OpenAI tests visual ads in ChatGPT image generation

By VK

Published 6 Oct 2026, 11:28 UTC

Reporting from TechCrunch, OpenAI, The Decoder, The Verge

OpenAI will test visual ads in ChatGPT image generation in the US this month and is adding attribution, measurement and brand safety partners.

What it means for founders

A new paid channel, but measurable on your own stack. Support for AppsFlyer, Branch, Triple Whale and Northbeam means a consumer or app startup can judge ChatGPT spend against Meta and Google in the tools it already uses. Businesses can apply at ads.openai.com now; ask for incrementality testing, not just click attribution, before scaling budget.
Visual brands get first-mover room. Products that look good in context, such as furniture, fashion, travel and food, fit a slot that sits beside image prompts. Prepare lifestyle creative now rather than repurposing search copy.
Platform risk for apps built on ChatGPT. As OpenAI sells attention inside the chat, rivals may eventually be able to buy placement near the moments your ChatGPT app depends on. Watch whether OpenAI adds rules for ads near third-party apps.
Signal to watch: whether the US test starting this month expands to more advertisers or markets by year end, and whether early cost per acquisition holds once more bidders arrive.

The story

OpenAI said on Monday that it will start testing a visual ad format in ChatGPT later this month, placing image ads from an initial group of US advertisers on screen while the chatbot generates a picture for a user. In the same announcement, the company widened its ChatGPT Ads measurement stack and started brand safety pilots with outside verification firms.

How the new ChatGPT visual ads work

Until now, ads in ChatGPT were simple units: a company name, logo and link in a sponsored section under the answer, according to The Verge. The new format uses images meant to show a product in use or the experience it enables. OpenAI says the ads will be labeled, kept apart from the image being generated, and will not affect what ChatGPT answers. Paid Plus, Pro and Enterprise subscribers do not see ads.

Image generation is a natural first slot. The user is already waiting and already thinking visually, so a product picture competes with nothing but a loading state.

Measurement and brand safety

The bigger change for advertisers may be measurement. OpenAI now lists:

Conversion data sent in through Hightouch, Tealium and LiveRamp.
Attribution and Conversions API support from ten vendors, among them AppsFlyer, Adjust, Branch, Singular, Northbeam and Kochava.
Incrementality tests using geo-based experiments with Haus, Measured and WorkMagic, which OpenAI calls early work.
Brand suitability pilots with DoubleVerify and IAS (Integral Ad Science), run in a test environment without access to private conversations, plus Negative Phrases for qualifying advertisers.

OpenAI also cited partner results, including a DV Rockerbox finding that WeightWatchers saw its ChatGPT attributed cost per acquisition come in 15.3% under its blended benchmark for paid search. These are vendor-reported case studies, not audited figures.

The context is scale and money. OpenAI puts weekly ChatGPT reach at 1.2 billion people. The Decoder reports that ads, live since February and now running in over 40 countries, have reached a $1 billion annualized run rate, and that the company is targeting $100 billion a year from advertising by 2030.

What we don't know yet

Pricing, minimum spend and whether the visual format is bought by auction or by negotiated deal.
When the test leaves the US and opens beyond the first advertisers.
How targeting works inside an image request, and what signals from the prompt are used.
Whether the format spreads beyond image generation, which OpenAI hints at by calling it the initial placement.