Why OpenAI and Anthropic are really fighting about ads
The public spat between OpenAI and Anthropic over ads inside chatbots was noisy, but the substance sits elsewhere. This is an early test example of how AI platforms intend to fund themselves without undermining user trust.
Anthropic’s decision to air Super Bowl campaigns mocking the idea of ads inside AI assistants was done with clear, strategic intentions.
Its campaign, built around the tagline “Ads are coming to AI. But not to Claude” landed days after OpenAI confirmed it would begin testing advertising on ChatGPT’s free and low-cost tiers in the US.
OpenAI insists the satire misrepresents its plans. Ads, it claims, will be clearly labelled and will not influence responses, and the paid tiers will remain ad-free.
But the unusually long response from chief executive Sam Altman suggested the issue touches a nerve.
For all the back and forth earlier this week, both firms are responding to the same underlying pressure. Generative AI is extraordinarily expensive to run, and both titans need to find a way to monetise their platforms.
ChatGPT now serves hundreds of millions of users globally, the vast majority of whom pay nothing.
OpenAI has disclosed multi-billion-dollar operating losses driven by data centre costs and compute spend, and does not expect profitability until late in the decade.
But advertising, however carefully introduced, offers a scalable way for Altman to subsidise free access.
Different models, same bottlenecks
The Superbowl saga shows that Anthropic has chosen a different path, at least for now.
Its revenues are more heavily weighted towards enterprise contracts and paid subscriptions for Claude’s more capable models, giving it room to position itself as ‘ad-free’, and to use that stance as a differentiator while the market is still forming.
In this case, buying one of the most expensive advertising slots in the world to argue against advertising is less of a contradiction.
Anthropic is essentially telling users, regulators and enterprise customers where it intends to sit in the AI value chain, away from consumer media.
The tension also reflects how AI interfaces differ from previous platforms.
Ads embedded inside search results are globally familiar to users. But ads inside conversational tools, where users ask for advice on work, health or decisions, instantly raise questions about neutrality and liability.
And, even if answers are not directly influenced, the commercial context changes how outputs are perceived.
That concern is not limited to consumers, as businesses embedding generative AI into workflows should now think about governance and bias in a way that was less urgent when tools were funded primarily by subscriptions.
From OpenAI’s perspective, the risk cuts both ways. If it moves too fast, user trust erodes just as quickly.
But too slowly and infrastructure costs balloon while rivals search for revenue elsewhere.
The company’s careful framing, in calling the rollout a ‘test’, tightly controlling partner messaging, and limiting early metrics, proves it is acutely aware of these risks.