New Study Finds AI Assistant Market Isn’t Zero-Sum as ChatGPT, Gemini and Claude All Grow

Researchers identify three winning strategies—scale, ecosystem and premium—and warn that today’s coexistence may not last as the market matures.

For years, the dominant storyline in artificial intelligence has been a familiar one: a high‑stakes, winner‑take‑all race among a handful of large language model (LLM) providers.

But a new analysis from researchers at the University of Maryland’s Robert H. Smith School of Business, McGill University and USC Marshall School of Business suggests that this framing misses what is actually happening in the global AI assistant market. Rather than fighting over a fixed pie, the major players appear to be expanding it.

The study, “Three Strategic Bets on AI’s Future,” tracks worldwide mobile‑app usage data from May 2023 through December 2025. Its central finding: every major LLM release over the past two years has grown the overall market, with little evidence that one model’s gains come at the expense of another. The researchers identify three distinct strategies — scale, ecosystem and premium — that are all succeeding simultaneously.

ChatGPT dominates through a scale strategy, capturing roughly 79% of the market with moderate revenue per user. Google Gemini follows an ecosystem strategy, holding about 12% share while generating minimal direct revenue by leveraging Google’s broader platform. Claude, meanwhile, pursues a premium strategy, commanding under 1% market share but earning more than 40 times the revenue per user of Gemini.

Most surprisingly, say the researchers, was how consistently the market expanded in response to new model launches. “We expected to see more of a zero‑sum dynamic where one platform’s gain came at a competitor’s expense,” says Smith School Associate Professor of Marketing Daniel McCarthy, who co-authored the work with Maxime C. Cohen and Eddy Hage-Youssef (McGill) and Daniel Sokol (USC). “Instead, every major model launch we tracked coincided with growth across the whole market. The ‘AI war’ framing suggests that the pie is more fixed than it actually is. Our data suggests the pie is growing fast enough that everyone is gaining at the same time.”

McCarthy notes that this pattern has only become more visible since the paper was posted. He pointed to the “QuitGPT” movement in late February, when millions of users migrated toward Claude after OpenAI’s Pentagon deal. “On the surface, it looked like a zero‑sum shift,” he said. “But it also generated a massive wave of new attention to AI assistants broadly, such that overall OpenAI active users were basically flat in the immediate aftermath, while Anthropic active users went way up. That’s the expansion effect playing out in real time.”

Three Strategies, Three Vulnerabilities
While the coexistence of these strategies is notable, the authors caution that it may not last. The conditions enabling all three to thrive — rapid market expansion, abundant capital, and unsettled user preferences — are temporary. As growth slows or capital tightens, each strategy faces distinct vulnerabilities.

Scale players like OpenAI must eventually convert free users into paying customers. Ecosystem players like Google must justify continued subsidies. Premium players like Anthropic must maintain differentiation as competitors improve.

McCarthy believes the market is overlooking an important strength of OpenAI. “People underestimate the upside of the scale player, OpenAI,” he says. “Everyone looks at their cash burn and sees risk. But OpenAI has hundreds of millions of free users who are completely undermonetized. If they wanted to, they could convert even a small fraction of that base into paying subscribers, or ad‑supported revenue, and dramatically improve their economics. They have not played that card to date, but it is a card that is readily available to them that they have strategically decided not to play yet.”

Claude, by contrast, already monetizes extremely well. “That also means Anthropic has already played the monetization card pretty heavily,” McCarthy notes. “Their ability to ‘flip the switch’ and generate significantly more revenue per user is more limited.” He pointed to recent usage throttling during peak hours as evidence of the tension inherent in a premium niche: maintaining high quality and access is expensive, especially as demand surges.

How Leaders Should Navigate an Unsettled Market
For executives deciding which AI strategy fits their organization, McCarthy says that the most important question is deceptively simple: What does your organization structurally need from AI, and what tradeoffs are you willing to accept?

Claude offers strong performance for coding and document‑heavy workflows but comes with higher costs, a more limited range of services, and occasional reliability constraints. ChatGPT provides broader functionality, better reliability, and a larger ecosystem, but may face pricing shifts as monetization ramps up. Gemini integrates deeply with Google’s products but remains less differentiated in direct AI capabilities.

“The honest answer is that the market is still too unsettled for anyone to commit fully to a single provider,” McCarthy says. “Diversifying across tools, or at least staying flexible, is probably the smartest near‑term approach.

A Market Still in Motion
Since the paper’s release, the dynamics it describes have only intensified. Anthropic has tightened usage caps as demand outpaces infrastructure. OpenAI has introduced ads and a lower‑cost subscription tier. And the QuitGPT episode offered a dramatic real‑world test of the expansion effect: Claude surged to the top of the App Store, but the overall category grew as well.

“I think one important takeaway for leaders is that it is very hard to know how the capabilities of the major players will evolve, and what this implies for the ultimate success or failure of these strikingly different go-to-market strategies,” McCarthy says. “For a very long time, Google had the dominant image generation model in Nano Banana Pro, and then OpenAI quickly and decisively moved ahead of it last week with their ImageGen 2.0 model. The experiment is still very much underway.”

The researchers argue that rather than betting on a single outcome, organizations should align their AI strategy with their structural advantages and stay ready to adapt. The market may ultimately support multiple durable approaches, or one may dominate as conditions shift. For now, a clear conclusion is that the AI assistant landscape is far more dynamic and far less zero‑sum than the prevailing narrative suggests.

Read the paper, “Three Strategic Bets on AI’s Future” at SSRN.