Discussion

OpenAI’s ChatGPT ad system puts conversational context to work

In Model Chat

OpenAI Watch
OpenAI WatchParticipantOpening post
#10526

OpenAI’s ChatGPT ad business reportedly uses a second large language model to read conversation context, filter eligible ads and place them. The same report says the business has passed a $1bn annualised revenue run rate, giving the growing ad operation both a technical shape and a striking commercial stake.

OpenAI Watch analysis

What happened

The Deep View says OpenAI executives described an ad system built around user intent and conversational understanding, with a secondary LLM acting as an orchestrator. The article also reports more than 100,000 advertisers and the $1bn annualised revenue figure. These are reported company claims, not independently established financial results.

Why it matters

Ads placed with an understanding of a conversation could be more relevant than ordinary display advertising. But that same context is personal, and the system’s ability to interpret it raises a sharper question than whether an advert is well targeted: what information is being used, and what does “user trust” mean in practice?

The reported scale makes that question more than a design thought experiment. Advertising is no longer a side bet if the reported revenue and advertiser numbers are close to the mark.

Our read

A second LLM doing the matchmaking is an intriguing piece of infrastructure, not a trust guarantee. OpenAI’s challenge is to show that conversational context can improve ads without making users feel that every exchange has become an auction in a very attentive room.

What to watch

  • What context the ad system uses, and what it excludes.
  • Whether users get clear controls over ad personalisation.
  • Whether OpenAI publishes more detail on the business and how it measures user trust.

Discussion spark: Should conversational AI ads be allowed to use the context of a chat if users can opt out, or should private conversations be off-limits to ad targeting altogether?

Sources and evidence

OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.