Advertisers are treating OpenAI ad trials like a bet on the future. The question now is how long they are willing to wait until the debt is paid off.
So far, the wait hasn’t been easy. Since the test launched in February, advertisers have reported chronic underdelivery – campaigns not achieving targeted impressions, leaving budget unspent and the results hard to justify, according to several advertising executives, all of whom asked to remain anonymous because of the sensitivity surrounding the test.
One of them spent just $2,500 of a $250,000 commitment over four weeks, taking on 200 clients in the process. Other agencies described similar experiences, with another agency executive saying that when their first ad aired in February, almost nothing happened for almost a month. The supply, to be honest, is almost non-existent.
This frustration was even more pronounced at last month’s Digiday Programmatic Summit in Palm Springs.
A senior executive from a large agency told Digiday that OpenAI could only make about $100 per client per week until supply starts opening up around mid-April. In the town hall session, participants were more outspoken. “I think they don’t know the revenue and do a lot of things before they’re ready,” one of them said. Another put the problem in structural terms, “Less than 10% of the eligible base is part of the trial. We’re just having a general large-scale problem.”
The $250,000 minimum commitment doesn’t help. While OpenAI advertising representatives were clear that the funding would have to come from the innovation budget – funds set aside for tests that don’t need to produce immediate results – the amount set expectations that inventory could never meet. One agency participating in the pilot called the minimum spending commitment “a smoke and mirrors thing in hindsight. There was never enough inventory at launch to absorb that level of spending.
That’s starting to change. Over the past six weeks, four out of seven advertising executives Digiday spoke with said they were seeing more ads running on apps and, as a result, spending more. The pattern they describe is consistent: users ask ChatGPT a question, get an answer, and an ad. Ask for a follow-up, get another one. Fill rates are increasing — by as much as 30% to 50% compared to at launch, a representative for one advertising agency said.
When the ad landed, its performance surprised many people. One of the agency’s clients, tracking conversions via URL redirects before OpenAI measurement tools were available, found efficiencies approaching those of Google’s non-branded searches. For a CPM-only product in the first week, those results are not results that anyone can emulate. It turns out that the destination environment is completely different. Users aren’t searching and clicking like they used to. They are considering options and accepting in a way that passive exploration rarely produces.
That’s the kind of data OpenAI advertising executives can point to when asking for patience. For them, the downside is the cost of doing this right. Before an ad goes live, the platform filters for sensitive conversations, then contextual relevance, and only then considers auction dynamics. The limits on each step are deliberately set high. If you get it wrong, OpenAI will not only lose advertisers, it could also lose users, and the entire premise of the platform.
Whether marketers are sympathetic to that depends on how clearly they see what is going on. Building an advertising business without destroying the trust that makes a product worth advertising is very difficult. Advertisers who understand this have different expectations than those who don’t. Most of them don’t expect a miracle. But they expected more than this.
For some people, the gap between the two is enough to drive them away. A number of blue-chip advertisers have told their agencies they will not return to OpenAI advertising without a trusted intermediary, arguing that its delivery and reporting fails to meet basic expectations, said one advertising executive who works with them. For high-performing advertisers – those who have to work as hard as possible to earn every dollar – pilot programs are not a viable option.
The pilot was still effective in the cold-start phase when initiated. One ad format, limited inventory, and basic reporting — meaning advertisers are testing the unknown too much, rather than having the underlying mechanisms that make performance marketing provable and deliverable.
There is a slight mismatch of expectations; advertisers were repeatedly told this was a test and treated it almost like an early version of Google or Meta, while OpenAI was still developing the basics in real-time. Could this have had an impact on initial perception? Of course. Under-delivery and limited transparency are not easy to forget when advertisers want their dollars to work hard. For some, OpenAI then falls into the “not ready” or “not feasible” category, even though OpenAI has been very clear about it from the start.
There’s nothing missing in OpenAI. CPC bidding, conversion tracking, and pixel-based measurement APIs have all been added since launch, and CPA is still to come. The self-serve ad manager, now open in the US, eliminates minimum commitments altogether. Persuading advertisers of a clear direction, rather than convincing a company to keep going the way it wants, is a more difficult task.
It’s even harder because OpenAI’s own actions show that they are still learning how the advertising business actually works. When advertisers want unspent budgets released, OpenAI has trouble understanding why the request makes sense, according to a senior executive at one media network. Rather than canceling commitments outright, they should quietly onboard and remove clients from pilot projects to save on expenses.
It is an avoidable friction, and an open one.
Netflix also faced a similar situation when its ads didn’t meet its viewing guarantees in the early months, Netflix proactively offered cash back, garnered significant goodwill, and the move became something of a case study in how to handle a rocky start. OpenAI doesn’t do that. Not out of bad faith, the source said, but because executives didn’t understand why they had to do this.
For media companies, offering cash back is almost a distant memory. In contrast, OpenAI, which was born out of the software subscription and cloud economy, is approaching the situation more like a platform trying to overhaul spending than a media company managing a viewership shortage, says advertising technology expert Shirley Marschall. “It’s the kind of mismatch you’d expect from a company that is, in many ways, still an advertising baby.
“Additionally, with chat advertising becoming the industry’s newest obsession, this is not the time or market for buyers to hold a grudge over awkwardness in early trials,” Marschall said. “There are too few alternatives, and too much curiosity about what will happen next.”
OpenAI did not respond to Digiday’s request for comment.
PakarPBN
A Private Blog Network (PBN) is a collection of websites that are controlled by a single individual or organization and used primarily to build backlinks to a “money site” in order to influence its ranking in search engines such as Google. The core idea behind a PBN is based on the importance of backlinks in Google’s ranking algorithm. Since Google views backlinks as signals of authority and trust, some website owners attempt to artificially create these signals through a controlled network of sites.
In a typical PBN setup, the owner acquires expired or aged domains that already have existing authority, backlinks, and history. These domains are rebuilt with new content and hosted separately, often using different IP addresses, hosting providers, themes, and ownership details to make them appear unrelated. Within the content published on these sites, links are strategically placed that point to the main website the owner wants to rank higher. By doing this, the owner attempts to pass link equity (also known as “link juice”) from the PBN sites to the target website.
The purpose of a PBN is to give the impression that the target website is naturally earning links from multiple independent sources. If done effectively, this can temporarily improve keyword rankings, increase organic visibility, and drive more traffic from search results.