Sam Altman spent two years publicly calling advertising a “last resort” and telling anyone with a microphone that he liked the fact people paid for ChatGPT knowing the answers weren’t influenced by advertisers. Then, on February 9, OpenAI quietly placed its first ads inside ChatGPT conversations — and the premium pricing it launched with has been in freefall ever since. The $60 CPM that early advertisers paid at launch had cratered to as low as $25 within ten weeks. By April 15, OpenAI had abandoned the impressions model entirely, pivoting to cost-per-click pricing at $3 to $5 per click. And on May 5, the company opened a full self-serve Ads Manager to global advertisers with a minimum spend slashed from $250,000 to $50,000.
That’s not a confident rollout of a new revenue stream. That’s a company watching its business model crack in real time and scrambling to plug the holes before investors notice.
The CPM Collapse Nobody’s Talking About
When OpenAI launched ChatGPT ads in February, it charged a $60 CPM — cost per thousand impressions — with a minimum spend between $200,000 and $250,000. That put it in the same pricing tier as premium Super Bowl adjacency. The early roster was impressive: Target, Ford, Adobe, Expedia, Mrs. Meyer’s. Within six weeks, the pilot had crossed $100 million in annualized revenue with several hundred advertisers.
Then the floor fell out. A leaked StackAdapt deck from late March showed CPMs being offered as low as $15 — a quarter of the launch rate. Advertisers who’d paid premium prices in February watched their competitors get the same inventory at steep discounts weeks later. By April, CNBC was reporting that the test was “moving too slowly to meet the hype” and that OpenAI “can’t prove the ads are working” due to the absence of mature measurement tools.
Here’s the thing about a 60% CPM collapse in ten weeks: it doesn’t mean the product failed. It means OpenAI priced the product based on the story it was telling investors, not on the value advertisers were actually receiving. A $60 CPM works when you’re the only AI chatbot in town and advertisers are buying novelty. It stops working the moment those advertisers start asking what the click-through rate actually is.
The CPC Pivot Is an Admission, Not a Strategy
On April 15, OpenAI shifted ChatGPT advertising from CPM to cost-per-click pricing, with bids between $3 and $5 per click. The minimum spend dropped to $50,000. The company framed this as a natural evolution. It wasn’t. It was the ad-tech equivalent of a fire sale.
CPM pricing works when you control enough attention that advertisers will pay just to be seen. That’s Google Search. That’s Instagram Stories. It requires enormous scale and proven conversion data. OpenAI has neither — ChatGPT has roughly 800 million monthly users, but the free and Go tiers that see ads represent a fraction of that, and the conversion tracking pixel the company launched only supports basic events like lead creation and subscription starts.
CPC pricing, by contrast, shifts the risk from the advertiser to the platform. OpenAI only gets paid when someone actually clicks. For advertisers, that’s great — it’s the model they already use on Google and Meta, and it gives them a metric they understand. For OpenAI, it means revenue is now directly tied to whether users engage with ads in a context where they came to ask questions, not browse products. The fundamental bet is that a user asking ChatGPT about the best running shoes will click on a sponsored Nike card with the same intent they’d click a Google Shopping result. That’s far from proven.
The Self-Serve Ads Manager: Democratization or Desperation?
On May 5, OpenAI launched its ChatGPT Ads Manager — a full self-serve platform that lets any US business create, manage, and optimize ad campaigns inside ChatGPT conversations. No agency required. No $250,000 minimum. Just $50,000 and a dashboard. International expansion to Australia, New Zealand, and Canada followed within 48 hours. The UK, Brazil, and Japan came next.
OpenAI hired Shivakumar Venkataraman, a 21-year Google veteran who led Google’s search ads business, to run this. It partnered with StackAdapt for programmatic placement, Smartly and Criteo for conversational ad formats connecting to Criteo’s network of 17,000 advertisers. Agency partners include Dentsu, Omnicom, Publicis, and WPP. The infrastructure is serious.
But the timing tells a different story. You don’t slash your minimum spend by 80% and open self-serve access three months after launch because things are going well. You do it because the enterprise advertisers who paid premium rates aren’t renewing at those rates, and you need volume from smaller businesses to compensate. OpenAI is projecting $2.5 billion in ad revenue for 2026, scaling to $11 billion by 2027, $25 billion by 2028, and $100 billion by 2030. To hit the 2026 target, it needs roughly $300 million per month in ad revenue for the rest of the year — starting from a base of about $100 million annualized in March. That’s a 30x increase in nine months.
