ChatGPT Ads don't work like Meta, TikTok, or Google. There's no keyword bidding, no audience builder, and one ad slot per response, full stop. What decides whether you get that slot isn't what you'd expect from a normal auction.
A brand asked us recently if they should be testing ChatGPT Ads. Then they asked how to bid on the right keywords.
There are no keywords.
That's usually where the conversation gets interesting, because it means most of what performance teams know from running Meta, TikTok, or Google doesn't transfer directly.
ChatGPT Ads don't target, they match
Every platform you already run builds an audience from demographics, interests, or keyword intent you declare up front.
ChatGPT Ads work backward from that. OpenAI's system reads the topic and context of the live conversation someone is having and matches a sponsored card to it, placed underneath the answer, clearly labeled, never blended into what the model actually says. There's no keyword list to build and no audience to define. The targeting is the conversation itself.
That's also why OpenAI is strict about what advertisers see back. You get aggregate performance data: impressions, clicks, and conversions you can tie to your own events through the Conversions API. You never see the conversation that triggered the ad. That's a deliberate wall, not a limitation you can work around.
Why timing matters more here than anywhere else
People don't open ChatGPT to browse. They open it mid task, mid decision, usually after they've already narrowed things down. That puts these ads later in the funnel than almost anything else you run.
The ad isn't introducing your brand. It's confirming a decision someone is already close to making.
Which means the real competition for that slot isn't only another advertiser. It's the model's own organic answer, sitting directly above it. If your brand isn't already showing up in the answer ChatGPT gives for free, the paid slot is working twice as hard to make up the difference.
This is the reason answer engine optimization has stopped being a side project for a lot of the teams we work with. The organic answer and the paid slot now reinforce or undercut each other in the same conversation.
What this means for your signal strategy
Two brands can run the exact same campaign type on ChatGPT Ads and get very different results, for the same reason two brands sending identical purchase volumes to Meta get different ROAS: one is feeding the optimization layer clean, matched signal, and the other isn't.
The platforms have changed. The underlying problem hasn't. Ad platform AI, whether it's Meta's delivery algorithm or OpenAI's ad matching, can only optimize toward what it can actually measure with confidence. Match rate still decides how much of that signal the platform can use.
Getting ready for ChatGPT Ads isn't really about ChatGPT. It's about whether your conversion signal is clean enough to compete for a slot that doesn't forgive a weak one.
Journify already sends that signal to ChatGPT Ads, through the same pixel and Conversions API setup already running your Meta, TikTok, and Google feeds. We don't bid and we don't manage the campaign. We make sure the signal reaching ChatGPT Ads is complete enough for it to optimize correctly.
If you want to see what that looks like on your own account, book a call.
