You ask ChatGPT to imagine your living room repainted in sage green, with a large sofa and plants. While the image is being generated, another appears beside it: a very real sofa, from a real brand, with a "Learn more" button.
This is the new advertising format OpenAI announced on Monday.
What was announced
OpenAI will test image ads in ChatGPT, initially during image creation. The test begins later this month, in the United States, with an initial group of advertisers. The company has announced nothing for the UK at this stage.
The idea, according to OpenAI, is to help people imagine how a product or service could fit into their lives: showing an object in context, a use, an experience.
The company stresses three promises. Ads will be clearly labelled. They will remain separate from the image you are creating. And, it says, advertising does not influence ChatGPT's answers.
This is not the first advertising in ChatGPT
Since February, OpenAI has already been testing ads in the United States, in the form of messages placed at the bottom of conversations, linked to the topic being discussed. They only concern users of the free version and the cheapest plan. Subscribers to the more expensive paid tiers do not see them.
The new part is the image. And the timing is no accident: when you ask it to imagine a living room, an outfit or a trip, you are stating very precisely what you want. For an advertiser, it is hard to dream of anything better.
What really changes
On the internet, an ad appears next to a page or a video. In ChatGPT, it appears next to a conversation. And a conversation says far more than a page: your plans, your tastes, sometimes your worries.
OpenAI acknowledges as much in its announcement, quoting a partner: in ChatGPT, it is the conversation itself that defines the context of the ad. The company also announced new tools to help brands measure the effect of their ads, and to prevent them from appearing next to conversations they would consider inappropriate.
OpenAI says it keeps your conversations private from advertisers. That is an important promise. It deserves to be watched closely, because it is what makes all the difference.
For its ads already being tested, OpenAI lets you hide an ad, give feedback, see why it was shown to you and adjust ad personalisation. If these ads ever reach you, these are the first settings to go and look at: they tell you what the service has inferred from your exchanges.
Why OpenAI is doing it
The answer comes down to one figure: more than a billion people use ChatGPT every week, and the vast majority pay nothing. Yet every answer has a very real cost, as we explained in our article on why ChatGPT is free.
Advertising is the oldest way of funding a free service. It is how television, radio, search engines and social networks survive. ChatGPT is now entering that model too.
What to take away
An assistant that helps you imagine your living room is useful. An assistant that at the same moment shows you a brand's sofa may be useful too. But it is no longer quite the same relationship.
The real question is not advertising itself. It is what it does to the trust we place in a tool to which we entrust so much. We devote our reflection this evening to that.