The Privacy Bomb Nobody Defused
The day ChatGPT ads launched, Zoe Hitzig, a researcher at OpenAI, resigned. In a New York Times piece, she described ChatGPT’s conversation logs as “an archive of human candor that has no precedent” and warned that OpenAI risks following “the same path as Facebook.” Her concern was precise: while advertisers don’t see user conversations, OpenAI processes conversation content internally to serve contextually relevant ads. Health conditions, financial distress, relationship problems, personal struggles — all being analyzed and categorized to decide which ad you see.
OpenAI says targeting is purely contextual, matched to the topic of the current conversation, with no demographic or third-party data. It says advertisers receive only aggregated performance data. It says users under 18 don’t see ads. But the structural tension doesn’t resolve: the same intimate conversations that make ChatGPT useful are the signal that makes the ads work. That’s not a privacy policy problem. That’s an architectural one.
A boycott campaign called QuitGPT has gathered more than 200,000 sign-ups since late January. An Ipsos survey found nearly two-thirds of US adults say ads in AI search make them trust the results less. These aren’t fringe reactions — they’re the market telling OpenAI that a chatbot is not a search engine, and users don’t treat it like one.
Anthropic and Perplexity Are Betting the Opposite Way — And Winning
While OpenAI builds an ad machine, its two closest competitors have staked their entire positioning on being the anti-OpenAI. Anthropic spent millions on Super Bowl commercials with titles like “Deception,” “Betrayal,” “Treachery,” and “Violation” — tagline: “Ads are coming to AI. But not to Claude.” The result? An 11% jump in daily active users and a climb to seventh on the App Store. Claude positioned itself as the trusted alternative to the ad-supported chatbot, and users responded immediately.
Perplexity tested sponsored follow-up questions in 2024 and 2025, then abandoned advertising altogether, citing user trust concerns. It’s now targeting $500 million in pure subscription revenue as the explicitly ad-free AI search engine.
The split matters because it creates a permanent positioning difference in the market. Once you introduce ads, you can never credibly claim to be ad-free again. OpenAI made that choice, and every competitor that didn’t now has a marketing advantage that compounds with every privacy scandal, every CPM collapse, and every user who switches because they don’t want their therapy-adjacent conversations feeding an ad algorithm.
Follow the Money: $14 Billion in Losses and No Plan B
The financial pressure behind every decision here is blinding. OpenAI generated $13 billion in revenue in 2025 and is currently producing roughly $2 billion per month. It closed a $122 billion funding round at an $852 billion valuation in March, led by Amazon, Nvidia, and SoftBank. But projected losses for 2026 are approximately $14 billion on compute, research, and infrastructure. Profitability isn’t expected until 2030. An IPO filing is planned for the second half of this year, with a 2027 listing targeting up to $1 trillion.
Advertising is the fastest path to closing the gap without another funding round or subscription price increase. The US market for AI search advertising is projected to grow from $1 billion in 2025 to $25.9 billion by 2029. OpenAI’s $2.5 billion target would make it a major player immediately — if it can get there. But even $2.5 billion only covers about 18% of the projected $14 billion loss. Ads won’t save OpenAI’s bottom line. They’ll just make the loss look slightly less catastrophic to IPO investors who need a growth story beyond subscriptions.
The Verdict
OpenAI isn’t building an ad business because it discovered a brilliant new revenue stream. It’s building one because it’s burning $14 billion a year, its CPM model collapsed in ten weeks, and it needs something — anything — to show IPO investors that revenue can scale faster than costs. The self-serve Ads Manager, the CPC pivot, the slashed minimums, the agency partnerships — they’re all symptoms of a company that priced itself like Google before it proved it could convert like Google.
Meanwhile, Altman’s competitors are eating his credibility for breakfast. Anthropic’s ad-free positioning is converting users. Perplexity walked away from ads entirely. And 200,000 people have signed up to boycott the product. The $2.5 billion target isn’t a business plan — it’s a prayer dressed up in a self-serve dashboard. And when the IPO filing drops later this year, that dashboard’s numbers will tell investors exactly how well the prayer is working.
