A Billion Users and a Shrinking Share: How Both Are True
ChatGPT's simultaneous achievement of one billion monthly active app users and its first sub-50% market share is not a paradox — it is an arithmetic consequence of category explosion. Sensor Tower's State of AI data places total AI app engagement at roughly 36 billion hours for the half , which means the pool of AI assistant users grew faster than any single product could hold. ChatGPT got there faster than any prior app in history, passing Google Maps, TikTok, Instagram, and YouTube , and still ended May as a minority player within the category it invented.
That structural position — category creator now in market-share decline — is the one OpenAI's growth narrative has never had to account for. The billion-user milestone was the story in early June; the 46.4% figure is the story that follows it, and the two together describe a company whose absolute scale is still growing while its relative grip is loosening faster than its product roadmap anticipated.
The DoD Deal as a Durable Trust Threshold
ChatGPT uninstalls rose 132% in April compared to the prior year , and the timing maps directly to the period when OpenAI's Department of Defense relationship became broadly publicized. A user on Bluesky described the mechanism plainly: "Trust is now a variable users act on" — not a passive sentiment, but a behavioral trigger with measurable consequences in Sensor Tower's uninstall data.
This matters because trust-driven exits operate differently than feature-driven churn. A user who leaves because a competitor launched a better model is recoverable when OpenAI ships the next version. A user who leaves because they concluded the tool has institutional allegiances incompatible with their own priorities is not recovered by a GPT-5o announcement. Each subsequent government contract OpenAI signs adds to the evidence base that drove the original exit decision — making this a compounding problem, not a one-cycle news story. OpenAI's monetization strategy and its consumer retention strategy are now structurally in tension in a way the pre-May data did not reveal.
Gemini and Claude Inherited a Push, Not a Win
The competitive reading of the Sensor Tower numbers is seductive but partially wrong. Gemini's rise to 27.7% and Claude's to 10.3% look like wins earned through superior product development, but the growth dynamic is better described as inherited displacement. Both challengers benefited from a push force — the DoD deal and the associated uninstall spike — that arrived independently of anything Anthropic or Google built in their respective models.
Claude's trajectory is particularly telling. Its 10.3% share accrued without the consumer advertising infrastructure OpenAI deploys, through a combination of developer-community trust and enterprise relationships that predated any mass consumer moment. Anthropic's navigation of its own institutional pressures in Washington has, at least in the Sensor Tower reading, translated into a public positioning that consumer users read as less entangled with government priorities. Whether that positioning holds as Anthropic's own government engagements deepen is the question the market-share data cannot answer — but forces into view.
What the Market-Share Ceiling Tells Enterprise Buyers
For enterprise procurement teams, the sub-50% reading introduces a vendor-risk dimension that was absent from the original calculus. A category with a dominant majority-share player is one where standardizing on that player carries manageable institutional risk. A category where the former leader lost its majority through a trust event tied to a specific government contract — not a product failure — is one where tool selection carries reputational and political dimensions that a feature evaluation cannot capture.
The Sensor Tower data arrives at the same moment that AI is reshaping employment patterns across industries, which means enterprises are simultaneously accelerating AI adoption and watching the market-structure assumptions that justified their initial tool choices shift underneath them. The compliance and procurement teams that locked in ChatGPT integrations before May now manage a tool whose institutional associations are a documented source of user attrition. That calculation will appear in the next round of renewal conversations — and unlike a pricing dispute, it cannot be resolved by OpenAI offering a discount.
OpenAI's Position After the Threshold
OpenAI's situation after the Sensor Tower report is not catastrophic — 46.4% of a rapidly expanding category remains an enormous user base — but the direction of travel is the story. The billion-user milestone gave OpenAI a headline that obscures the trend; the market-share data reveals it. The company is simultaneously the largest AI consumer product ever built and a product whose users are departing at an elevated rate for reasons unrelated to model quality.
The Forrester comparison to AI's BlackBerry identified the risk abstractly. The Sensor Tower data puts a specific mechanism on it: OpenAI's government-facing revenue strategy is generating consumer-facing attrition, and Gemini and Claude are the direct beneficiaries. The enterprises and individual users who left on trust grounds are already inside Anthropic's and Google's products — and the model updates OpenAI ships will not bring them back.