ChatGPT Ads Add Sponsored Agents: Testing in the US with Select Advertisers Only

On September 16, 2026, OpenAI's official announcement 'Reimagining advertising with AI' rolled out four ChatGPT ads updates at once: the in-testing 'Sponsored Agents', the ability to create an ad by simply writing a few prompts in ChatGPT Work, new AI creative assistance and text customization in Ads Manager, and integrations with HubSpot and Shopify. This article focuses on the one that turns an ad inside ChatGPT into something you can open and talk to.

This article was fact-checked on September 18, 2026, drawing on the announcement itself, OpenAI's page describing the current state of the ads test, and the advertiser introduction page; all three URLs were re-read on the day of the check. This site has never logged into the ChatGPT ads dashboard, never opened a Sponsored Agent conversation, and never created or bought an ad, so it offers no purchasing or ad-placement advice. Everything here about how the feature works and who it is open to is OpenAI's own description.

Ads You Can Open and Talk To: What Was Announced on September 16

OpenAI describes Sponsored Agents in one sentence: after clicking an ad in ChatGPT, a user can start a conversation with an agent sponsored by a business. The verb OpenAI uses for this is 'testing'; the last sentence of that same section narrows the scope further, stating that this test is currently running only with select advertisers in the United States — 'testing', 'select advertisers', and 'the United States' are three conditions in that sentence, and the description is incomplete without any one of them.

OpenAI draws two boundaries around this conversation: it is kept separate from ChatGPT's usual answers, and separate from the conversation the user was already having. OpenAI writes that after seeing a relevant ad, the user can choose for themselves whether to start this clearly labeled conversation, with a link to the business's own website when needed.

The other three announcements are tools for the advertiser's side: the announcement's opening line says creating an ad takes just a few prompts in ChatGPT Work, while the paragraph on creating, updating, and analyzing campaigns states that this happens in ChatGPT through the Ads Manager plugin. Ads Manager also gained an AI tool that suggests copy and images based on the landing page and campaign objective. Ads also now connect to HubSpot and Shopify, which OpenAI calls its first CRM partner and first ecommerce partner. These are tools for how advertisers set things up, not the focus of this article — that topic, including the four-layer account structure, bidding methods, and region targeting, is covered in another article on this site.

The Boundaries OpenAI Draws, and This Round of Changes at a Glance

The advertiser introduction page spells the boundaries out more fully in three lines: ads are clearly identified, ads stay separate from ChatGPT's responses, and people control how their data is used for ads. These three lines are principles for the ChatGPT ads system as a whole, not commitments specific to Sponsored Agents.

The announcement also states that this update does not change its existing ads principles, and that protecting people's trust in ChatGPT remains its top priority. This is OpenAI's own statement, and this site cannot verify it independently.

For HubSpot, the announcement says a connection is available 'starting today', but that line does not say which regions it covers. Shopify comes in two steps: US-based merchants could use the ChatGPT Ads app in the Shopify app store starting the day of the announcement, and it is only 'starting September 23' that the app becomes available internationally — that sentence itself prints no year, and this article reads it as September 23, 2026 based on the announcement's own September 16, 2026 dateline; that is this article's own editorial reading, not something OpenAI wrote. And what opens up is 'markets where ChatGPT Ads are available', not every place ChatGPT is used.

Looked at together, the items announced this time differ in who they are open to, which region, and their current status; the table below summarizes the state as read on the day of the check.

Who Sees Ads: Your Plan Sets the Threshold, Your Region Sets Whether You See Any

OpenAI's page on the current state of the ads test states this plainly, and that text is part of the page's original post from February 9, 2026, describing the scope of this test: only logged-in adult users on the Free or Go plan see ads; the five plans Plus, Pro, Business, Enterprise, and Education have no ads at all. The same page also gives two ways to avoid seeing ads: upgrading to Plus or Pro, or staying on the Free plan and trading fewer daily free messages for ads turned off. The page does not otherwise explain what kind of plan Go is, beyond grouping it with Free on the side that sees ads.

The same page stacks several updates, with the newest at the top, dated August 11, 2026: ChatGPT ads have now launched in the United Kingdom, Mexico, Brazil, Japan, and South Korea. Before that, an update from May 7 says that in the coming weeks OpenAI 'plans' to expand the ads pilot to those same five markets. Earlier still, the March 26 update says that in the coming weeks it would begin with pilots in Canada, Australia, and New Zealand; the earliest version, from February 9, 2026, is where the US test began. Stacked together, these dates show the covered regions being added one at a time, not opened all at once from the start.

None of the three pages checked for this article mentions Taiwan anywhere, and none carries an exclusion list either. Based on what was there as of the September 18, 2026 check, Taiwan is not among the markets OpenAI lists as open — but that is not the same as OpenAI saying it won't open there; the page itself says it is continuing to expand to more markets. Update, October 3: on September 23 OpenAI announced that ads were beginning to roll out gradually in Taiwan (see this report); that announcement does not mention Sponsored Agents.

Ads That Rewrite Themselves: Customization and Automatic Translation

The announcement adds a feature advertisers can choose to turn on for themselves: AI text customization. OpenAI writes that once enabled, the system adjusts the advertiser's existing headline and description text to fit the conversation, and automatically translates the ad copy into the user's preferred language. In other words, the ad text a user sees may be a version the system has rewritten and translated; the announcement does not say whether advertisers can see or review these rewrites and translations beforehand.

OpenAI attaches one condition: the original text uses the phrase 'opt into', meaning advertisers must turn it on themselves. How translation quality is checked, which languages are supported, and whether users can see the original wording are not addressed in the announcement or the other two pages, and this article does not answer on OpenAI's behalf.

Another AI feature in Ads Manager is creative suggestions: the system suggests copy and images based on the landing page URL and campaign objective the advertiser enters. OpenAI states that what advertisers see are 'suggestions', which they can review and edit before deciding whether to add them to a campaign — OpenAI describes this step as remaining under the advertiser's control, not the system deciding automatically on the advertiser's behalf.

What OpenAI Doesn't Say, and Three Things Readers Can Check for Themselves

Reading the three pages together, there are things OpenAI does not say: the announcement does not state when this test ends, how many advertisers are involved, or how an advertiser would join this particular test; none of the three pages says whether Sponsored Agent conversations are used to train models or how long they are retained. On what advertisers can see, the status page has one passage written about ChatGPT ads generally: advertisers do not get a user's conversations, chat history, memories, or personal details, only aggregate performance numbers such as views and clicks — and that passage does not specifically mention Sponsored Agent conversations. Pricing does not appear anywhere either; this announcement lists no change to any fee or billing method.

In its March update, OpenAI made one claim about results in its own words: it says it is seeing no impact on user-trust metrics, a low ad-dismissal rate, and ongoing improvement in relevance; the page attaches no methodology, sample size, or time period, so this is OpenAI's own observation, not an externally verified result. The advertiser page separately names four brands — Newegg, Best Buy, Lowe's, and VistaPrint — saying advertisers are already seeing the benefits of AI-native advertising, but it publishes no figures for any of the four; the four names are listed participants, not evidence you can point to.

There are three things readers can check for themselves: if a Sponsored Agent conversation appears, look at how the clear labeling OpenAI describes is actually worded; check whether your own subscription plan is Free or Go; and remember that this site has never seen what Sponsored Agents looks like once opened — every sentence in this article comes from OpenAI's own pages, not hands-on experience.

2026 AI News Roundup: Highlights and Daily Life Applications from January to September

ChatGPT Ads Open Up to Self-Serve Buying: The Tools, Billing, and Targeting from OpenAI's May Announcement

• AI tools
• Apps & software
• AI news