OpenAI has spent its first four months laying the groundwork for its ad business. Now it’s turning its attention to the various formats that will power it.
Three job listings on OpenAI’s careers page point to text, image, video, native, conversational and interactive ad formats in the works.
The company is looking for an ad formats software engineer with at least seven years experience — a role described as a “foundational role” across the full stack — that sits within the monetization team. The role is responsible for building the infrastructure and tooling, as well as defining how ads are structured, rendered and delivered across different surfaces, platforms and media types, per the job spec.
The other two ad format software engineering roles are similar, but require at least four years experience and focus on iOS and Android experiences, respectively.
“The roles they are recruiting for need to tackle the thorny questions of the marketing ecosystem: attribution, brand safety and device modelling,” said Rob Webster, CEO of TAU Marketing Solutions. “Setting this up is not going to be easy. This is a different type of situation because no one knows what the right way to run ads in OpenAI is right now.”
What’s interesting about these roles is that privacy and safety are baked into what the engineers build. All three roles list upholding “the highest levels of safety, privacy, fairness and policy compliance across all ad rendering and delivery systems”, while the iOS and Android engineers specifically list building “policy-aware UX patterns” and “format validation”.
That hardline stance gets at one of the bigger tests facing OpenAI’s ad ambitions. The company has been explicit that the whole push hinges on not breaking trust in ChatGPT itself. The product only works if people don’t feel like they’re being sold to mid-conversation, so any misstep on privacy or intrusive ad experiences threatens the thing that makes the chatbot valuable to begin with. Building the guardrails in at the engineering stage, rather than retrofitting them later, could be the difference between ads that coexist with the product and ads that erode it.
“The real challenge is what advertising looks like in an advice system,” said Andrew Frank, research vp and analyst at Gartner. “How do you reconcile what we call the dual-alignment problem? Are you optimizing for user trust or advertiser value? Those objectives are often incompatible. Solving that problem — if it’s solvable at all — will take more than innovation around ad formats.”
Bigger picture
All three roles are exclusively based in San Francisco (with the offer of relocation assistance if needed), and the compensation package for each is $230,000 to $385,000 plus equity.
Since launch, the ad format has stayed simple: a headline, short description, an image and a link. Changes to it have been minor and iterative as OpenAI has tried to settle on the right balance of text and image so as not to detract from how people use the chatbot. Last month, Digiday viewed mockups of an iteration of OpenAI’s existing standard unit, which included a larger image and an optional call-to-action button that advertisers could personalize. Since then, only a minor tweak has arrived to show a slightly narrower format — previously it was 480px, now it’s 440px.
As recently as last week, it looked like OpenAI would sit with this format for a while. Ads boss David Dugan was vague when asked about what’s next. But more formats were always coming. Even OpenAI’s vp of monetization Benji Shomair said as much at a press roundtable in May, when he said “creative variation has been a real key to success”.
According to eMarketer’s principal analyst, AI in marketing and commerce, Nate Elliott, it’s about time OpenAI started testing creative formats. They’ve been one of the slower threads in OpenAI’s fast-moving ad story. The reason, Dugan told Digiday last week, is the team has been heads-down on getting the basics right, with the roadmap shaped by feedback from test advertisers helping to define what comes next.
“So far [OpenAI] has just been plowing ahead with a global launch of the one single format and placement they’ve ever tested, without any idea if that actually works best for advertisers or for users,” said Elliott. “The reality is, different types of ads will perform better for different audiences, different advertiser goals and different types of conversations. It’s vital that they test and learn.”
OpenAI did not respond to Digiday’s request for comment.
More in Marketing
While Instagram, TikTok and YouTube remain dominant social platforms for influencer partnerships, creators are increasingly being used across other media channels, including retail media, CTV and display ads.
[
How 3 brands are using TikTok Shop to expand abroad ](https://digiday.com/marketing/how-3-brands-are-using-tiktok-shop-to-expand-abroad/)
As these brands venture into the U.K., Europe and Japan, they say that the key to a global approach to TikTok Shop involves tailored marketing campaigns.
[
How beauty brands are popping up around World Cup festivities ](https://digiday.com/marketing/how-beauty-brands-are-popping-up-around-world-cup-festivities/)
Beauty and skin care in particular have taken a bigger spotlight this year, exemplifying just how closely beauty and sports have become intertwined.