OpenAI Ads Might Finally Be Graduating From Toy to Tool

We’ve been experimenting with OpenAI ads since they launched. Until recently, the experience was interesting mostly because of what the platform might become.

The controls were rudimentary. Reporting was limited. And while ChatGPT is obviously an intriguing place to reach people while they’re actively researching a problem, advertisers had very little ability to tell OpenAI which clicks were actually valuable.

That is starting to change.

OpenAI has now made conversion optimization available in open beta across its ads platform. It calls the model conversion-optimized CPC, or oCPC. Instead of simply optimizing for people likely to click an ad, advertisers can select a conversion event and ask OpenAI to favor clicks more likely to produce that outcome.

You still pay by the click. The conversion goal simply gives the system a better idea of which clicks are worth pursuing.

That may sound like table stakes to anyone accustomed to Google or Meta. It is. But table stakes are rather useful if you’re trying to play poker.

Early Results Are Encouraging

We moved some of the campaigns we manage to conversion optimization, and the early results have been noticeably better. We’re seeing higher-quality traffic at a CPC that makes economic sense.

It’s far too early to draw sweeping conclusions from that. But it’s the first time we’ve felt that OpenAI ads were behaving less like an interesting experiment and more like a performance marketing platform.

One Important Tracking Tip

If you’re testing OpenAI ads, there’s another thing we’d recommend doing before spending much money: manually add UTM parameters to your destination URLs.

Don’t assume GA4 will automatically understand where this traffic came from.

At minimum, we’d use something along these lines:

utm_source=chatgpt&utm_medium=cpc&utm_campaign=campaign_name&utm_content=ad_name

That gives GA4 enough information to identify the traffic source, campaign and individual ad.

This matters for conversions too. The UTM parameters don’t create the conversion in GA4; you still need your conversion measurement configured correctly. But they preserve the acquisition information so that conversions can be attributed back to your OpenAI campaigns rather than ending up in a murky reporting bucket with no obvious source.

Given how little reporting OpenAI currently provides, having an independent view in GA4 is especially valuable.

The Next Frontier: What Triggered the Ad?

Conversion optimization solves one important problem. The biggest remaining one, in our view, is transparency.

Search marketers are accustomed to seeing the terms people searched before clicking an ad. With ChatGPT, the equivalent information would be enormously useful: What kinds of questions or conversations caused OpenAI to decide that your ad was relevant?

Right now, advertisers are largely left to infer that through process of elimination.

OpenAI provides performance reporting by things like campaign, ad, product, country and device. What advertisers still don’t get is meaningful visibility into the queries or conversations that caused an ad to appear.

That makes optimization harder than it needs to be. You can test messages, landing pages and conversion goals, but you still can’t see the part that may ultimately make ChatGPT advertising uniquely interesting: the intent expressed in the conversation that preceded the ad.

OpenAI ads are getting better. Conversion optimization is a meaningful step. Giving advertisers some useful insight into why their ads appeared would be an even bigger one.