OpenAI says ChatGPT Ads reached a $1 billion annualized revenue run rate less than 200 days after launch. Digiday explains the important caveat: an annualized run rate projects the current monthly pace across a year. It is not a billion dollars already earned. Still, it is enough to show that ads in ChatGPT are becoming a real business.
The ad system now looks familiar. OpenAI says tens of thousands of advertisers use it, ads are available in more than 40 countries, and self-service buying is opening across India, Europe, the Middle East and North Africa. Cost-per-click and outcome-optimized bidding now account for most campaigns. Product feeds, custom audiences, a tracking pixel and a Conversions API make this a measurement platform, not a handful of sponsorships.
What is new is the conversation itself. OpenAI's current help page says its ad system considers the context and intent of the current chat. When personalization is enabled, it may also use signals from past chats and memory. Among eligible ads, both relevance and advertiser bids help decide what appears first.
There is an important privacy distinction here. OpenAI says advertisers do not receive people's chats, history, memories or personal details. But the company's own ad system can still use conversation context to choose a paid placement. Turning personalization off stops past chats and memory from informing ads; it does not stop the current thread from doing so. “Private from advertisers” is not the same as “unused for advertising.”
OpenAI promises a wall between the answer and the ad. Its advertising principles say ads are labeled, appear separately, and do not influence ChatGPT's answers. The help page says the ad system runs separately from the chat model and advertisers cannot shape, rank or alter responses. OpenAI's placement policy also excludes ads from sensitive contexts such as personal health, mental health and politics.
Those are meaningful rules, but today's evidence is still mostly OpenAI describing its own system. The milestone post says its measures of trust, relevance and overall experience “remained strong,” without publishing the values, method or an external audit. Answer independence is therefore a clear company commitment, not yet a public measurement result.
What changes now is the incentive. OpenAI says it plans more formats, objectives, buying tools and measurement, plus “more native” ways for businesses to interact with people inside ChatGPT. A conversation often contains a goal, constraints and the alternatives someone is weighing. That can make an ad more useful than one matched to a search keyword. It also gives the ad system a closer view of the decision someone is making.
The same audit problem applies to any agent that decides what people see. Citations can reveal selected evidence; they cannot show every candidate source or ranking system that shaped the selection. Paid placement therefore needs its own visible label and a separate trail from evidence.
Watch two things: whether OpenAI publishes evidence that answer independence survives more interactive ad formats, and whether people can understand and control when the current chat, past chats and memory inform an ad. The revenue number shows traction. Those boundaries will decide whether ads can scale without changing what people think an answer is.
Source graph: Semble source collection