ChatGPT Ads Just Went International: 7 Countries, Multi-Advertiser Auctions, and New Ad Tools Terms

One month of changes at a glance

ChatGPT Ads shipped product changes roughly weekly through June. The pace matches the pattern we tracked in May, when access opened to every US advertiser on May 5 and budgets, geo-targeting, and CTA buttons landed in a single week. June's wave was different in kind: less about new campaign controls, more about expanding where the platform operates and writing the rules for what comes next.

Three of the seven rows are the story. The international expansion changes who can buy. The multi-advertiser auction changes how ads are priced. The Ad Tools Terms change what advertisers will be able to do with their own data and creative. The rest of this post takes them in that order, then separates what is verifiably live from what is still gated, because, as we show below, even the AI engines themselves currently disagree on that question.

ChatGPT Ads is now open in seven countries

OpenAI's Help Center lists Ads Manager availability in Australia, Canada, Japan, Korea, New Zealand, the United Kingdom, and the United States as of July 7, 2026. The United Kingdom was the milestone: ads went live there on June 6, announced on LinkedIn by OpenAI's VP of monetization Benji Shomair, making it the first market outside North America and Oceania and the first where English is not the only language of the regulatory environment. Coverage of the UK launch named fashion retailer Zalando as a confirmed launch advertiser, with agency Dentsu participating through client campaigns.

The sequence matters for anyone planning international spend. OpenAI announced the expansion on May 7, naming the UK, Japan, South Korea, Brazil, and Mexico. Two months later, three of those five are on the availability list and Brazil and Mexico are still pending. OpenAI's own caveat is that country availability may continue to evolve as testing expands, so treat the Help Center list, not the announcement, as the source of truth. The scale context comes from OpenAI's March 26 disclosure that the ad business crossed $100 million in annualized revenue within six weeks of launch, and reporting since has put the company's internal target at $2.5 billion by the end of 2026.

For B2B marketers with pipeline in the new markets, the practical note is that launch-window inventory tends to be cheap relative to attention. The UK launched with managed, register-interest access and a handful of advertisers before Ads Manager availability followed, which is the same access-then-controls sequence the US walked from February through May. If the UK, Japan, or Korea matter to your pipeline, the Ads Manager walkthrough covers the setup flow once your account has access.

The announcement named five countries. The availability list shows three arrived. When a platform is shipping this fast, the documentation, not the press release, is the source of truth.

Multi-advertiser placements and a second-price auction

On June 8, Search Engine Land reported that OpenAI began testing placements that group several relevant ads together instead of showing a single sponsored result. Eligible ads are priced through a second-price auction, the mechanism most digital ad platforms use, in which the winner pays just above the runner-up's bid rather than their own maximum. OpenAI framed the goal as improving product discovery for users while creating more opportunities for advertisers in high-intent conversations. The test runs on a small subset of ads with no announced timeline.

Two implications are worth planning around. First, grouped placements mean your ad can win the auction and still sit next to a competitor's, which raises the bar on creative differentiation: in a single-ad unit, adequate copy wins by default, but in a grouped unit it competes side by side. Second, a maturing auction changes the economics. More competing bidders in one placement generally lifts clearing prices over time, which is exactly why the current window, while inventory is cheap relative to attention, favors advertisers who measure return properly and can defend the spend when prices move.

The same June 8 update quietly shipped the campaign-management features advertisers had been asking for since launch: you can now switch a campaign from a lifetime budget to a daily budget (enforced as an average daily budget with weekly pacing flexibility), clone a CPM campaign and convert it to CPC bidding in one action, set a custom max CPM bid on impression campaigns, and bulk-edit inside the Ads Manager. Combined with the cost-per-action objective that reached early access on June 5, the bidding stack now spans reach, clicks, and conversions.

The June 17 Ad Tools Terms: rules written before the features arrive

On June 17, OpenAI published its Ad Tools Terms, a policy document covering three tool families: Audience Tools, Creative Tools, and Conversion Tools. Platforms publish legal frameworks like this shortly before features scale, which makes the document worth reading the way we read Google's 2026 terms refresh: as a map of what the platform is about to do. Digiday covered the terms the same day.

