Strategic Fit Beats Market Share: Surprising Smart AI Study

Strategic fit, not market share, is what separates the winners in AI. That is the central finding of a study published in California Management Review by researchers from the University of Maryland, McGill University and USC Marshall, who tracked the worldwide mobile usage of ChatGPT, Gemini and Claude from May 2023 to December 2025. The three assistants hold wildly different shares of the market — 78.4%, 12.5% and 0.5% of mobile daily active users respectively — yet the study concludes that all three are winning, because each has chosen a position that matches its structural strengths.

That conclusion cuts against most of the coverage the artificial intelligence race receives. League tables of users and downloads imply a single scoreboard, and an eventual single champion. The researchers argue the scoreboard itself is misleading: the real test is strategic fit — whether the position a firm pursues lines up with the assets, distribution and economics it already has. A company can trail badly on share and still be executing the best AI strategy available to it.

This article unpacks what the study measured, why its authors say the AI market is not zero-sum, what strategic fit means in practice for the three model providers, and — because most readers are buyers rather than builders of AI models — what the same logic says about how your own organisation should choose its position.

Table of contents

Inside the Strategic Fit Study
Strategic Fit vs Market Share: Three Very Different AI Winners
Why the AI Market Is Not Zero-Sum
What Strategic Fit Means in Practice
Strategic Fit Lessons From Enterprise AI
How the Market Has Moved Since the Study
How to Apply Strategic Fit to Your Own AI Decisions
FAQ: Strategic Fit and the AI Winners
References

Inside the Strategic Fit Study

The paper behind the headline is “Three Winning AI Strategies”, written by Daniel McCarthy, associate professor of marketing at the University of Maryland’s Robert H. Smith School of Business, with Maxime C. Cohen and Eddy Hage-Youssef of McGill University and D. Daniel Sokol of USC Marshall. It reached a wider audience this week when the university’s summary was republished under the headline that strategic fit, not market share, is the key to AI success.

What the researchers measured

The team used worldwide mobile app usage estimates from Sensor Tower covering May 2023 through December 2025, and ran an event study across 15 major model releases in that window. For each release they measured what happened to the launching firm’s own usage and to its competitors’ usage. The design matters: rather than asking who is biggest, it asks whether one firm’s gains actually come at another’s expense — the assumption baked into every “AI war” headline.

The single most surprising number

Across the 15 releases, the effect on competitors averaged just minus 0.1%, while the effect on the launching firm’s own usage averaged plus 7.0%. Big launches, in other words, grew the launching assistant substantially while barely denting anyone else. “We expected to see more of a zero-sum dynamic where one platform’s gain came at a competitor’s expense,” McCarthy said. “Instead, every major model launch we tracked coincided with growth across the whole market.”

The study’s share figures make the case for strategic fit vividly, because the gaps are so extreme. ChatGPT held 78.4% of worldwide mobile daily active users. Gemini held 12.5%. Claude held just 0.5% — and yet earned revenue per active user more than 40 times Gemini’s and roughly 3 times ChatGPT’s. On the market-share scoreboard, two of the three look like also-rans. On the strategic fit scoreboard, all three are executing coherent, defensible positions.

Worldwide mobile daily active user share, study window to Dec 2025

ChatGPT 78.4%

Gemini 12.5%

Claude 0.5%

Scale leader: ChatGPT

OpenAI’s position is the classic scale play: serve everyone, monetise moderately per head, and let volume do the work. McCarthy’s shorthand is blunt: “ChatGPT is selling to everyone.” He also cautions against writing the position off as unimaginative — “People underestimate the upside of the scale player, OpenAI.” Scale brings data, brand default status and distribution gravity that compound over time.

Ecosystem player: Gemini

Gemini’s 12.5% share generates minimal direct revenue, and the study argues that is the point. Google monetises across complementary products — search, Workspace, Android, cloud — and Gemini’s job is to protect and extend that estate. “Google is giving Gemini away to protect its search business,” as McCarthy puts it. Judged on assistant revenue, Gemini looks weak; judged on strategic fit with Google’s ecosystem, it is doing exactly what it should.

Premium niche: Claude

Anthropic’s Claude serves a small professional base with high willingness to pay — the premium niche position. Its 0.5% share would be fatal for a scale player, but its revenue per user multiple shows a different game being won. McCarthy adds a caution here too: “That also means Anthropic has already played the monetization card pretty heavily.” A premium position that has already harvested much of its pricing power must keep earning it with capability.

Position	Example	Mobile DAU share	How value is captured
Scale leader	ChatGPT	78.4%	Moderate revenue per user across a very large base
Ecosystem player	Gemini	12.5%	Minimal direct revenue; value captured across Google products
Premium niche	Claude	0.5%	Revenue per user 40x Gemini’s and about 3x ChatGPT’s

Why the AI Market Is Not Zero-Sum

The event-study numbers explain why three such different positions can all win at once: the market is expanding faster than any rivalry can shrink it. When a major model ships, the launching firm gains about 7% in usage while competitors lose almost nothing. New capability pulls new users and new use cases into the category rather than shuffling a fixed pool between vendors.

