OpenAI is bringing ChatGPT ads to one of the app’s most used features. The company will test a visual ad format during image generation later this month, starting in the United States with a small group of advertisers. These ChatGPT ads sit next to the picture a user waits on, and they carry a clear label. For a product that reaches over a billion people a week, that is a real shift in how the assistant behaves.
What the new ChatGPT ads look like
Image generation leaves a user sitting and waiting, so OpenAI has decided to fill that pause. The format uses pictures instead of plain text links. An advertiser can show product ideas, real world uses, or the kind of feel a service promises. OpenAI says these ads stay labeled and stay apart from the image on screen.
The company also repeats an older promise: ads do not shape the answers the model gives. That line matters, because the trust gap between a search result and an assistant reply is wide. People treat a chatbot answer as advice, not as a listing. Any hint of paid influence inside the text itself would break that trust fast.
Why OpenAI wants ad money from 1.2 billion weekly users
ChatGPT now reaches 1.2 billion people every week, and only a slice of them pay for a plan. Running models at that size costs a fortune. The free tier drives growth, yet it has never paid its own bills. ChatGPT ads turn that huge free audience into money without asking anyone to subscribe.
The tracking that arrives with ChatGPT ads
The ad slot itself may be the smallest part of this news, since the bigger change sits behind it. OpenAI is also adding tools that let brands measure if an ad led to a sale or another action. Conversion data flows through Hightouch, Tealium and LiveRamp. Attribution partners now include AppsFlyer, Adjust, Branch, Triple Whale and Kochava, among others.
Anyone who works in digital marketing will recognize that list, because those firms tie clicks, app installs and purchases together across devices. A brand can then prove its spend worked. With them in place, ChatGPT ads plug into the same tracking machine that already measures the rest of the web. That push on measurement tells you OpenAI treats ChatGPT ads as a real business line, not a side experiment.
The privacy problem with ads inside a chat
A search box sees a few words at a time, while a chat sees far more context: plans, worries, purchases, problems. That gap matters once ads arrive, because the targeting signals get much richer than any keyword auction sees. OpenAI has not said which signals pick a placement during image generation. Until it does, users can only guess how much of a session shapes the ChatGPT ads they see.
How the safeguards work in theory
OpenAI says its safeguards keep ChatGPT ads away from sensitive or emotionally vulnerable chats. Brand safety work runs through DoubleVerify and Integral Ad Science. Both firms normally police where ads land on the open web, and that job gets harder inside a chat window. Ads here sit around live conversations, not around pages with known content.
The company adds that these outside partners never see private chats while they check the guardrails. Those promises rest on OpenAI’s own build, since no user can audit them from the outside. Still, the point stands: some conversations should never carry an ad at all. The harder question is who defines sensitive, and how a filter draws that line in real time.
What a VPN changes, and what it does not
Privacy tools help with this shift, though they help unevenly, and the split is worth spelling out. A VPN hides the IP address behind each request, which removes a cheap location signal and keeps an internet provider from logging which services someone uses. That matters while ChatGPT ads stay locked to one country. The protection stops at the account line, however.
Signing in ties every session to a profile, and no amount of IP masking undoes that link. Tracking kits inside advertisers’ own apps work the same way. They identify a device or a logged in customer, not a network route. So a VPN shrinks the metadata trail around the service, while the account layer stays exactly where it was.
What users can do while the test runs
The test covers the United States and a small group of brands for now, so most people will not see these ChatGPT ads yet. Anyone who wants less exposure can start with the data controls in their account. Turning off chat history helps, and so does keeping sensitive plans out of accounts tied to a real name. Pairing a VPN with a separate browser profile also keeps assistant use apart from daily browsing.
None of that blocks ads from arriving, but it does narrow how much context follows a user from one place to the next. That part stays under the user’s control. Habits formed now will age better than a scramble later, once more ad slots open up.
Final Thoughts
Ads inside an AI assistant were always coming, and the first format lands before most users have thought it through. A labeled picture beside a loading image looks harmless on its own. The tracking layer under it deserves more attention, because conversion data and attribution partners turn casual chat use into measured shopping behavior. ChatGPT ads will spread past image generation soon enough, so the habits people build now around accounts, data controls and network privacy will count for more with every new format.
