OpenAI said in an announcement on Monday, October 5, that it will begin testing ChatGPT visual ads in the US later this month, initially when users generate images. The format puts images of sponsored products and services on the user’s screen.
Until now, ChatGPT’s ads have reportedly shown only a company’s name, logo and a link to its product, in a ‘sponsored’ section beneath the chat. A sponsored image is a bigger step: it is more prominent than a logo and link, and it appears at a moment when the user is making something visual.
What did OpenAI announce about ChatGPT visual ads?
OpenAI announced a visual ad format for ChatGPT, with a US test starting later this month and image generation as the first place users will meet it. The announcement’s qualifier was ‘initially’. That marks image generation as the starting point, but the available accounts do not say where the format might go next, so we treat any extension as an open question.
Two accounts of the announcement differ in small ways. One frames it as a launch, the other as a test beginning later this month; we describe a test, since that matches a start date still ahead. On placement, one says the ads appear when you generate images, while the other says they show up after the results appear. We say only that they are tied to image generation and do not describe the layout.
OpenAI’s own announcement text was not available to us, so these details rest on how it was relayed.
How do the new ads differ from ChatGPT’s first ads?
The main difference is the content: images of sponsored products and services now join, or replace, a name, logo and link. The table sets the two formats side by side. The earlier format and its February start come from a single account, so treat them as reported.
| Earlier ads | Visual ads | |
|---|---|---|
| Content | A company’s name, logo and a link to its product (reported) | Images of sponsored products and services |
| Where users see them | A ‘sponsored’ section beneath the chat (reported) | Tied to image generation; exact layout not confirmed |
| Start | February (reported) | Test begins later this month |
| Scope | Not stated in the available material | US only for now |
How did ChatGPT advertising get here?
ChatGPT advertising has moved in stages: ads reportedly arrived in February, and this announcement adds images. BriefFlash earlier reported that OpenAI said ChatGPT ads hit a $1 billion run rate as self-serve access went global. We rely on that headline only and add no figures beyond it.
On the subscription side, BriefFlash also covered how OpenAI paused $200 ChatGPT Pro sign-ups as Astra demand overwhelmed its systems. Nothing in the available material links that report to the ad test.
Why does a sponsored image in image generation matter?
It matters because an ad inside a creative moment competes for attention differently from a link under a reply. This is BriefFlash analysis, not reporting. The premise: a user opening image generation has just stated a visual intent. The inference: a related sponsored image there is likely to be more noticeable than a logo beneath a chat answer. We hold that view with moderate confidence, because placement and labelling are not confirmed.
For advertisers, the format is a visual product slot in a context where users are looking at pictures. Whether it is worth buying depends on pricing and targeting, and neither has been stated. For people who generate images, the practical question is whether the ads stay clearly apart from their results.
For makers of rival assistants, this is a live test of how far commercial placement can go before it changes the product. We would not draw conclusions about their plans from it, and the available accounts do not compare the format with ads in other assistants or image tools.
What do we not know yet about the ChatGPT ad test?
Several practical questions remain open, and none was answered in the accounts available to us:
- How the ads are labelled and separated from generated images.
- Whether the content of a user’s image prompt influences which ads appear.
- Which ChatGPT plans or tiers will see the ads.
- How advertisers buy the format, and at what price.
- When, or whether, the test reaches countries beyond the US.
- Whether the format extends beyond image generation.
No advertiser names or revenue figures for the new format have been given.
What happens next?
The US test is due to begin later this month. Tests can change or be dropped before any broader rollout, and a one-country test is not a global rollout.
The first things to watch are OpenAI’s own documentation on labelling and prompt use, and any statement on tiers and markets. US users who generate images will be the first to show what the ads actually look like.
Frequently asked questions
When will ChatGPT visual ads start appearing?
OpenAI said on Monday, October 5, that it will begin testing visual ads in the US later this month. It is a test rather than a full launch. OpenAI has not said how many users will see the ads, or when the format might reach other countries.
Will I see ads every time I generate an image?
That is not yet clear. OpenAI has said the ads will initially appear when users generate images, but the available accounts do not say how often they show, which ChatGPT plans see them, or whether the content of an image prompt affects them.
Are the new ChatGPT ads available outside the US?
Not at the moment. The test is described as US only for now, and no timeline for other markets has been given. Because it is a test, the format could also change or be dropped before any wider rollout.
How are visual ads different from ChatGPT’s earlier ads?
The earlier ads reportedly showed a company’s name, logo and a link in a ‘sponsored’ section beneath the chat. The new format shows images of sponsored products and services, tied to image generation. The description of the earlier format comes from a single report.
Does OpenAI use my image prompts to choose ads?
Unknown. The available accounts do not say whether the content of an image prompt influences which ads appear. Until OpenAI publishes details on targeting and privacy, the question stays open, and we are not assuming either answer.
