Introduction
In the span of one week in February 2026, the two most important AI companies in the world publicly split on a question that will shape everything that comes next: should AI make money from advertising?
Anthropic said no, emphatically, with Super Bowl ads that mocked the very idea. OpenAI said yes, quietly at first, then defiantly, launching ads in ChatGPT days after Anthropic’s campaign aired to 120 million viewers.
This is not a minor branding spat. It reveals fundamentally different visions for what AI should be, who it should serve, and how the economics of artificial intelligence will work for the next decade. And for anyone who advertises (or plans to), the outcome matters more than you might think.
The Super Bowl Moment
February 2026. Anthropic buys four Super Bowl ad slots (some of the most expensive airtime on the planet) and runs spots titled “Betrayal,” “Deception,” “Treachery,” and “Violation.” Each depicts a humorous but pointed scenario: a user asking their AI assistant for genuine advice about fitness, therapy, or travel planning, only to have the response hijacked by an ad for protein powder, a therapy app, or a discount airline. The tagline: “Your AI should work for you. Not advertisers.”
The ads were funny. They were also a $30 million attack on OpenAI’s business model, timed to land days before ChatGPT’s ad launch on February 9.
Sam Altman responded publicly, calling Anthropic’s campaign “dishonest” and arguing that it misrepresented how ChatGPT ads actually work. He was right on the specifics: ChatGPT ads appear below responses, not within them, and OpenAI has stated they run on separate systems from the AI model. But Anthropic’s campaign wasn’t really about the mechanics of ad placement. It was about a deeper question: once advertisers become a stakeholder in a conversation between a user and their AI, does the nature of that conversation change?
The AI industry now has two competing answers to that question. The stakes are higher than either company wants to admit.
From “Last Resort” to $60 CPM
It is worth remembering how recently OpenAI’s position was identical to Anthropic’s. In 2024, Sam Altman described advertising as “a last resort” and told an interviewer that “ads plus AI is sort of uniquely unsettling.” The implication was clear: ChatGPT would find another way.
By January 2026, OpenAI was charging $60 CPM with a $200,000 minimum commitment and had hired Dave Dugan, Meta’s former VP of global clients, to lead ad sales. That is not the hiring decision of a company treating advertising as experimental. It is the hiring decision of a company building a permanent revenue channel.
$15B
Projected OpenAI cash burn in 2026
What changed? The math. OpenAI projects $15 billion in cash burn in 2026. The company generates substantial subscription revenue, but roughly 75% of it comes from consumers, and the free tier’s 480 million-plus users generate massive inference costs without producing any revenue at all. Every free conversation has a compute cost. At scale, those costs are staggering.
Ads solve this problem directly. They turn the free tier from a cost center into a revenue generator without requiring users to pay. OpenAI is not the first technology company to reach this conclusion. It is following the same path that Google, Facebook, and every other company that tried to sustain a free product eventually walked.
The question is whether the path leads to the same destination.
The Trust Paradox
ChatGPT built its value on being a trusted, unbiased advisor. People don’t use it the way they use Google. They ask it for help with personal decisions: financial planning, health questions, career moves, relationship advice. They share context they would never put into a search query: their salary, their diagnosis, their anxieties about a business decision.
This level of intimacy is what makes ChatGPT valuable to advertisers. It is also what makes advertising inside ChatGPT fundamentally different from advertising on a search engine or social feed. When you search Google for “best running shoes,” you expect ads. The context is transactional. When you ask ChatGPT to help you build a marathon training plan and it recommends a specific shoe brand, the context is advisory. The same ad can feel helpful in one context and manipulative in the other.
OpenAI is aware of this tension. Ads are architecturally separated from responses. They appear below the AI’s answer, visually delineated with a “Sponsored” label. The company has stated repeatedly that advertising content does not influence the model’s output.
