Your ChatGPT Ads problem this week is not another map pin.
On September 24, 2026, MARKETECH APAC reported that OpenAI is rolling ChatGPT Ads into Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam, and Taiwan, taking the product past sixty countries. TechNode Global filed the same seven-market list the same day. PPC Land’s Luis Rijo, also September 24, traced the disclosure to Benji Shomair’s LinkedIn post and an OpenAI newsroom note dated September 23 on the US calendar, then flagged the one-day start-date gap that shows up whenever West Coast announcements hit Southeast Asia mornings. The buying routes are familiar: OpenAI Ads Solutions, agency partners (dentsu, Havas Media, Omnicom Media, Publicis Groupe, WPP), technology partners, and self-serve Ads Manager for eligible businesses.
That sentence is what every regional Slack channel will paste into a brief. It is also the wrong control surface if you treat it like a US or Australia campaign with seven new country checkboxes.
Country targeting is a button. Eligibility, inventory shape, commerce context, and who writes the creative are the job.
What landed on September 24
Four facts. Keep them operational.
Fact one: seven markets, sixty-plus footprint, self-serve named on day one. MARKETECH APAC and TechNode both list Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam, and Taiwan. Both say ChatGPT Ads is now available in more than sixty countries, following earlier Asia Pacific launches in Australia, New Zealand, Japan, South Korea, and India. PPC Land’s Help Center check on September 24 found sixty-three countries marked Available for self-service, including all seven new markets. That is the operational contrast with India: Rijo notes Indian advertisers went live August 29 with ads running and no self-service door, while these seven open with Ads Manager on the menu from the first day, closer to the Brazil and Mexico pattern.
Fact two: Free and Go inventory only, ads labeled off the answer. MARKETECH APAC restates the boundary: ads appear only to users on Free and Go plans; Plus, Pro, and Enterprise stay ad-free. TechNode adds Business and Edu to the ad-free list. OpenAI’s standard assurance travels with the launch copy: ads are clearly labeled, separate from ChatGPT’s answers, and do not influence responses. Advertisers do not receive people’s conversations, chat history, or personal details. If your forecast assumes Plus or Enterprise reach in Jakarta or Taipei, rebuild the forecast before you rebuild the creative.
Fact three: Help Center still gates international self-serve by home country. OpenAI’s Create Campaigns article (accessed September 24) is blunt under “Choosing countries to target.” Some new self-serve ad accounts can initially advertise only in their home country, the country associated with the ad account. Selecting other countries does not override that restriction. The account becomes eligible for other supported countries only after identity verification is approved and the account reaches the required advertising spend from running campaigns in the home country. Creating a campaign or setting a budget alone is not enough. Ads Manager Availability adds the incorporation rule: the legal entity that advertises and is billed must be based in a country listed as available. A Singapore hub can buy. A Vietnam-registered retailer can buy. A company stuck off the list can sign up and wait. Geo open is not “clone US and tick seven boxes.”
Fact four: Shopee is the named commerce partner, and OpenAI is selling complement, not replacement. MARKETECH APAC’s exclusive interview quotes OpenAI positioning ChatGPT Ads as a way to reach people while they explore a need, compare options, or decide what to do next, and as “a complement to the channels advertisers already use, including commerce platforms.” Shopee is named in the announcement. Peggy Zhu, Executive Director for Brand and Growth Marketing at Shopee, said the platform is extending a collaboration that followed ChatGPT Ads in Brazil into Southeast Asia and Taiwan. Dave Dugan, Vice President and Head of Global Ads Solutions, framed the expansion as opening the channel to businesses of every size. OpenAI also told MARKETECH it had seen strong waitlist interest in Southeast Asia and expects to keep learning across markets with different languages and business needs.
