OpenAI plans to begin testing visual ads during image generation in ChatGPT in the United States later in October, adding product and service imagery to its advertising offering. The company announced the format on October 5 and said the initial test would involve a limited group of advertisers.
The placement comes with a distinction that matters for people using ChatGPT to create images: OpenAI says the ads will be clearly labeled and kept separate from the image being generated. It also says advertising will not influence ChatGPT’s answers. Those are the company’s stated boundaries for the format, rather than findings from an independent assessment of the test. OpenAI’s announcement
Alongside the format, OpenAI is expanding advertising measurement partnerships and developing brand suitability evaluation pilots. The announcement therefore covers both what users will see and how advertisers will assess the surrounding environment and campaign results.
How the visual ads will work
The initial setting is ChatGPT’s image generation workflow. OpenAI describes ads that use images to show product inspiration, product usage or experiences associated with a service. Its stated aim is to help people understand an offering and consider how it could fit into their lives.
That description supports a visual advertising placement alongside the creative process. It does not establish that a sponsor’s product will be inserted into the image a user requests. OpenAI explicitly says the advertisement will remain separate from the image being created, while TechCrunch describes the ads as appearing alongside image generation results. TechCrunch’s reporting
According to The Verge, the earlier ChatGPT ad format displayed a company’s name, logo and product link in a sponsored section beneath the chat. The new format adds imagery to that advertising proposition. The Verge’s reporting
For users evaluating the change, the central issue is the boundary between the requested output and the sponsored placement. Labeling identifies the commercial content, while separation preserves a distinction between an advertisement and the generated image. OpenAI’s additional promise that ads do not influence answers addresses a related question: whether advertising changes the assistant’s response. The announcement states that principle but does not provide an independent evaluation demonstrating it.
A limited US test, with subscription distinctions
OpenAI says testing will begin later in October in the US with an initial group of advertisers. The evidence supports a scheduled test, rather than general availability of the format or a completed international rollout. It does not establish a precise launch day, how frequently the visual ads will appear or the eventual geographic scope.
The Verge reports that ads will not appear for subscribers to ChatGPT Plus, Pro or Enterprise. The report identifies those subscription exclusions, but the forthcoming visual-ad test has not yet established its behavior in practice.
The visual format extends an advertising business already aimed at free and lower-cost access. The Verge dates ChatGPT’s initial ad introduction to February, while Thurrott reports that OpenAI rolled out advertising to Free and Go users. TechCrunch reports an expansion to India, and Thurrott describes an expansion to 31 European countries. Those reports concern the existing advertising offering; they do not make the new visual format available in those markets. Thurrott’s reporting
Measurement expands beyond the new imagery
The advertiser-facing changes extend across ChatGPT Ads. According to TechCrunch, OpenAI’s expanded click attribution partner ecosystem includes AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin. The report also names Fospha, Measured and INCRMNTAL as partners for advanced, full-funnel measurement.
OpenAI is additionally working with Haus, Measured and WorkMagic to help advertisers explore experiments based on geography. These efforts sit alongside the attribution partnerships as ways to examine advertising impact. The evidence identifies the partners and types of measurement being pursued, but does not supply campaign results or demonstrate that visual ads produce better outcomes than the earlier format.
For technical decision-makers, that distinction limits what can be concluded from the announcement. A wider measurement ecosystem gives advertisers more named routes for evaluating campaigns; it is not itself evidence of improved campaign performance. Nor does the partner list establish particular integration steps, data schemas or implementation requirements. The reporting does not specify those details.
Guardrails and brand suitability remain separate questions
OpenAI is developing brand suitability evaluation pilots with DoubleVerify and Integral Ad Science, according to TechCrunch. The pilots are intended to help advertisers understand the ChatGPT advertising environment. They are described as work in development, rather than a completed evaluation or a certification of every possible placement.
Separately, The Verge and Thurrott report that OpenAI has placement guardrails intended to keep ads out of emotionally vulnerable, sensitive or otherwise unsuitable contexts. This is a statement about the intended behavior of the placement system. The evidence does not quantify how reliably those guardrails work or provide results from the brand suitability pilots.
The user-facing safeguards and advertiser evaluations address different concerns. Keeping sponsored material separate from a generated image helps identify what the user requested and what an advertiser paid to show. Assessing conversational suitability concerns whether a commercial placement belongs in a particular context at all. Both will matter when evaluating the test, but neither can be judged solely from the announcement.
The immediate development is a defined experiment: visual advertising during image generation, starting with US users and an initial advertiser group later in October. OpenAI has specified labeling, separation and answer independence as its commitments. The remaining questions concern the test’s implementation and results, including placement frequency, the effectiveness of contextual guardrails and measured advertiser outcomes. The announcement does not yet answer them.