But the trust problem is not about whether ads actually influence answers. It is about whether users believe they might. Once a user starts asking “is this recommendation genuine or is someone paying for it?” that behavioral shift is difficult to reverse, even if the answer is that no one is paying for it. The mere presence of advertising creates a perception of influence, and perception shapes behavior.
Forrester research found that 83% of users say they will keep using the free tier despite ads. That number sounds reassuring, but it masks a deeper concern. Continued usage and continued trust are not the same thing. Users may keep using ChatGPT while simultaneously discounting its recommendations, cross-referencing answers, second-guessing suggestions, treating the tool as less authoritative. If that happens, the very quality that makes ChatGPT ads valuable (the trust users place in the AI’s advice) erodes quietly.
What Users Actually Experience
The theory of ChatGPT ads is clean: relevant, unobtrusive placements that help users discover useful products. The early reality has been messier.
PCMag tested the ad experience shortly after launch and found large, poorly-targeted ads that contradicted OpenAI’s promises of relevance. On mobile, ads took up nearly the entire screen, pushing conversation history out of view. In one documented case, a StubHub ad appeared during a conversation about the AI industry, a mismatch that undermines the premise that ads will be contextually useful.
Reddit communities surfaced similar complaints. Users reported ad saturation that felt disproportionate to the value of the free tier, with some explicitly stating they were switching to Claude because of the ad experience. The most common criticism was not that ads existed at all, but that they were intrusive and irrelevant, the two things OpenAI had specifically promised they would not be.
OpenAI’s own data tells a different story. The company reports low dismissal rates and no measurable decline in trust metrics. But these are self-reported by the platform that benefits most from positive numbers. Early ad performance data from any platform should be interpreted with that conflict of interest in mind. Independent user sentiment, measured through app store reviews, social media discussion, and third-party surveys, is more skeptical than OpenAI’s internal metrics suggest.
This gap between official data and user sentiment matters. If users feel the ad experience is poor but OpenAI’s metrics say otherwise, the metrics may be measuring the wrong thing. Low dismissal rates could simply mean users don’t know the dismiss button exists, not that they welcome the ads.
Anthropic’s Counter-Model
Anthropic’s commitment that “Claude will remain ad-free” is framed as a principled stand: the company says it wants Claude to “act unambiguously in our users’ interests” without the structural conflict that advertising introduces. It is a compelling position. It is also a position that their business model happens to support.
Anthropic derives most of its revenue from B2B API sales. Companies like Amazon, Canva, DuckDuckGo, and thousands of startups pay for Claude access through the API. Consumer subscriptions exist but are secondary to the business. This means Anthropic does not face the same free-tier economics that forced OpenAI’s hand. When your revenue comes from enterprise customers paying per token, you do not need to monetize consumer conversations with ads.
This is not to say Anthropic’s position is cynical. The company may genuinely believe that advertising compromises AI integrity. But it is easier to hold that belief when your business model does not require you to test it. OpenAI’s leadership probably believed the same thing in 2024, before $15 billion in projected costs changed the calculation.
The strategic question for Anthropic is whether the ad-free commitment survives scale. If Claude’s consumer user base grows to match ChatGPT’s (800 million-plus weekly active users), the inference costs become enormous. At that scale, B2B API revenue alone may not cover the infrastructure bill. Anthropic would then face the same difficult math that OpenAI faces today, and the “never” in “ad-free forever” would be tested against financial reality.
The Google Search Precedent
History has a useful parallel here. In 1998, Larry Page and Sergey Brin wrote in their Stanford research paper that “advertising funded search engines will be inherently biased towards the advertisers and away from the needs of consumers.” They argued that ad-supported search would inevitably compromise result quality and that the incentives were structurally misaligned.
Google went on to become the largest advertising company in the world. By 2025, more than 80% of Alphabet’s revenue came from advertising. The wall between organic search results and paid placements held, until it didn’t. Over two decades, the visual distinction between ads and organic results shrank. Sponsored listings moved from a clearly separated sidebar to positions above and within organic results. The labels became smaller. The formatting became more similar.