Objectives and floors sit in the same Help article: Views (CPM), Clicks (CPC), or Conversions (oCPC or oCPM billing). Minimum daily budgets now include local currencies operators will actually see: SGD 4.00, IDR 14,000, MYR 5.50, PHP 50.00, THB 33.00, VND 19,000, TWD 60.00, beside the familiar USD 25 floor. Low floors are not a license to treat seven markets as one experimental line item.
Bright line against last month’s ChatGPT Ads tickets
This essay is not a rematch of the last thirty days of Vault ChatGPT Ads theses. Say the bright lines once, then leave them alone.
Not The catalog is the campaign: Shopify feed craft. Not The bill is not the click: oCPM billing. Not The click is the conversation: Sponsored Agents. Not CRM is the ads console: HubSpot closed-won joining spend. Not The pipe is not the placement: Amazon DSP path versus placement control. Not yesterday’s The search term is not the journey or Monday’s upgrade-is-not-a-toggle Google stack: AI Max briefs and settings migrations.
Today’s ticket is geo expansion as a media decision: who can buy, which inventory you actually reach, how super-app commerce already owns the transaction moment, and whether your staffing model can localize without cloning Sydney or San Francisco creative into Bahasa Indonesia and Vietnamese.
If your working group is still arguing feed quality, oCPM math, or Sponsored Agent click paths under a “SEA launch” subject line, you are in the wrong thread.
What operators get wrong
Mistake one: treating country targeting as the unlock. Checking Indonesia in Ads Manager feels like progress. Help Center says new self-serve accounts may still be stuck on home-country until verification and required home-country spend clear. Selecting other countries does not override that. Teams that plan a seven-market flight without checking eligibility and legal-entity country will discover the gate after creative is approved and the CFO is watching.
Mistake two: cloning the US or Australia playbook into seven languages. English proof, AU seasonal hooks, and US retail calendars do not survive Ramadan timing, 11.11 commerce peaks, or markets where Shopee and other super-apps already own comparison and checkout. OpenAI is selling a complement to commerce platforms. If your offer logic ignores that stack, you bought conversation inventory that dumps into a journey the shopper already finishes elsewhere.
Mistake three: forecasting Plus/Pro/Enterprise reach. Inventory is Free and Go. Paid tiers stay ad-free. Media plans that smear “ChatGPT users in SEA” into one TAM number will overstate reachable volume and understate how price-sensitive the Free cohort is.
Mistake four: assuming self-serve and agency are the same product. MARKETECH and TechNode both name the holding-company path beside Ads Manager. Self-serve lowers the door for smaller advertisers. Agency paths matter for multi-market compliance, category policy, and measurement scaffolding still expanding. Dugan’s MARKETECH briefing comments underline two policy layers: OpenAI’s own category and safety rules, plus local regulatory work before launch. Early focus categories include household and consumer goods, local services, travel and entertainment, and digital products and education. If you are regulated or multi-brand, “we clicked self-serve” is not governance.
Mistake five: staffing one English pod as seven markets. Language is the untested variable PPC Land flags after Japan and Korea. Bahasa Indonesia, Malay, Filipino contexts, Thai, Vietnamese, Traditional Chinese for Taiwan, plus English in Singapore, are not bid modifiers. They are creative, moderation, and QA queues. Machine-translating a master English set will buy the wrong creative model even if the country checkbox works.
Mistake six: measuring like Search or Meta because the objective picker looks familiar. Views, Clicks, and Conversions map onto CPM, CPC, and optimized billing. Conversation context still decides delivery. OpenAI told MARKETECH that delivery draws on the current conversation and, when available and enabled, broader account context. If your weekly readout only shows CPC and ignores whether ads showed up in comparison chats versus idle prompts, you will optimize the wrong moment.
Mistake seven: calling waitlist interest a performance case. OpenAI cited strong advertiser interest and waitlist demand for Southeast Asia. It did not publish regional usage or outcome benchmarks in the MARKETECH interview. Tens of thousands of advertisers globally and a late-August one-billion-dollar annualized revenue run rate are scale color, not your Jakarta CPA target. Pilot with local creative and local commerce truth, or you will present waitlist enthusiasm as ROI.