Audience Tools: custom audiences get their rulebook

The terms define Audience Tools as features that let an advertiser upload or import audience data, and they define that data broadly: first-party data, identifiers, lists, segments, suppression data, exclusion data, rules, and instructions. In plain terms, this is the legal frame around custom audiences, the include-and-exclude list targeting that surfaced in a gated rollout on May 14 supporting CSV or TXT uploads of raw or SHA-256 hashed emails and phone numbers. The explicit mention of suppression and exclusion data confirms that keeping people out of a campaign, current customers, active deals, churned accounts, is a first-class use case, not a workaround.

The data-broker ban

The most consequential clause is a warranty every advertiser makes: audience data must be genuine first-party data, not purchased or licensed from data brokers, data marketplaces, or other third-party data suppliers. You may use a service provider, including an identity-resolution vendor, to process, format, hash, or match your data, but only if that provider is not the source of the data and does not enrich it with third-party data. The terms also enumerate twelve categories of prohibited data, including health status, sexual orientation, race or ethnicity, religious beliefs, political affiliation, immigration status, union membership, and financial distress, with health data called out under Washington's My Health My Data Act. For B2B marketers the message is clear: the CRM list you built is welcome; the list you bought is banned.

Creative Tools: AI-generated ads, with the liability on you

Creative Tools are defined as AI-powered features that generate, modify, transform, optimize, localize, or translate ad creative from an advertiser's own materials: catalogs, website content, images, logos. Any generated creative you approve is treated as a regular ad, and the terms make the responsibility split explicit. Per the document, OpenAI is not responsible for “errors, omissions, outdated information, or inconsistencies” in your materials or for claims arising from generated creatives you approve. Advertisers must independently verify every material representation, and the terms ban unauthorized digital replicas of another person's voice or likeness. Neither Audience Tools nor Creative Tools has an announced general-availability date, which is exactly why the terms matter: the rules arrived before the features.

The rules arrived before the features. When a platform publishes the legal framework first, the features are close enough to plan for.

Why AI engines disagree about what's live (we checked)

Here is the part that should give every marketer pause. On July 7, 2026, we asked the major AI engines whether ChatGPT Ads custom audiences are available, using the AI-Advisors citation-tracking pipeline that powers our own platform. Three engines gave three materially different answers about a feature on OpenAI's own ad platform.

The disagreement has a simple cause: the public record is genuinely ambiguous. The feature was never announced by OpenAI; it was spotted in the Ads Manager interface on May 14 and documented by trade press from screenshots. The June 17 terms formalized the rules without confirming availability. And OpenAI's Help Center, which documents twelve articles in its ChatGPT Ads collection, has no custom-audiences article as of July 7. Each engine resolved that ambiguity differently, which is a live demonstration of why the sources AI engines cite end up defining what buyers believe about your product category.

The operational takeaway is dated-status discipline: check your own account rather than trusting any summary, including an AI engine's. The audience upload, when your account has it, appears at the Create Ad Group and Ads step of campaign creation. If it is not there yet, the preparation below means you lose zero days when it arrives.

Four roadmap signals worth planning around

We read OpenAI's April 30 privacy policy update the same way when it telegraphed conversion tracking: the pixel and Conversions API shipped five days later. The June documents carry four signals of comparable weight.

• Creative automation is next. The Creative Tools section is the longest new addition to the terms, and OpenAI's ads lead David Dugan told Digiday that creative variation has been a real key to success on the platform. Expect asset-feed style ad generation, closer to how Google assembles responsive ads than to a blank-page creative studio.
• Third-party measurement is acknowledged, not scheduled. In a June 30 Digiday interview, Dugan said OpenAI will consider “the most trusted third-party partners” to collaborate with as the platform evolves. No vendors, no timeline. Until it ships, dashboard numbers remain self-graded homework; our measurement-stack guide covers how to build an independent read.
• The single ad format is ending. A July 1 Digiday report found OpenAI job listings for image, video, native, conversational, and interactive ad formats. The e-commerce product-feed pipeline, up to one million SKUs per advertiser, is already being built with Criteo as the first ad tech partner.
• Lookalikes remain the visible gap. The terms cover uploading your lists but say nothing about expanding them. Audience modeling needs conversion-event volume, which the June 5 cost-per-action rollout is now generating at scale. We flagged converter-based audiences as the likely next product in our measurement roadmap analysis, and nothing in the June documents changes that call.

