ContentsStart0%
ChatGPT Ads began rolling out to Taiwan in September 2026 alongside Indonesia, Malaysia, the Philippines, Singapore, Thailand and Vietnam. OpenAI’s announcement is dated September 23; the original Taiwan article recorded the rollout on September 24 local time. The expansion brought the advertised footprint above 60 countries and included self-service access for eligible businesses.
This is an archive of that Taiwan launch and its market implications, not a current worldwide setup guide. Account eligibility, supported markets and product formats can change after the dates discussed here.
What did the launch make available?
OpenAI described ads displayed separately from ChatGPT’s answers and identified as sponsored content. Its launch materials put Free and Go users within the ad-supported audience, while maintaining ad-free paid tiers such as Plus and Pro. The official explanation of ads also describes account, age and personalization conditions; those conditions should be checked for a present-day use case rather than assumed from this archive.
At launch, advertisers could work with OpenAI’s ad team, agencies or technology partners. Eligible businesses could also create campaigns in Ads Manager. Eligibility and an available budget did not guarantee exposure in a particular conversation: the ad system still had to select an ad under its relevance and bidding rules.
OpenAI said paid ads did not determine its independent answers and advertisers did not receive private conversations. It described conversational context as relevant to ad selection, with broader signals subject to personalization settings. These are the platform’s stated boundaries, not an independent audit of every delivery decision.
| Experience | What it means | What it does not establish |
|---|---|---|
| An ad beneath an answer | A labeled paid placement distinct from the answer. | That the advertiser purchased the recommendation in the answer. |
| A Sponsored Agent conversation | A user chooses to open a separate branded conversation after an ad. | That the format was available in Taiwan at this launch. |
| An agent buying on a user’s behalf | A delegated transaction requiring relevant tools and authorization. | That conversational advertising includes checkout or payment authority. |
The Chinese original used an illustration placing an ad and branded conversation side by side. This English version expresses the distinction in text so a concept image is not mistaken for a Taiwan product screenshot.
How the US pilot led to the Taiwan expansion
The original timeline starts with the February 9 US ad test, followed by Criteo’s participation in March and the May 5 self-service Ads Manager beta. The platform subsequently added bidding and measurement capabilities, expanding the means through which advertisers could run and evaluate campaigns.
On August 31, OpenAI reported a US$1 billion annualized revenue run rate less than 200 days after launch, with tens of thousands of advertisers. An annualized run rate extrapolates the revenue pace at that point; it is not US$1 billion of cumulative cash collected.
September brought integrations with HubSpot and Shopify. OpenAI’s September 16 announcement also described AI tools for drafting creative material and managing campaigns through natural-language instructions. These changes matter to agencies because the platform increasingly performs routine work that previously justified service fees.
Amazon Ads announced a ChatGPT integration on September 10, initially with selected US advertisers. This was another distribution route for advertisers, not evidence that Amazon’s entire catalogue or checkout had been connected to Sponsored Agents.
What early performance evidence could—and could not—tell Taiwan advertisers
Individual advertisers reported different outcomes. The original coverage cited Speechify describing useful new business from relevant tool-comparison contexts, while Launch10 reported lower click-through rates than Google Ads and fewer creative-testing options. Those are named advertiser experiences reported in Business Insider’s coverage; they do not establish a platform-wide or Taiwan-wide return.
Criteo reported that traffic from large-language-model platforms converted at roughly 1.5 times the rate of other referral channels in an observation involving 500 US retailers in February 2026. The denominator matters: this was traffic referred by AI platforms, not an isolated evaluation of paid ChatGPT advertising. Applying the figure to Taiwan campaign forecasts would change both the traffic type and the market.
OpenAI also said it did not observe a decline in trust measures during its early pilot. That is a platform-reported user-response observation. User acceptance and advertiser profitability answer different questions.
The comparison of nine named advertising cases keeps spend, clicks, attributed revenue and ROAS separate. A test budget should depend on those definitions, attribution windows and the advertiser’s own customer economics, not on an AI referral headline.
Why Taiwan’s existing search budgets matter
There was a substantial incumbent budget to compete for. The DMA 2025 market statistics published by the Taipei Association of Advertising Agencies put Taiwan keyword advertising at NT$16.72 billion. Against a digital advertising total of NT$66.177 billion, that is approximately one quarter of spending. These are 2025 market estimates, not ChatGPT ad revenue.
LS-Labs’ interpretation is that a conversational ad channel can compete with part of search advertising when people use AI to compare suppliers or products. It may also compete for budget with SEO and GEO, because all three can be presented to the same budget owner as ways to enter a buyer’s consideration set.
This does not mean an ad replaces every purpose of content. Accurate product information, documentation and independent research continue to support decisions before and after a click. It means an article count or an AI-mention count is a weak justification for spending unless the buyer understands what the work contributes.
A useful comparison asks which audience each activity reaches, which actions it enables, and what evidence connects it to an inquiry or sale. Advertising can be bought directly; durable information has a different cost and lifespan. They should be compared without assuming identical outcomes.
Why agents could accelerate the competition for attention
If a user delegates product selection to an AI agent, more comparison may occur before any visit to a brand’s website. The operator of that interaction can create additional paid opportunities to reach the user. That is a market hypothesis, not a behavioral change proven by Taiwan’s launch.
Sponsored Agents illustrate one such format: users can ask a brand questions after choosing an ad. At the September announcement, the format was being tested with selected US advertisers. It was not a purchasing agent, and its existence did not prove advertisers could buy control over an independent agent’s recommendations.
“AI-native advertising” can describe advertising designed for an AI interaction. “Agentic advertising” can describe agents performing work such as planning or managing a campaign. The terms concern different functions and are not used here as synonyms.
The strongest counterargument to a rapid budget shift is straightforward: a new interface does not guarantee economically valuable customers. If attributed conversions are not incremental, or acquisition costs remain unattractive, novelty will not sustain spending. As platforms automate campaign setup, the work brands may continue to value is deciding who to reach, what to offer and whether reported results represent worthwhile business.
