Daily active users of the Gen AI Chatbot app market in the US have contracted 1.5% month-over-month in April. DAUs still remain elevated 3.8% from February. The only apps consistently growing in 2026 are Google Gemini and Claude.
Claude has moved from a market share of 1.5% in January 2026 to a share of 13.1% in April. Google Gemini has roughly maintained a 25% market share this calendar year. ChatGPT has fallen from a 45.3% share in January to a 38.1% share in April.
Looking globally and adding in three more apps (Microsoft 365 Copilot, Le Chat by Mistral AI, Meta AI), we see continued year-over-year growth for DAUs but that growth is decelerating.
Overall time spent on AI apps (in the United States) as a percentage of all time spent on mobile devices is growing (see below).
Engagement Flow Trends
AI usage on mobile is bifurcating, and the gap between high frequency users and everyone else is widening fast. Apptopia’s mobile app data shows that high-frequency AI users grew from 6.8% to 8.5% of the user base between December ’25 and April ’26, a 25% relative increase in four months and the fastest-growing segment in the data.
Once a user gets to that level of engagement, they almost never leave. About 68% of high-frequency users stay high-frequency the following month, and effectively zero churn outright. When they do step back, they downshift to medium usage rather than dropping AI entirely. That is the cohort investors should be focused on, because it behaves more like a locked-in subscriber base than a casual app audience.
The middle of the funnel is where the story gets messier. Only about a third of medium-frequency users stay medium each month. Roughly half downgrade to low usage or churn, and only 17% upgrade to high-frequency. The medium tier is volatile, with the share dropping straight to churn jumping to 22.7% in February before recovering. It is also the single most important feeder into the power-user cohort, supplying roughly 23% of the high-frequency pool every month.
Below that, the picture is worse. Low-frequency retention deteriorated steadily from 28.7% to 22.9% across the four periods, and the share of low-frequency users churning each month rose from 46.9% to 50.0%. Half of casual AI users now drop off in any given month.
Even with that leakage at the bottom, total adoption is still expanding. The never-AI share of the base shrank from 42.3% to 40.3%, with about 9 to 10% of never-users trying AI in any given month. Total active AI users grew from 32.1% to 35.0% of the base over the period. The rest are churned users. Churn is more reversible than one may think, with 24 to 25% of churned users returning to AI usage each month, mostly at the low-frequency level.
