Technology

RevenueCat data shows AI apps churn faster

subscription economics weaken as models become interchangeable, conversion gains fail to translate into retention

Images

REvenuecat: AI vs Non-AI apps by categoryImage Credits:RevenueCat REvenuecat: AI vs Non-AI apps by categoryImage Credits:RevenueCat Image Credits:RevenueCat
Image Credits:RevenueCat Image Credits:RevenueCat Image Credits:RevenueCat
AI vs non-AI apps by subscription plan typeImage Credits:RevenueCat AI vs non-AI apps by subscription plan typeImage Credits:RevenueCat Image Credits:RevenueCat
ScreenshotImage Credits:RevenueCat ScreenshotImage Credits:RevenueCat Image Credits:RevenueCat
Sarah Perez Sarah Perez techcrunch.com

AI-powered subscription apps are converting users faster but keeping them for less time, according to new data from RevenueCat, a subscription management provider used by more than 75,000 developers. In its 2026 State of Subscription Apps Report, RevenueCat says AI apps churn annual subscribers about 30% faster than non‑AI apps at the median, while showing weaker retention at both monthly and annual horizons.

The dataset is large enough to matter: RevenueCat says its tooling manages more than one billion in-app transactions and helps developers generate more than $11 billion in annual revenue. Within that universe, AI-powered apps now account for 27.1% of apps across categories, TechCrunch reports, with the highest concentration in photo and video (61.4%) and the lowest in gaming (6.2%).

The pattern in the metrics is awkward for the popular idea that “AI” itself is a defensible moat. AI apps do better at the top of the funnel—trial-to-paid conversion is 8.5% versus 5.6% for non‑AI apps, and monetisation per download is slightly higher (2.4% versus 2.0%). But the back end deteriorates: annual retention is 21.1% for AI apps versus 30.7% for non‑AI apps, and monthly retention is 6.1% versus 9.5%. Refund rates are also higher—4.2% versus 3.5%—with a wider upper bound, suggesting more volatility in realised revenue.

One explanation is simply substitution. Many “AI apps” are thin wrappers around interchangeable foundation models, and users can switch when a competitor launches a cleaner interface, a cheaper plan, or access to a newer model. If switching costs are close to zero, churn becomes rational behaviour: subscribe for a burst of use, cancel, and rotate to the next best offer. That makes pricing power hard to sustain unless the app owns something harder to copy—proprietary data, a distribution channel, or deep workflow integration that makes leaving painful.

The report’s numbers also hint at a business model mismatch. AI features are often most compelling in the first week—when a user tests image generation, summarisation, or chat—then fade into a commodity utility. Subscription economics, by contrast, depend on habitual use and predictable renewal. Developers can buy installs and juice conversions, but they cannot buy long-term product differentiation if the core capability is rented from the same model providers.

RevenueCat’s data suggests the AI gold rush is real at launch, and less reliable at renewal.

At the median, AI apps are generating higher lifetime value per paying user while losing those users sooner—a combination that looks profitable until acquisition costs rise or competitors cut prices.