OpenAI is preparing to widen access to ChatGPT ads next month, while also trying to fix some of the weakest parts of the initial pilot. The Information reports that the company is planning to streamline ad buying through partnerships with ad tech firms or through its own ad management system. That effort makes sense. The first wave of ChatGPT advertising appears to have reached the market with plenty of curiosity from brands and a much thinner operational layer than most media buyers would normally accept.
The core issue in The Information’s article is fairly plain. Early advertisers have struggled to prove business impact.
The article says buyers in the initial pilot had to work through phone calls, emails, and spreadsheets rather than through an automated buying system. The same article says OpenAI provided little beyond views and clicks, without the kind of audience detail, conversion visibility, or downstream performance data advertisers usually expect from a serious digital channel. Two agency executives working with early ChatGPT advertisers told The Information they have not yet been able to prove measurable business outcomes for their clients.
A product can survive weak tooling for a while if the audience is strong enough. ChatGPT clearly has that advantage. The Information says OpenAI had around 920 million weekly active users at the end of February, while only about 5 percent of weekly active users were paying as of last July. A platform with that much reach is always going to attract experimental spend.
Interest, however, is not the same thing as proof.
The version of ChatGPT ads described in The Information reads much closer to premium experimental inventory than to a mature advertising channel. Marketers, agencies, and finance teams should judge it on those terms for now.
Demand Arrived Before the Product Was Fully Built Out
The most revealing part of The Information’s reporting is the order in which the market is developing.
OpenAI already has scale, attention, investor expectations, and a large population of nonpaying users it would like to monetize. The ad product still appears to be catching up on the basics. Advertisers reportedly committed meaningful budgets in advance, paid a premium rate of $60 per thousand ad views, and entered a buying process that sounds more like old style direct sponsorship sales than modern digital media buying. The article says OpenAI required advertisers to commit at least $200,000 each for the initial launch.
A company can get away with that during an early rollout if buyers believe the inventory is scarce, high attention, and strategically important. ChatGPT checks all three boxes. Marketers want firsthand knowledge of how ads perform inside an AI interface that people return to frequently. Agencies do not want to wait too long and enter after the strongest learning period has already passed.
Even with that context, the product described in The Information still looks like a channel built around possibility more than proof.
One detail from the article captures the state of the rollout especially well. More than halfway through the pilot, one agency executive said its clients had spent only around 15 percent to 20 percent of their committed budgets. A slow ramp is normal in many pilots. A slow ramp paired with thin reporting and low delivery leaves buyers in a difficult position. They have committed real money, but the campaign has not yet generated enough scale or enough evidence to make a credible case for performance.
Premium Pricing Does Not Automatically Create a Premium Channel
OpenAI priced the first run like premium inventory. The Information says the company charged $60 CPM, a level associated with high end media environments like live NFL games.
That pricing sends a clear signal. A premium rate suggests premium conditions, whether those conditions come from audience quality, scarce inventory, strong attention, brand safety, or some combination of those factors.
The reporting stack described in The Information does not yet sound premium in the same way.
Views and clicks can tell a buyer that delivery happened. Views and clicks do not tell a buyer much about who saw the ad, how that exposure connected to later behavior, whether the campaign influenced business outcomes, or whether the same spend would have produced more value elsewhere. Established digital media platforms built their ad businesses partly by making those questions easier to answer, even when attribution in practice remained imperfect.
OpenAI’s current offer appears to ask buyers to accept premium pricing while working with early stage visibility into results.
Large brands can absorb that kind of uncertainty more easily. An experimental budget can cover a test like this. A large advertiser may value access, internal learning, PR value, and category presence even when hard measurement remains limited.
Smaller brands operate under different constraints. The Information points directly to that issue when it reports that one agency executive would not recommend ChatGPT ads to smaller clients who cannot afford experimental budgets. That sounds like a sensible conclusion. Smaller advertisers usually need digital media to do more than generate curiosity. The spend has to connect to revenue with enough clarity to justify doing it again.
The Buying Experience Still Sounds Manual and Fragile
Another important thread in The Information’s article is how low tech the initial buying process appears to have been.
Advertisers reportedly had no automated way to purchase ad space. Buyers relied on calls, spreadsheets, and emails with OpenAI representatives. A manual launch process is not unusual for the first stage of a new ad business, especially when the seller wants close control over who gets access and how campaigns run. Even so, a manual process tells you a lot about the maturity of the product.
A premium media business can absolutely sell inventory through direct relationships. Plenty of them have done that for years. A platform that wants to become a repeatable part of a broader digital media plan usually needs much more than access. Buyers need pacing, campaign management, reporting, trafficking, creative variation, budgeting controls, and eventually a self serve layer that lets teams move without waiting on human back and forth every time.
The Information says OpenAI is planning its own ad manager and is already testing a self serve version with some partners, though the company has not announced a broad rollout timeline. That development feels necessary. A self serve system would not only make buying easier, it would also give OpenAI more control over campaign structure, reporting, and the cadence of broader adoption.
A smoother interface, though, only solves one part of the problem. Easier campaign management does not automatically create better targeting, stronger measurement, or greater advertiser confidence.
