OpenAI advertising is rolling into ChatGPT. Reports say ads will appear at the end of answers and be labelled “Sponsored”. In early trials, advertisers are being charged on a pay-per-impression basis, not per click, and some have been asked to commit under $1 million for short trial runs. Free accounts and the new ChatGPT Go tier will see ads. Paid tiers such as ChatGPT Plus and ChatGPT Pro are not included for now.
Pay-per-impression changes the incentives. Advertisers pay for views rather than actions, which gives OpenAI a steadier revenue stream without depending on clicks. It also changes what the ad system needs to know. If ads sit inside an AI response, the prompt, the conversation context, and the account tier can all become useful signals. That blurs the line between the answer and the targeting.
The privacy questions come in two parts: data collection and profiling. If ad selection uses the text of your conversation, that text has to be inspected somewhere, whether briefly or for longer retention. Even if ads are sold on impressions, the system can still build account or device profiles to make those impressions more valuable. The trial details already suggest that free and low-cost tiers are the first targets, so account tier itself is a signal.
The practical checks are straightforward. Look at your account plan and privacy settings. Read the data retention and processing terms in OpenAI’s policy. Export or delete conversations you do not want kept around. If you want to avoid ads altogether, a paid tier that excludes them is the cleanest option while the trials are still running.
The real test is how much of this sits in policy and how much is backed by actual controls. I would want to see clear limits on what text is logged and how long it is kept, ad metrics that are aggregated rather than tied back to raw conversation data, and access controls that keep ad tech away from sensitive prompts. OpenAI could use anonymisation, short retention windows, in-memory targeting, or client-side hashing to reduce the risk. None of that means much unless it is stated clearly or shown in practice.
So the shape of it is simple enough. Pay-per-impression ads in ChatGPT are plausible and likely to land in free and low-cost tiers first. That brings in more revenue, but it also adds more privacy surface for users to manage. If you rely on ChatGPT for sensitive prompts, treat ad-supported tiers as less private until the handling is clear.