Average usage effect of a major model launch, 15 releases 2023–2025

Launching firm’s own usage +7.0%

Competitors’ usage −0.1%

McCarthy frames the implication directly: “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.” In a growing market, obsessing over relative share is a distraction from the question strategic fit actually asks — are you capturing the kind of value your structure lets you capture?

That does not mean rivalry is irrelevant forever. Growth phases end, and positions harden. The study’s argument is narrower and more useful: right now, a firm’s results are explained far better by strategic fit than by its rank on the share table, and decisions made to chase rank — discounting a premium product, or bolting a consumer app onto an enterprise franchise — actively destroy the fit that was working.

There is a practical corollary for anyone reading vendor announcements. A rival’s flagship launch is weak evidence that your chosen provider is losing: over the study window, the biggest releases in the category moved competitors’ usage by a rounding error. Judging providers by launch-week headlines therefore measures marketing noise, not strategic fit. The durable signal is whether each provider keeps deepening the position its structure supports — shipping enterprise controls if it is courting professionals, or distribution deals if it is playing for scale.

What Strategic Fit Means in Practice

The researchers are precise about the term. Strategic fit is not a synonym for strategy in general; it is the alignment between the position a firm pursues and the structural advantages it brings — distribution, complementary products, cost position, brand permission and existing customer relationships. The study’s sharpest sentence is worth quoting whole: “The strategic error is not choosing the ‘wrong’ universal model—it’s pursuing a position that doesn’t fit the organization’s strengths.”

Three questions that define the position

The paper distils strategic fit into three diagnostic questions, one per winning position. A firm that can answer yes to one of them honestly has found its lane; a firm that answers yes to none of them is competing on hope.

Position	Diagnostic question from the study	What it demands
Scale leader	“Do we have a credible path to serving and monetizing a very large user base?”	Capital, infrastructure and mass-market brand permission
Ecosystem player	“Can our distribution or complementary offerings capture value the focal product does not?”	An estate of adjacent products the assistant strengthens
Premium niche	“Can we solve an important problem well enough to command higher willingness to pay?”	Deep capability on problems a defined audience values highly

Fit is a constraint, not a preference

Notice what the questions have in common: none of them asks what management would like to be true. Strategic fit is discovered by auditing what the organisation already has, then choosing the position those assets can defend. Gemini could not credibly run Claude’s premium play — Google’s economics and brand pull it towards ubiquity. Anthropic could not credibly run the scale play against two rivals with vastly bigger distribution. Each firm’s constraint became its strategy.

Strategic Fit Lessons From Enterprise AI

The study covers the consumer assistant market, but its logic lands hardest in the enterprise, where the record of AI projects is famously uneven. MIT’s widely cited 2025 report on generative AI in business found that only about 5% of AI pilot programs achieve rapid revenue acceleration, while the vast majority stall — the “95% of pilots fail” statistic that dominated boardroom conversations for a year. Read through a strategic fit lens, that failure rate looks less like a technology problem and more like a fit problem: projects chosen because AI is fashionable, not because they align with what the organisation is structurally placed to do.

The same MIT research offers a striking supporting number: purchased specialist tools succeeded about 67% of the time, while internally built systems succeeded only around a third as often. Most companies’ structural advantage lies in their domain, data and customer relationships — not in building AI models — so buying capability and fitting it to the business shows better results than imitating a frontier lab.

Enterprise generative AI outcomes, MIT 2025 figures

Pilots achieving rapid revenue acceleration 5%

Pilots that stall 95%

Purchased specialist tools that succeed 67%

Enterprise buying behaviour echoes the not-zero-sum finding too. Andreessen Horowitz’s January 2026 survey of enterprise AI adoption found average spend on model APIs rising from $4.5 million to $7 million year on year, with $11.6 million projected next — and 81% of enterprises running three or more model families at once. Even inside single companies, multiple providers are winning simultaneously, each where its strengths fit the workload. That mix is not indecision; it is strategic fit applied at the level of individual workloads.