Ad Tech Partners Can Improve Access Before They Improve Measurement
The Information also reports that OpenAI is working with outside firms that can help sell ads on its behalf. The company has reportedly held talks with several ad tech businesses and recently announced an integration with Criteo, which focuses on retail advertising and can provide both a buying interface and targeting related technology. Existing ChatGPT advertisers are reportedly getting priority access to the Criteo rollout, although money spent in the pilot will not count toward the new commitments Criteo has been pitching.
The operational value of those partnerships is easy to understand. Outside firms can bring OpenAI into more buying workflows, widen the pool of advertisers who can access the inventory, and give agencies a more familiar system for planning and execution.
The targeting side appears more limited right now. The Information says one agency executive believed OpenAI would share only limited ad targeting data with Criteo, while another person who spoke with OpenAI said the company is limiting what it shares with ad tech firms to very basic information. If that description is accurate, then Criteo may help buyers purchase inventory more easily without immediately solving the deeper intelligence problem.
Aggregated data around impressions, clicks, and spending gives a campaign some reporting structure. It does not provide the audience visibility or outcome clarity buyers are used to from Google or Meta. Those businesses built powerful ad systems over many years by turning user behavior into planning, targeting, optimization, and measurement capabilities. OpenAI is entering the market with audience scale already in place while the commercial infrastructure is still under construction.
That imbalance is going to shape the next stage of adoption.
ChatGPT Ads Belong in Experimental Budgets for Now
The Information describes a pattern that shows up often in new ad markets. Marketers are testing ChatGPT ads with small portions of broader budgets, drawing from funds set aside for unproven formats rather than pulling meaningful spend away from established channels.
That is exactly where a product like this belongs at the current stage.
A disciplined advertiser can still learn a lot from an early market without pretending the market is already mature. Teams can study how creative behaves inside a conversational interface. They can examine whether text and image variation affects delivery. They can learn whether the environment feels additive for certain categories. They can assess whether the product has any unusual brand effect, attention quality, or category fit.
The mistake would be treating the current version like a dependable performance line item.
The Information’s reporting points to an ad product that still needs more delivery, more measurement, more buying infrastructure, and more evidence before marketers should describe it as a scalable channel. OpenAI’s own spokesperson language fits that reading. The company says it is encouraged by early signals from users and participating brands, is seeing a wide range of results, and is rolling ads out slowly because ChatGPT is a trusted and personal environment.
That is reasonable language for a company still learning.
Marketers still need to judge the product by what can be defended in a media review, not by how strategically exciting the platform feels.
OpenAI’s Revenue Pressure Helps Explain the Timing
The timing makes sense from OpenAI’s side.
The Information says OpenAI projected in January that it would generate $17 billion in consumer revenue this year, including ads as well as subscriptions. Once that number is placed next to the size of the nonpaying user base, the move toward advertising becomes much easier to understand. A company with massive usage, rising infrastructure costs, and a relatively small pool of paying users is always going to look at ads as a major revenue lever.
The timing also makes sense competitively. If OpenAI waits until every system is fully mature, it risks giving other platforms more time to define what AI advertising looks like. Early rollout creates advantages beyond immediate revenue. The company gets buyer feedback, learns how agencies want to plan against the product, and starts shaping how marketers think about ChatGPT as a media environment rather than only as software.
A lot of ad markets begin that way. The platform does not wait for perfect systems. The platform opens the door, watches where demand gathers, learns from the first budgets, and gradually builds the infrastructure underneath.
Buyers need to remember what role they are playing when that happens. An early buyer may gain access and insight. The same buyer may also end up funding part of the platform’s learning curve.
The Bigger Signal for AI Advertising
The most interesting part of The Information’s article may be broader than OpenAI alone.
AI products are beginning to attract ad budgets before the industry has settled on how performance should be measured inside them. That condition is likely to repeat across the category. User attention is moving into AI interfaces quickly. The commercial systems wrapped around those interfaces remain uneven. Budget will keep arriving anyway because advertisers follow behavior first and reporting standards second.
That creates a messy transition period.
Platforms will talk about opportunity. Agencies will talk about learning agendas. Buyers will justify tests through innovation budgets. Leadership teams will still want to know what the spend accomplished. The gap between platform momentum and advertiser proof is where a lot of weak decision making usually gets made.
The Information gives a clear early picture of that gap inside ChatGPT ads. OpenAI has scale, attention, and advertiser demand. The company is widening access, increasing the number of users seeing ads, building an ad manager, and working with outside partners to make buying easier. Early advertisers still appear to have limited evidence that the ads are producing measurable business outcomes.
A marketer can work with that reality, but only by describing it honestly. ChatGPT ads currently look best suited to brands that can afford a true experimental budget, want exposure to a fast developing environment, and understand that much of the value may come from learning rather than immediate proof. Smaller advertisers and performance driven teams would be wise to stay cautious until the product offers a clearer picture of audience quality, targeting depth, and business results.
OpenAI may close those gaps quickly. The company has every reason to try.
For now, The Information’s reporting points to a straightforward conclusion. Demand showed up first. Measurement is still trying to catch up.
References
The Information, “OpenAI’s First Advertisers Can’t Prove ChatGPT Ads Work,” March 2026