ChatGPT Ads Get a Makeover: OpenAI Tests New Formats with CTAs and E-Commerce Features

OpenAI has spent months building the plumbing for ChatGPT ads. Now, it’s working on how they actually look.

A week after the platform made it easier for e-commerce companies to bulk create their shopping ads, it’s now giving advertisers more options to customize their ads around ChatGPT conversations.

The platform is testing a new ad format — an iteration on its existing standard unit — with a larger image and an optional call-to-action button that advertisers can personalize, according to mockups viewed by Digiday. OpenAI itself confirmed this, sending an email to advertisers stating that the platform is “beginning an early test” of a new ad experience on a small subset of ads. Those ads include dynamic CTAs, such as “shop now”, “book now”, “sign up” and “learn more”.

It’s also introducing a dedicated e-commerce format that can be displayed in either portrait or landscape orientation and pulls in shopping data including price and customer reviews. The portrait version is designed to stack, potentially allowing three or four ads to sit side-by-side, opening the door to a carousel-style placement.

Until now, advertisers had a single, simple format: a headline, a short description, an image and a link. The new formats, while minor updates to that design, give marketers more control over how their brands appear and, for the first time, a direct response element within the ad unit itself.

But it won’t stop there. Bigger changes to the actual ads are on the way. OpenAI ad execs have made that clear to marketers publicly and privately.

“Creative variation has been a real key to success,” said Benji Shomair, OpenAI’s vp of monetization, during a recent press roundtable.

His comment speaks to some of the deeper conversations happening around the embryonic stage of ChatGPT ads. Chiefly, that it is a high-intent, conversational environment where the context of a query — researching running shoes versus holiday trips — shapes what an ad needs to do. That makes creative performance more variable than in news feed or a search box, and more sensitive to execution. With enough live data now flowing through the platform, OpenAI appears to be amassing the breadth of signal needed to act on that.

What does that mean in practice? A self-serve ads manager means advertisers no longer need a direct sales relationship with OpenAI to run campaigns. It opens the platform to the long tail of advertisers — the mid-market brands, the performance shops and the agencies testing on behalf of clients who aren’t yet ready to commit big sums — and dramatically increases the volume of creative and spend data flowing through the system. That data is raw material for knowing what formats work, in which contexts and for what objectives.

Adthena’s CMO Ashley Fletcher, for example, said that his team has more U.S. folks wanting to join via the ads manager, but they haven’t been able to get their accounts up and live yet. “It’s the same for Adthena, I am still waiting for our approval for our account and I see a message that reads: ‘We are reviewing your information. Due to a high volume of signups, verification may take some time’.”

It may get longer before it gets shorter on OpenAI’s current roadmap: audience targeting — i.e. allowing advertisers to retarget existing customers or exclude them entirely, with lookalike audiences likely to follow — is in gated rollout. Outcome-based optimization, the feature that would unlock serious performance spend by letting advertisers target against CPAs rather than impressions and clicks, is in development but without a public timeline. New ad formats, beyond those already announced, are being drawn up. More granular serving status visibility and campaign start and end data controls are also on the way.

All of that is on top of what’s already been added in short order: conversion tracking, role-based account access, spend reporting at the ad level, daily budget controls and the self-serve manager itself. The pace of iteration is, by any measure, aggressive.

For advertisers, it’s like watching an ad offering being built in real time. And while the demand is clearly there, bigger budgets are unlikely to materialize until more advanced features are built in.

One exec, who currently has advertisers participating both directly and via OpenAI partners, said there are two big wish list items they want to see: scale and the ability to optimize toward business outcomes, such as product sales or lead generation.

“Business objectives are what’s going to actually drive business outcomes,” they said. “They’re not [yet] giving advertisers the ability to specifically target ROAS or cost per acquisition, efficiency amount. I’ve heard from many other clients [not yet in the pilot] that they want to do ChatGPT ads, but they’re going to wait until they are able to target business outcomes.”

All of this further confirms OpenAI’s urgency to want to generate as much revenue as soon as possible from its ad business to support the platform’s bigger goal. The AI company is reportedly expected to confidentially file its IPO prospectus as early as Friday, according to the Wall Street Journal, with a view to go public as early as September.

OpenAI did not respond to Digiday’s request for comment.