Second-order effects for media and agencies
Holding-company pods get a staffing fork. dentsu, Havas, Omnicom, Publicis, and WPP are named routes. Work will split between self-serve test budgets inside brand teams and managed buys that need local language creative, category counsel, and multi-market reporting. Agencies that bill “SEA activation” as one workstream without naming home-country eligibility owners will miss the Help Center gate and blame the platform.
Commerce and performance have to sit in the same room. Zhu’s quote is not a logo slide. It is OpenAI admitting the region’s transaction gravity lives in super-apps. ChatGPT Ads is sold as the exploration moment beside those platforms, not as their replacement. Media that cannot explain incrementality versus Shopee or rival retail media will lose the budget fight to whoever owns those dashboards.
Measurement politics get sharper across languages. TechNode notes performance data will accumulate across different languages and ad ecosystems. That only helps if you refuse a seven-country blended ROAS before creative is localized. Home-country spend requirements also mean early learning may concentrate where the legal entity sits, not where the brand dream map points.
Policy becomes launch work. Dugan’s two-level frame means market-by-market category checks belong in the kickoff, especially for health-adjacent or financial advertisers who assume US rules travel intact.
Entity structure becomes media strategy. If your APAC bill-to is Singapore but your growth bet is Indonesia plus Vietnam, plan the home-country unlock before you promise seven live markets. No entity in an Available country means partner or waitlist path, headline or not.
What to do this week
Run this as a geo eligibility and localization sprint, not a map-coloring exercise.
- Audit account eligibility before creative. Confirm legal-entity country against Ads Manager Availability. Confirm whether the account is still home-country gated. Document identity verification status and whether required home-country spend has cleared. Do not brief seven markets until that sheet is green or explicitly partner-routed.
- Pick one home market and one expansion candidate, not seven. Choose the market that matches your entity and language strength. Run Views or Clicks long enough to learn conversation contexts, then decide if Conversions wiring is honest. Leave the other five as a sequenced backlog with owners, not muted checkboxes.
- Rewrite creative for Free/Go and local commerce reality. Lead with offers that make sense when the shopper already has Shopee in their pocket. Local currency, local holidays, local return norms. Kill US shipping claims and AU seasonal jokes. Keep ads on-brand for labeled inventory that sits beside answers, not inside them.
- Staff language owners, not translators of a master English RSA. Name a writer or agency lead per live language. Require human QA on every variation. Machine translation can draft; it cannot own brand safety in Thai or Vietnamese. Put that cost in the media plan so finance does not treat localization as free.
- Choose the buying path on purpose. Self-serve for a single-market test with a clean entity. Agency or Ads Solutions when you need multi-market policy, category navigation, or measurement scaffolding you lack in-house. Do not pretend those are the same SKU.
- Build a readout that matches the inventory. Segment by country and by conversation theme where Insights allow. Track Free/Go assumptions explicitly. Refuse a seven-market blended ROAS until each market has localized creative and a named owner.
- Brief leadership on complement, not conquest. Use Zhu’s framing and OpenAI’s own complement language. ChatGPT Ads is an exploration surface beside super-apps, not a reason to pause retail media. Ask for incremental tests, not channel replacement theater.
The operator close
OpenAI did not hand you a playbook when it opened seven markets. It handed you a map, a self-serve door for eligible entities, agency rails for everyone else, Free and Go inventory, and a Help Center sentence that still says home-country first.
Country targeting is the obvious button. The control surface is eligibility plus localization plus staffing. Teams that clone a US or Australia campaign into seven checkboxes will buy the wrong creative, the wrong measurement story, and the wrong people to run it.
Ship the eligibility audit. Pick one home market that can actually spend. Write local creative for Free and Go shoppers who already live inside super-apps. Treat every extra country as a staffing decision, not a checkbox victory.
