AI is rapidly reshaping the app economy, but new data suggests that adding artificial intelligence is far from a guaranteed path to durable subscription revenue.

RevenueCat, whose infrastructure processes more than a billion in-app transactions for tens of thousands of developers, analyzed performance across iOS, Android, and the web. Its latest State of Subscription Apps report finds that while AI apps are strong at converting users into paying customers, they are markedly weaker at keeping those subscribers over time.

Across the dataset, AI-powered apps see annual subscribers cancel about 30 percent faster than subscribers to non-AI apps. At the 12-month mark, only 21.1 percent of AI subscribers are still paying, compared with 30.7 percent for non-AI apps. Monthly retention shows a similar gap, at 6.1 percent for AI apps versus 9.5 percent for others.

The only bright spot on the retention front is at the weekly level, where AI apps slightly outperform, with 2.5 percent weekly retention compared with 1.7 percent for non-AI products. But weekly plans are a minority of AI subscriptions, limiting the impact of that advantage.

Despite these churn problems, AI apps are powerful revenue engines in the short term. They convert free trials to paid subscriptions 52 percent better than non-AI apps, and they monetize downloads roughly 20 percent more effectively. Their realized lifetime value is also higher: a median of $18.92 per paying user per month, versus $13.59 for non-AI apps, and $30.16 versus $21.37 on an annualized basis.

AI is also spreading unevenly across categories. Photo and video apps are the most AI-heavy segment, with 61.4 percent of titles marketing themselves as AI-powered, reflecting the boom in AI filters, image generation, and editing tools. Gaming remains relatively untouched at 6.2 percent, while travel and business apps also show low AI penetration.

The report points to several likely reasons for the retention gap. The underlying technology is evolving so quickly that users frequently jump between competing apps in search of better models, features, or pricing. Refund behavior underscores that instability: AI apps have a 20 percent higher median refund rate than non-AI apps, and a higher upper bound, suggesting more volatility in realized revenue and lingering questions about long-term user value and product quality.

The emerging picture is nuanced. AI can supercharge acquisition and early monetization, but sustaining trust and habit appears harder. For developers, the message is clear: AI may open the door to revenue, but product depth, reliability, and ongoing innovation still determine who gets to keep it.