Inside OpenAI's Ad Rush: A $1 Billion Bet With Mixed Results

Eight months is a long time in the attention economy. On January 16, 2026, OpenAI announced it would begin testing ads inside ChatGPT — a decision the company framed as a way to “expand access to AI” by subsidizing the free tier of a chatbot whose compute bills were becoming legendary. This week, the Wall Street Journal published “Inside OpenAI’s Ad Rush,” a detailed look at what that experiment has become: an artificial-intelligence company “adding tools and features to woo both top brands and smaller advertisers, with mixed results.”

That last clause — mixed results — is doing a lot of work. Because by every top-line metric, ChatGPT Ads has been a runaway success. The business crossed a $1 billion annualized revenue run rate on August 31, according to CNBC, barely seven months after the first sponsored message appeared in a US chat. Tens of thousands of advertisers now use the service, which has expanded from the United States to India, Europe, the Middle East, and North Africa. What began as a closed pilot with 50 brands brokered through WPP and Omnicom has become, by advertiser count at least, a mid-sized ad network.

And yet the Journal’s framing captures the tension that defines the business: the machine is enormous, but nobody is quite sure yet whether it actually sells things.

Why OpenAI needs the money

The strategic logic of ChatGPT Ads was never subtle. Each query costs roughly $0.003 in electricity alone, and with hundreds of millions of users on the free tier, the unit economics of “AI for everyone” only work if someone else pays the bill. Advertising is the classic answer — it is how Google and Meta turned free consumer products into two of the most profitable businesses ever built. OpenAI anticipates “low billions” of dollars from advertising in 2026, according to reporting that dates back to the January launch, and analysts have sketched scenarios where the segment scales toward $25 billion by the end of the decade.

The timing matters for another reason: OpenAI is preparing for a highly anticipated initial public offering, and every durable revenue line it can point to strengthens the narrative. A billion-dollar run rate in year one is a good slide. But an IPO roadshow also invites uncomfortable questions — and the gap between ChatGPT Ads’ volume and its performance is exactly the kind of question that gets asked.

What the toolkit actually looks like

What the Journal describes as an effort to “woo both top brands and smaller advertisers” maps onto a concrete product roadmap that has shipped at remarkable speed:

• Self-serve with no floor. The minimum advertiser commitment was $200,000 at the February launch. By May, OpenAI had eliminated spending commitments for some buyers entirely, opening a self-serve Ads Manager with CPC bids in the $3–5 range and a $200 daily cap.
• First-party audience uploads. New Ad Tools Terms published in June define “Audience Tools” — custom audiences built from advertisers’ own customer data, explicitly barring broker and third-party lists — and “Creative Tools” for AI-generated, localized ad creative.
• Product feeds and commerce. A September 4 update made shopping results inside ChatGPT exclusive to brands with connected product feeds, mirroring the feed-only structure Google adopted for Shopping years ago. Brands without a feed are now invisible in shopping placements “regardless of ad spend.”
• Measurement that finally measures. The ChatGPT Pixel now supports hashed phone numbers, names, regions, and postal codes; the Conversions API handles plural list fields and Android GAID; a LiveRamp partnership bolsters measurement; custom audiences can be edited incrementally and support GAID identifiers.
• A conversational Ads Manager plugin that lets advertisers build and optimize campaigns through plain language, with every recommended change previewed and requiring explicit confirmation.
• Programmatic plumbing. On September 10, Amazon began piloting ChatGPT Ads inventory through its DSP as a managed service — CPC and CPM pricing, OpenAI retaining control of all delivery decisions — with Delta Vacations among the first US brands testing it. OpenAI had earlier partnered with the retail media networks of Target and Albertsons, and reportedly discussed a deal with The Trade Desk before landing on Amazon.

Each of these moves does double duty: it lowers the barrier for the long tail of smaller advertisers while giving big brands the measurement and data infrastructure they require before committing real budgets.

The mixed part of “mixed results”

The underwhelming side of the ledger is just as well documented. Advertiser-reported click-through rates have come in as low as 0.91% per Adweek, against a roughly 6.4% benchmark for Google Search. Analysts at Adweek’s sibling coverage have argued the business is on pace to miss some of its own forecasts by as much as 90%. Competitors have piled on, and industry reporting has repeatedly noted marketers who remain unconvinced that a conversational interface — where the user is mid-thought, talking to a machine — behaves anything like a search results page where clicking an ad is a natural next step.

There is a structural argument behind that skepticism. A chatbot session is not a list of ten blue links; it is a dialogue. Users who ask ChatGPT which laptop to buy are asking for an answer, not a marketplace. Ads that interrupt the dialogue feel, per one widely cited characterization, more interruptive than helpful. OpenAI’s counter is that its users are “super intentional” — they arrive mid-decision, with context that no keyword query provides — and that ads which are genuinely useful in that moment should command a premium, not a discount.

That bet is unproven, but it is not untested. The September 4 update is effectively OpenAI conceding that commerce, not clicks, is the near-term business: if you cannot beat search’s CTR, build a shopping surface instead, and charge for presence in it. The feed requirement converts ChatGPT from a general interest ad network into a product-data-driven commerce channel. The upcoming view-through conversion model, which will bill on impressions rather than clicks, similarly acknowledges that in a conversational interface, influence may arrive without a click at all.

The trillion-dollar context

The ad rush also plays into a bigger story. OpenAI is reportedly in early talks for a funding round at a valuation that could approach $1.2 trillion, and advertising is one of the few levers that can diversify revenue beyond subscriptions and API calls at scale. The company’s second-quarter sales showed what the Journal separately called “tepid growth” compared with Anthropic’s, which makes every new revenue line more consequential.

There are real headwinds. Europe’s expansion puts ChatGPT Ads squarely inside the EU’s stricter AI Act advertising-transparency regime. OpenAI has already drawn a line banning ads for rival AI products, which itself became a story about how much control the company will exercise over its own commercial surface. And the free-tier users who fund the whole experiment have shown, repeatedly across the industry’s history, that tolerance for ads is finite and revocable.

What to watch

The Journal’s piece lands at the moment the experiment becomes a business. The markers that matter from here:

1. Whether CTR converges toward search benchmarks as formats mature, or whether OpenAI successfully redefines the metric entirely around view-through and commerce outcomes.
2. Whether the Amazon DSP pilot expands beyond “select brands” to broad availability — Amazon neared $70 billion in ads revenue last year, and even a small spillover would move OpenAI’s numbers.
3. Whether the IPO narrative holds if the segment’s growth rate decelerates from its blistering early pace, as analysts already warn it might.

Eight months in, OpenAI has built the scaffolding of a top-ten ad business: global reach, self-serve access, programmatic pipes, first-party data tools, and commerce formats. What it hasn’t yet proven is the harder thing — that the most intimate computing interface ever mass-deployed can carry advertising without breaking the spell that made it worth a trillion dollars in the first place. The Wall Street Journal’s “mixed results” is, for now, the honest scorecard.