Sensor Tower, a mobile app analytics firm, puts OpenAI's chatbot at 46.4% of the category as of May 31, 2026, still the top ranked product but no longer a majority, and the first measurable challenger to a market OpenAI has owned since launch.
For two years, "AI assistant" has been a synonym for ChatGPT. That sentence stops being true on May 31, 2026, when Sensor Tower, a mobile app analytics firm whose estimates are widely cited but not independently audited, put ChatGPT's share of the AI assistant app category at 46.4% (Gizmodo, reporting the Sensor Tower data via TechCrunch). The product is still the top-ranked AI assistant by users. It is no longer a majority of the category it defined, and the share it lost did not drift away. It went to a named rival.
The rival is Anthropic, the maker of Claude, OpenAI's closest competitor and the company that has spent the last year selling enterprise customers on a different posture: a public refusal to help build autonomous weapons or mass domestic surveillance, even when that stance cost it access to the Pentagon. In March 2026, OpenAI took the opposite deal, signing with the Department of Defense while Anthropic was publicly feuding with the Trump administration over the same terms (Gizmodo). The Pentagon contract is not why ChatGPT slipped below 50%. It is part of the same story, the moment a category with one boss stopped behaving like a category with one boss.
The shift is structural, not a collapse. ChatGPT held more than half of the AI assistant market at the start of 2026, per the same Sensor Tower dataset, and the 46.4% figure as of late May is the first time the line has been crossed since the product launched in late 2022. A market moving from a single dominant vendor to a top vendor plus a credible number two is a normal stage of a software category maturing. It is also the stage where the leverage moves. When one company owns the category, pricing, defaults, and roadmap priorities are set inside that company. When a second company holds measurable share, every buyer has a reference price, every user has a fallback, and every product decision has to clear a public comparison.
The Sensor Tower number is the load-bearing claim here, and it carries the usual caveats for a single-vendor mobile-app dataset. The figure describes share of the AI assistant app category, not enterprise AI spending, not model API usage, and not chatbot traffic on the open web. It is one analytics provider's read of consumer mobile app installs and active users, surfaced by TechCrunch and re-reported by Gizmodo, with no independent corroboration in hand (Gizmodo). Treat 46.4% as the number on the record, not the number on the wall.
The number is more credible because the direction of travel matches the decisions OpenAI has been making in public. The company rolled advertising into ChatGPT in the first half of 2026, a product choice that enterprise buyers and privacy-sensitive users had asked it not to make. It shut down the Sora app, its short-form video product, in a move that read as a retreat from a consumer surface it could not defend. Inside the company, leadership has reportedly declared a "code red," OpenAI's internal shorthand for an emergency strategic shift, and pivoted engineering attention toward enterprise revenue (Gizmodo). Sam Altman has himself acknowledged, in public remarks, that the Pentagon deal was "opportunistic and sloppy," a rare mea culpa from a CEO whose company is still the market leader (Gizmodo). Each of these decisions has a defensible business rationale. Read together, they describe a company optimizing for the buyer who writes the largest checks, at the cost of the consumer who gave it the brand.
The QuitGPT campaign, a user-led push to delete ChatGPT accounts after the Pentagon deal, claimed four million sign-ups. That is the campaign's own number, not an audited figure, and it belongs to QuitGPT, not to the market (Gizmodo). What is auditable is that the same window in which QuitGPT claimed its four million also produced the first time ChatGPT dropped below 50% on Sensor Tower's read. Correlation is not causation, and the share drop is bigger than any single campaign. But the timing is the timing.
Anthropic is not a folk hero in this story, and the framing matters. The company has its own controversies, its own product stumbles, and a public stance on safety that critics read as marketing. Its rise is real and measurable, and it is also the rise of a competitor that, on the enterprise side, has been willing to lose deals to take positions ChatGPT will not. The competitive dynamic that produced a 46.4% ChatGPT share is not OpenAI losing. It is OpenAI meeting a buyer that is allowed to say no.
The phrase to retire is "just ChatGPT it," the way a generation retired "just Google it" once Google became a genericized trademark (a brand name so dominant it became the generic word for the product it sold), a brand name so dominant it became the generic word for the thing it sold. For two years, "AI assistant" and "ChatGPT" were the same noun in casual speech. As of late May 2026, they are not. Whether the next equilibrium is a durable duopoly, a rotating cast of category leaders, or a fragmenting field of specialized assistants is the question the next quarter of Sensor Tower data will answer. The first era of consumer AI ended not with a defeat but with a second name on the leaderboard.