Update: OpenAI launches image ads in ChatGPT with DoubleVerify and IAS pilots

Full analysis

What's new since we last covered this: OpenAI expands ChatGPT ad formats with brand-safety vendor pilots and measurement partner.

OpenAI is building an ad business the grown-up way. Image ads inside ChatGPT, a new measurement partner in Kochava, and brand-safety pilots with DoubleVerify and IAS, the two vendors that tell buyers whether an ad landed next to something that would embarrass the brand. The constraint sits in plain sight: there isn't enough ad space inside ChatGPT today to soak up the budgets advertisers want to spend.

What's actually being decided here isn't OpenAI's. It's whether the ad-tech ecosystem treats ChatGPT as a real new channel to plan for, or waits. That's easy to undo either way, which is exactly why most operators will wait too long. The deadline nobody controls is the holding companies: once they green-light test budgets, the clock starts on whether the inventory can scale to meet the demand.

The Market Analyst. Being named by OpenAI is a gift to DoubleVerify and IAS, and both stocks need it. Their core job, verifying ads on the open web, is under slow pressure as spend drifts to walled gardens. Getting tagged as the verification layer for an AI chat surface reframes them as category-agnostic infrastructure, not open-web specialists. For the informed outsider: these two are the ones who certify to a brand that its ad didn't run next to garbage, and now they certify it inside ChatGPT too. The quieter loser is The Trade Desk's story that it sits at the center of the premium open internet. If OpenAI stands up a walled garden with agency-grade measurement, it courts the same brand-safety-conscious budgets. That's a 2026 fight, not a today fight.

The Skeptic. The inventory line buried at the end is the whole story. Brand-safety pilots with DV and IAS are there to get CMOs to approve test money, not because a scaled, measurable channel is waiting. ChatGPT is where people go to finish a task. They are not browsing in a receptive, ad-friendly mood. The CPMs OpenAI needs to justify its economics sit far above what an unproven channel commands at scale. For the outsider: OpenAI gets a trust checkbox, DV and IAS get a logo, and the gap between how much advertisers want to spend and how little space exists to run it goes unchanged. Early pilot prices will anchor expectations that have no relationship to what survives scale.

The Operator. The work is real and it's new. A ChatGPT conversation is structurally different from a web page or a pre-roll video. The classifiers DV and IAS built to read a page can't just be pointed at a chat thread and trusted. They need retraining, or at minimum re-tuning, for "AI chat" as its own environment. That's a product sprint, not a press release. Kochava's inclusion tells you OpenAI wants mobile-attribution hygiene baked in from day one, which is a shot across the bow of the mobile-measurement incumbents like AppsFlyer and Adjust. For the outsider: OpenAI is wiring in the plumbing that lets a buyer know whether a tap turned into a download before it ever needs to.

The Customer / End User (the advertiser). Advertisers pushed for this, which is why it exists. But wanting a channel and being able to feed it are different. A brand buyer can approve a ChatGPT test today and still find there's nowhere to put the next million dollars. The verification partners matter because no holding-company buyer places real budget without third-party checks ticked. For the outsider: the agency won't spend until someone it trusts, not OpenAI, confirms where the ad ran. So DV and IAS aren't a nice-to-have. They're the unlock. The thing advertisers can't get yet is scale, and no measurement partner fixes that.

The CFO. For DV and IAS, a new environment tag is a new revenue line, but it's small and it's front-loaded with cost. You pay to retrain classifiers now; you collect fees later, and only if OpenAI's inventory actually grows into the budgets. For OpenAI, ads are a margin story layered on top of a business that burns cash on compute, and a thin, task-focused surface won't throw off search-grade ad revenue soon. For the outsider: everyone is spending today against a payback that depends entirely on whether ChatGPT becomes a place people discover products, not just finish chores.

Where the council splits. The Market Analyst and the Strategist see a moat being poured: lock in the measurement graph agencies trust, unlock holding-company budget, build the audience, justify a bidded marketplace, the Google playbook compressed. The Skeptic says none of that matters while the supply-demand mismatch holds, because task-completion sessions won't command premium prices no matter how clean the measurement is. The second split: the Operator thinks the DV and IAS work is hard and context-specific; the Market Analyst is happy to pattern-match their YouTube wins onto chat. Those can't both be cheap.

What it hinges on. One belief: is ChatGPT a high-intent discovery surface, or a task tool people want to leave fast? If it's discovery, the premium CPMs are real and the moat argument wins. If it's task completion, the inventory stays thin, the prices stay soft, and the verification deals are legitimacy theater that gets CMOs to wire test money. The council leans toward the Skeptic on timing and the Analyst on direction: the DV and IAS win is real and good for those two stocks right now, but the channel itself won't carry meaningful budget in 2026. Before anyone reprices a media plan around ChatGPT, verify two things: the actual available impression volume, and whether pilot CPMs hold once the whitelist opens up.

Prediction: DoubleVerify or Integral Ad Science will name an AI-chat or generative-AI environment as a distinct measured category in its investor materials on its Q4 2026 earnings call (February 2027), while reported revenue from that category stays immaterial to total revenue.

Confidence: Medium — the marketing incentive to claim the category is certain; the dollars behind it are not.

Why: OpenAI has publicly named DoubleVerify and IAS as brand-safety partners, and both companies badly need a growth story to offset slow pressure on their open-web verification business, so each has every reason to foreground an "AI environment" line for investors. But the same article states plainly there isn't enough ChatGPT inventory to absorb the budgets advertisers want, which means the fees DV and IAS can earn against that inventory are tiny today. The gap between a company's stated growth narrative and the money actually flowing is the whole call: they will talk up the category long before it pays. The opposite outcome, meaningful revenue from AI-chat verification by early 2027, would require ChatGPT ad inventory to scale roughly tenfold from a standing start in under two quarters, which the supply constraint rules out.

Revisit by 2027-02-28: We're right if DoubleVerify or IAS presents AI-chat or generative-AI as a named measured environment in Q4 2026 investor materials and no segment-level revenue from it is material. We're wrong if either company reports material AI-environment verification revenue, or if neither mentions AI-chat as a distinct environment at all.