-terms report.

Policy and measurement can also shift under a campaign. OpenAI's ad policies, updated 10 September 2026, put the advertiser, creative, landing page and placement through review and largely bar financial, health and legal services ads outside the US. They also keep the early phases focused on a limited set of consumer categories, and the September revision spells out OpenAI's right to decline ads that conflict with its business interests or competitive position, which a B2B software vendor competing with OpenAI should read before budgeting a test. Measurement moved as fast: view-through conversions sat outside the Conversions total in August, inside it by late September, and now feed oCPM optimization. A B2B test is credible only if the business is eligible, the landing page can pass review and success is defined in the CRM before any money is spent.

Next to the channels that already deliver demand and professional audiences, as a measured experiment rather than a replacement. Google Ads still offers observable search demand and a mature conversion stack, and LinkedIn Ads still offers explicit targeting by company, role and seniority. What ChatGPT Ads adds is the chance to enter a conversation while the user is still framing a problem or comparing approaches.

Wait until three conditions hold: your market has a real buying route, the Pixel or Conversions API and the CRM handoff are ready, and the budget can produce a learning sample without taking money from campaigns that already generate pipeline. Our Google Ads service page and LinkedIn Ads service page explain the established roles that the new channel would need to complement.

The implementation order in our Google Ads best practices still applies here: reliable measurement before optimization, then CRM signal feedback before advanced bidding. On ChatGPT Ads that means click billing first and oCPM only once the CRM shows which conversion event predicts pipeline, since view-through credit is the hardest part of OpenAI's report to check against your own data.

The launch is worth watching because it puts a paid surface at the point of research, but it doesn't yet justify a forecast dressed up as a benchmark. Register, write down the hypothesis, isolate the test by market and offer, and judge it on qualified pipeline rather than novelty.

Did ChatGPT Ads launch in Europe on 24 August 2026?

Yes. OpenAI expanded to 31 European markets, with buying first available through its Ads Solutions team, agency partners and technology partners, and self-service through Ads Manager followed on 31 August 2026. Check the availability page for your country before you plan a launch.

Can every European advertiser open Ads Manager today?

Not every one, but most can: OpenAI's availability page marks all 31 launch markets as available for self-service. The billed legal entity has to be based in a listed country, and a new account may at first be limited to its home country, so plan the first test there.

How are ChatGPT ads targeted?

OpenAI uses conversational intent, the landing page, ad copy and advertiser-provided context hints, plus settings such as country and platform. In the EEA and Switzerland ads aren't personalized yet and OpenAI advises against custom audiences, so write context hints that describe needs, not keywords.

What does ChatGPT Ads cost in Europe?

No reliable European B2B benchmark is public yet, and OpenAI says it has none of its own. Its $3-$5 starting maximum CPC bid is global bidding guidance and its €15 minimum daily budget is a floor, not a price. Build the benchmark from your own account.

Should a B2B company test ChatGPT Ads now?

Only if access, eligibility and conversion measurement are ready and the test doesn't displace a proven pipeline source. Start on click billing with an isolated learning budget, and judge it on qualified opportunities, not clicks or platform conversions.