Three days before the first ads were set to appear inside ChatGPT, the marketers testing them received a brief update: the launch date had shifted from Feb. 6 to Feb. 9. No explanation came with it – just a scheduling change from a company still working out how openly it wants to talk about becoming an ads business.
The same day OpenAI notified marketers about that shift, rival Anthropic began running a brand marketing campaign during its Super Bowl slot, taking aim at the idea of chatbots selling ad space — an unsubtle critique. For some marketers, OpenAI’s delay looked less like a technical adjustment and more like optics management: let the noise pass, then proceed.
Now, a month after OpenAI announced said ads were coming on Jan. 16, that abrupt switch reads as instructive. Each step forward is not only about building an ad business but about shaping how that business is understood by users from the outlet. It’s a communication exercise as much as a monetization strategy.
Get it wrong and the downside won’t be limited to ad revenue. It risks eroding the trust of its 800 million user base and, by extension, usage.
“OpenAI is being very careful because it [introducing advertising] is a fine line in a conversational experience, and it does not take much to upset users if ads feel distracting or change the experience,” said Wayvia CEO Anthony Ferry.
The guidelines OpenAI sent to test advertisers ahead of last week’s pilot launch reinforce that. The document, which Digiday has seen, runs nearly two pages of dos and don’ts, urging advertisers to act as “messengers of user value” when discussing the ads. It also asks them to submit any public communication that references the company or the revenue opportunity for review, to ensure alignment with its messaging that ads will not undermine the broader in-app experience.
“As far as we know, they haven’t been laying the groundwork for a long time,” said one exec, whose clients have been approached. “I don’t believe they were reaching out [to brands] more than three weeks ago, I think they only started reaching out to them recently [a week or so prior to launch].”
Normally, platforms worry about that after advertising is established, not before — and even then, rarely with this degree of prescriptive framing. But those ads businesses were not being built on what is, arguably, one of the most detailed records of private human thought ever assembled, at a moment when public wariness over companies monetizing it is sharper than in earlier platform cycles.
That leaves little room for mistakes, which helps explain why Fidji Simo, OpenAI’s CEO of applications, has taken a hands-on role in the launch, particularly in the absence of a dedicated ads chief. In recent weeks, she has been meeting agency execs and consultants as the company works to steady the rollout.
Most marketers will not get that level of access, at least not yet. Securing time with Simo or others on the team has quickly become its own hurdle. Of the 27 brands and agencies Digiday contacted, three said they had spoken with someone at the company, while the rest have yet to find a way in.
What emerged from those conversations suggests a more fluid pitch than the one publicly outlined so far. Initially, OpenAI asked agencies to bring forward clients prepared to spend at least $250,000 on the test ads. More recently, that figure has shifted to as much as $1 million a month — or at least according to one ad buyer with knowledge of the discussions.
And the ambition appears bigger still. OpenAI is eyeing the broader pools of media dollars agencies control through their commercial agreements, added the ad exec. For a platform, securing agency-managed budgets means repeat volume across multiple clients and influence over how dollars are allocated across channels. It is a sturdier growth model than chasing isolated brand experiments, especially when enthusiasm for ads inside a conversational AI interface is still, at best, being tested.
“I’d think our advertisers who said no would say yes when it’s [the test] is opened up and they don’t have to commit [a heavy minimum spend],” said Phillip Thune, CEO of Adthena.
He said a growing number of his roughly 400 clients have been approached about the pilot, with OpenAI asking for a $250,000 minimum commitment. Some agreed. Most declined. Thune expanded on the reason why: “They [test advertisers] will very quickly tie the CPM to either it works in the economics they want to see, or it doesn’t. They just don’t want to spend $250,000 to figure that out.”
In the end, even the most carefully managed rollout runs into a familiar constraint: price. The cost of ads in ChatGPT’s Free and Go tiers — pegged at around $60 per thousand views — has become a sticking point for many marketers. For a platform still finding its footing in advertising, that CPM feels steep. There is little historical performance data, limited benchmarks and no guarantee of return.
“None of our advertisers want to commit as much without any viable return, especially with it being impression focused and CPMs priced as aggressively as live sports,” said David Dweck, president at Go Fish, who added that a large number of his team’s client base is performance advertisers.
How widespread those concerns are will become clearer as the test phase continues and the initial list of advertisers either expands into a broad cohort, or narrows to a more selective group willing to stay the course. As it stands, the list is a who’s who of big brands and agencies:
WPP Media, with pilot partners across automotive, CPG, entertainment, gaming, luxury, retail, technology, and travel verticals include Adobe, Audemars Piguet, Audible, Ford, Mazda, Mrs. Myers as well as Target and Roundel Retail media business partners, among others
Dentsu has put forward “several clients” across CPG, grocery, hospitality, retail, software and travel
Omnicom Media has more than 30 clients partaking in the test, across apparel, automotive, beauty, CPG, hospitality, retail, QSR, technology, and telecommunications
Publicis
A number of Adthena’s clients
Fever
William-Sonoma
Bed, Bath & Beyond
The Knot WorldWide
HelloFresh
Albertsons Companies
Even at this early stage, much of the rationale behind these tests appears less about conviction and more about calculation: securing optionality, learning fast and paying only as much tuition as the risk justifies. Right now, that risk is high. The same week ads launched, OpenAI researcher Zoe Hitzig announced her departure, publishing an opinion piece in The New York Times and embarking on a press tour to explain her concerns. For marketers weighing brand risk, those headlines are part of the equation. Will public unease spill over into hostility toward ads, or worse, into attrition from the app itself? It is too early to know but it’s clearly something being monitored.
Whatever the outcome, OpenAI appears intent on managing the rollout tightly — controlling the narrative, the pricing and the pace — in the hopes that caution buys it the room to scale.
Simo has said as much. In an interview with Alex Heath and Ellis Hamburger on the ACCESS podcast, she acknowledged that there is definitely still a risk of people still believing that the ads are influenced by what users say to ChatGPT, noting that Facebook is a good example of this type of misperception.
“Facebook didn’t sell your [users] data to advertisers but tons of people think it did. Facebook doesn’t listen to your [users] microphone and tons of people think they do. So is there a perception risk? For sure,” Simo said. “We are doing everything we can in the product through education to educate people. The thing that I think is also helpful is because we have a conversational product, I think people will actually talk to the model and realize what’s going on.”
On that basis, its recruitment plans take on added significance. Current vacancies include roles for a software engineer within the “Ads Integrity” team, first reported by Business Insider, and another posted on Feb. 6, for a monetization product marketing manager, advertising, point to OpenAI’s focus on growing its advertising business in a safe, trusted, and controlled way, with expertise to professionalize and scale the effort.
Bottom line: OpenAI is building the guardrails at the same time as it lays the track. Whether that proves to be disciplined execution or accelerated improvisation will shape how this experiment is judged — by users and marketers alike.
In an emailed statement, an OpenAI spokesperson said: “We’re grateful to the partners participating in this test and for their alignment with our ads principles. We’re taking a deliberate, principle-driven approach and will continue listening closely to feedback to ensure that any advertising supports broad access to AI while keeping the trust that makes ChatGPT valuable.”
Comments (0)