Study	What it examined	Headline finding
Three Winning AI Strategies (California Management Review)	Consumer assistant usage, May 2023–Dec 2025, 15 launches	Strategic fit beats market share; launches grow the whole market
MIT, The GenAI Divide (2025)	300 deployments, 150 leader interviews, 350 employees	About 5% of pilots accelerate revenue; buying beats building
Andreessen Horowitz enterprise survey (Jan 2026)	Enterprise model spend and provider mix	Spend rising toward $11.6M average; 81% use 3+ model families

How the Market Has Moved Since the Study

The study’s data window closes in December 2025, and the share picture has kept moving since — which, if anything, strengthens its thesis. Sensor Tower’s State of AI 2026 reporting showed ChatGPT’s share of assistant usage falling below 50% on broader audience measures by mid-2026, with Gemini climbing to roughly 27.7% and around 662 million users, and Claude reaching about 10.3% and roughly 245 million. One careful caveat: those 2026 figures use web and cross-platform audience metrics, while the study’s 78.4% figure counts mobile daily active users — different denominators that should never be mixed in one chart.

The monetisation gap, however, persisted in exactly the direction the study describes. On 2026 US mobile figures, Claude’s revenue per user ran at $2.76 against ChatGPT’s $1.74, with a reported 13% paid conversion rate — the premium niche still earning multiples of the scale player per head. Share moved; strategic fit endured. That is precisely the study’s point: positions built on structural advantage survive the volatility of the league table. For ongoing coverage of the models themselves, our AI models, tools and releases hub tracks each provider’s releases as they land.

The lesson for forecasting is humility about share and confidence about structure. Nothing in the 2026 numbers changed who the scale player, the ecosystem player and the premium niche player are — the market simply grew around all three, exactly as the strategic fit framing predicts. Anyone who sold their view of Anthropic on a 0.5% share figure in 2025, or their view of OpenAI on a sub-50% figure in 2026, was reading the wrong column.

How to Apply Strategic Fit to Your Own AI Decisions

Most organisations reading this study are not building frontier models. But strategic fit applies with equal force to AI adoption, and McCarthy points buyers to the same discipline: “What does your organization structurally need from AI, and what tradeoffs are you willing to accept?” The honest starting point is an audit of your structural position, not a tour of vendor demos — the same sequencing we use in our AI strategy engagements.

Choose your position before your provider

The three positions translate directly into adoption postures. A scale posture standardises one assistant across the whole workforce for breadth and price. An ecosystem posture picks the AI embedded in the platforms you already run — Microsoft 365, Google Workspace, your CRM — so value accrues to systems you own. A premium posture pays more for the strongest model on the narrow tasks where quality is worth multiples: code, legal drafting, research. The wrong choice is not any one of these; it is a posture mismatched to your structure, or, as with any digital strategy, one copied from a company with different strengths.

Sizing that choice against your estate matters more than benchmark scores. A firm that runs on Google Workspace has ecosystem gravity no benchmark can see; a consultancy billing for premium expertise has the same natural claim to a premium tool. Strategic fit for buyers means letting those structural facts, rather than launch events, set the default — and revisiting the default on your renewal cycle, not the news cycle.

Expect a multi-model world

The study also cautions against premature commitment. “The honest answer is that the market is still too unsettled for anyone to commit fully to a single provider,” McCarthy says. The 81% of enterprises already running three or more model families are not being indecisive — they are matching each workload to the provider whose strategic fit serves it best, keeping large language model choices reversible while the market keeps growing. Treat AI models as a portfolio to be rebalanced, not a marriage.

Judge success by fit, not by headlines

Finally, apply the study’s scoreboard lesson internally. An automation programme’s success is not whether it uses the biggest-name model, but whether it compounds an advantage you already hold — your data, your workflows, your customer relationships. Chasing whatever artificial intelligence capability is loudest this quarter is the corporate version of chasing market share: activity that photographs well and fits nothing.

FAQ: Strategic Fit and the AI Winners

What does strategic fit mean in the AI market?

In the study’s usage, strategic fit is the alignment between the competitive position a firm pursues and the structural advantages it actually holds — distribution, complementary products, capital and brand. ChatGPT fits a scale position, Gemini fits an ecosystem position, Claude fits a premium niche. The error is not picking the “wrong” model; it is pursuing a position your organisation’s strengths cannot support.

Does market share matter at all for AI companies?

Share still matters as an input — scale brings data, defaults and distribution gravity. What the study rejects is share as the scoreboard. Claude’s 0.5% share with 40 times Gemini’s revenue per user is a winning position, and Gemini’s minimal direct revenue with 12.5% share is a winning position too, because each matches its owner’s structure.

Is the AI assistant market really not zero-sum?

Across the 15 major releases studied, launches grew the launching firm’s usage by an average of 7.0% while competitors lost just 0.1%. Growth is coming from category expansion, not from raiding rivals. The authors are careful to say this describes the current growth phase — consolidation pressures can return as markets mature.

How should a business use strategic fit when buying AI?

Ask the study’s three questions of yourself, not of vendors: whether you need breadth across everyone, leverage inside platforms you already own, or premium capability on a few high-value tasks. Then match providers to workloads, keep the choice reversible, and measure success by whether the deployment compounds an existing strength rather than by the logo on the tool.