AI App Economy: Store Charts, Revenue and Churn — 2026-09-25

Top developments

AI subscription churn remains the sector's structural weakness

RevenueCat's 2026 benchmark data continues to frame the AI app monetization debate: AI-powered apps generate 41% more revenue per payer but churn 30% faster than traditional apps, with 12-month payer retention on AI monthly plans 36% worse over a year. Median refund rates run about 20% higher for AI apps (4.2% vs 3.5%), and 12-month retention sits at roughly 21.1% for AI apps versus 30.7% for non-AI apps. For anyone tracking app-economics dashboards, the takeaway is that acquisition economics for AI subscriptions remain fragile despite strong topline revenue. |

Higgsfield claims $1B revenue run rate as gen-video rivals burn cash

Video-generation startup Higgsfield told Startup Fortune it is on pace to hit a $1 billion revenue run rate within a year while remaining cash-flow positive — a contrast to competitors Runway and Pika, which it says are still burning money. If accurate, this would place one of the youngest gen-media consumer apps among the highest-grossing AI subscription businesses, and comes amid broader VC skepticism about how AI startups count recurring revenue (Business Insider, ~2 weeks ago, reported VCs calling the math "murky").

In-app events firmly a ranking lever for Apple and Google in 2026

SEM Nexus, publishing yesterday, details how in-app events now function as App Store and Google Play ranking signals in 2026, with event calendars and cadence affecting algorithmic placement. For AI app marketers competing for chart visibility against incumbents like ChatGPT — which held the US App Store's #1 top-grossing slot for 310 of the last 350 days per AppRadar's August analysis — structured event planning is becoming a required ASO tactic. |

JPMorgan turns bullish on Meta's AI app

JPMorgan published a fresh, strongly positive take on Meta stock built around the performance of its newest AI consumer app, having softened its stance earlier this year; the bank now sees a "historic run" ahead. Meta's Meta AI app is thus joining the ranks of consumer AI apps with visible Wall Street earnings-model implications — a signal worth tracking for top-grossing chart movement.

Local view

FX168's weekly AI-app tracker (covering 2026.09.14–09.20, published ~5 days ago) leads with ByteDance updating Doubao's large model to version 2.1 Pro (0915 build) and launching a Doubao in-car assistant, keeping Doubao's push into consumer-grade AI traffic and ecosystem front and center.

Chinese bargain-hunting community Smzdm (2 days ago) catalogs the latest free AI-product perks as of September 23: Tencent's new Hunyuan image model rolled out as free image generation in Yuanbao plus two-week trials of three pro tools, while Sichuan province institutionalized token vouchers and Jiangsu enterprises can claim Doubao work benefits. B.AI (2 days ago) announced that from September 25 at 15:00 SGT, new MiMoV2.6 models take over discount campaigns — MiMoV2.6 Flash at 90% off and flagship Pro at 50% off for launch.

Context & numbers

• Majority-ranking context: ChatGPT was the US App Store's #1 top-grossing app for 310 of the last 350 days as of AppRadar's August 2026 analysis.
• Global in-app consumer spend reached $171 billion in 2025 (+10.7% YoY), with ChatGPT the third-highest grossing app of 2025 behind TikTok and Google One.
• Consumer AI adoption keeps deepening: Menlo Ventures' State of Consumer AI report (published this week) finds consumers spending more time and money with AI, across more tasks, and increasingly delegating authority to AI agents.

On the radar

• Weekly Chinese AI-app watch: FX168's Monday tracker is worth following for Doubao/Yuanbao/Qianwen download and ad-spend moves.
• Zhihu's LLM tracker (updated 4 days ago) flags a burst of agent-model updates across GPT-6 Sol/Luna, Claude Opus 5.5 and Grok 4.7 — expect downstream app feature pushes that could move chart positions.
• Watch for whether Higgsfield's $1B run-rate claim draws independent verification or commission-estimate cross-checks from Sensor Tower/Appfigures-style trackers.