The April 30 privacy policy predicted the pixel. The June 17 Ad Tools Terms are the same kind of document: read them as the roadmap, not the fine print.

What to do now, before the tools reach your account

Everything gated in this post shares one property: the preparation can be done today, and none of it is wasted if a rollout takes another quarter. The four steps below take roughly half a day.

Verify what your account can actually see

Walk your own Ads Manager before planning around any of this. Check the Create Ad Group and Ads step for the audience upload, the campaign settings for the budget-type switch, and the campaign list for the clone action. Note what is present and what is not, with dates. Feature access is rolling out account by account, and as the engine-disagreement table shows, no summary, human or AI, substitutes for looking.

Prepare and hash your first-party lists

Export your CRM segments by lifecycle stage: trials, customers, churned accounts, and target accounts. The gated rollout accepts raw or SHA-256 hashed emails and phone numbers in CSV or TXT files, and hashing before upload is the privacy-safer default. Normalize first, then hash:

# Normalize then hash each email: lowercase, trim, SHA-256, one per line
tr 'A-Z' 'a-z' < emails.txt | sed 's/^[[:space:]]*//; s/[[:space:]]*$//' | \
while IFS= read -r email; do
  printf '%s' "$email" | shasum -a 256 | cut -d ' ' -f 1
done > hashed-emails.txt

Most customer data platforms ship a built-in normalization-and-hash export, in which case use that. What matters is that the identifier is lowercased and trimmed before hashing, because a hash of “ Jane@Example.com” and a hash of “jane@example.com” will never match, and remember the data-broker ban: only lists you collected yourself belong in the file.

Decide the suppression strategy before the targeting strategy

Suppression is the underrated half of the Audience Tools definition, and it is where B2B budgets leak first. Excluding current customers from acquisition campaigns is the obvious move; excluding open opportunities so sales conversations are not shadowed by ads, and excluding recently churned accounts from win-back messaging until the cooling-off window passes, are the two that show up in cost-per-pipeline math.

Wire the measurement before the auction changes

Multi-advertiser placements will eventually change clearing prices, and you want a clean baseline before that happens. Confirm the conversion pixel fires, tag every destination URL with the canonical UTM convention (utm_source=chatgpt, utm_medium=cpc), and record your current average cost per click. If you are spinning up campaigns in the newly opened markets, the free Google Ads to ChatGPT Ads converter drafts the context hints and headlines, and OpenAI's crawler needs access to your landing pages, which our OAI-AdsBot guide explains.

Where AI-Advisors fits

The June wave widens the gap between running ChatGPT Ads and understanding them. More countries, more auction complexity, and soon more generated creative all raise the same question: which of this is actually reaching buyers, and does it pay back? That connective layer is where AI-Advisors sits.

• Pixel install without the tag-manager detour. Paste your OpenAI Pixel ID and AI-Advisors stages the conversion tag in your Google Tag Manager container for a one-click publish.
• Clean paid attribution. Consistent UTM parameters keep ChatGPT Ads clicks classified as paid search in GA4, so the new markets and formats roll up into a cost-per-pipeline read instead of a mystery bucket.
• Paid and organic on one screen. The AI Ads platform overlays your spend with your organic AI citation data, the same pipeline that produced the engine-disagreement table above, so you can see where paid and organic visibility compound on the same buyer questions.

The engines are already answering questions about your category, with or without your input. Paid placement is one lever; being the source the answer cites is the other. The June wave made the first lever bigger. The second one is still the cheaper of the two.

Frequently Asked Questions

Related Reading

• ChatGPT Ads Just Added Budgets, Geo-Targeting, and CTA Buttons
• ChatGPT Ads Just Opened to Every US Advertiser: What Marketers Should Do First
• ChatGPT Ads Conversion Tracking Is Real: What OpenAI's April 30 Privacy Policy Update Tells You About the Roadmap
• How to Use ChatGPT Ads: A Marketer's Step-by-Step Guide (2026)