OpenAI pushes visual ads into ChatGPT this month

The company said Monday it will begin testing a new ad format in the United States later in October, placing product and service images directly on screen. The ads will appear "initially" when users generate images through chat and will stay separate from the generated output, without influencing the model's responses.

What changed from the old format

Until now, ChatGPT ads amounted to a company name, logo and link in a "Sponsored" section below the conversation — a format launched in February. The new version adds a visual layer: actual images, not just text and a logo. In both cases the ads are kept apart from model-generated content, but the shift to pictures changes the user experience in a way that's hard to ignore.

Where and when

The test starts in the US later this month, focused first on image generation, not regular text chats. Plus, Pro and Enterprise subscribers won't see ads at all, turning the feature into another lever for subscription sales. OpenAI didn't say how many advertisers are in the first cohort, only that it's an "initial group."

Guardrails and filtering

The company says safeguards decide when a conversation is "suitable" for ads, aiming to keep them away from emotionally vulnerable moments, sensitive contexts or other inappropriate situations. No precise criteria or examples were published, so it's hard to gauge how well the filtering works in practice.

What's next

OpenAI notes it plans to test additional formats down the line, signaling the current design is just a first step in monetizing the visual interface. Since the model itself isn't affected by ads, at least per the company's statement, the real question isn't whether answers change — it's how fast users learn to tune out visual noise they never asked for.