Did search quality decline? The answer depends on who you ask. But nobody disputes that advertising fundamentally changed how Google prioritizes content. The company now optimizes for two audiences simultaneously (users who want the best answers and advertisers who want visibility), and those interests do not always align.
The parallel to ChatGPT is direct. A tool originally built to serve users is being adapted to also serve advertisers. OpenAI insists the wall between ads and answers will hold. Google’s founders said the same thing. The wall held for years. Then it didn’t. The question is not whether the wall will eventually thin (the incentives guarantee that it will), but how long it takes and how much trust erodes in the process.
What This Means for Advertisers
If you advertise or plan to advertise on ChatGPT, the trust debate is not abstract philosophy. It is a practical question about the long-term viability of the channel.
Right now, ChatGPT ads benefit from novelty. Users have not yet developed the “ad blindness” they have for Google search ads or Instagram sponsored posts. Engagement rates are high. Conversion rates from ChatGPT referrals are reported at 1.5x to 4x higher than other channels, depending on the category. For early advertisers, the economics are genuinely attractive.
But novelty fades. The long-term value of ChatGPT as an ad platform depends entirely on whether OpenAI can maintain a strict separation between advertising and the AI’s recommendations. If that separation holds, ChatGPT could become one of the most valuable ad channels ever created: a platform where users describe their problems in paragraphs of detail and ads are matched with genuine relevance. If the separation erodes, either in reality or in user perception, the premium CPM becomes impossible to justify.
Metrics to Watch
Advertisers should track trust metrics alongside performance metrics. The numbers that matter are not just impressions, clicks, and conversions. Watch for:
- Declining click-through rates over time: a signal that users are developing ad blindness or distrust
- Increasing dismissal rates: users actively rejecting ads rather than ignoring them
- Negative sentiment in ad feedback: qualitative signals that the ad experience is degrading user perception
- Conversion quality decline: if downstream metrics like close rates or retention drop even as click-through holds steady, users may be clicking out of curiosity rather than genuine intent
If these trends emerge, they compound. Declining trust leads to lower engagement, which leads to lower ROI, which makes the premium CPM harder to justify. The advertisers who recognize these signals early will be able to adjust their spend before the economics turn against them.
The Verdict
Neither side is entirely right.
OpenAI’s ads may be necessary for keeping AI free and accessible to 800 million-plus users. Subscription revenue alone does not cover the cost of serving hundreds of millions of free users, and the alternative (making everyone pay) would cut off access for the people who benefit most from free AI tools. If advertising is the price of universal access, that is a defensible trade-off.
Anthropic’s ad-free stance may not survive their own scaling costs. It is easy to promise “ad-free forever” when your consumer user base is a fraction of your competitor’s. The promise becomes harder to keep when inference costs scale with user growth and B2B API revenue hits a ceiling. Anthropic may find other solutions (higher subscription prices, deeper enterprise partnerships, government contracts), but the math will get more difficult, not less.
The real question is not “ads or no ads.” It is “how well can the wall between ads and answers hold?” That is an empirical question, not a philosophical one. It will take years to answer definitively. Google’s wall held for about a decade before the erosion became obvious. ChatGPT’s wall is seven weeks old.
For advertisers, the pragmatic response is not to pick a side in the trust debate. It is to prepare for the reality that ChatGPT ads exist and are growing, while staying attuned to the signals that indicate whether user trust is holding or eroding. Lapis supports ChatGPT as a target platform for businesses that want to be ready regardless of how the trust debate evolves. If conversational AI advertising works (and early data suggests it can), the companies that prepared first will have the advantage. Start building your ChatGPT ad strategy while the channel is still new and the opportunities are largest.
For deeper context on how ChatGPT ads work in practice, read our complete guide to ChatGPT ads. For the question of paid ads versus organic visibility, see our analysis of paid ads vs. organic mentions on ChatGPT. And if you’re a founder building an AI-first marketing strategy, our guide to agentic ads covers how autonomous AI can manage campaigns at scale.
