Muse retention is suddenly the most-watched metric in the AI wars. New Sensor Tower data, shared on X on October 12 by Menlo Ventures partner Deedy Das, shows Meta’s personal AI agent bringing back 73.3% of new users the day after they install it — nearly twice the 41.3% day-one retention of OpenAI’s ChatGPT. The numbers, covering September 2026 across 81 countries, went viral within hours and have Wall Street talking.
Jim Cramer called Meta a “monster long” on the back of the data, and Meta shares climbed about 4% on October 12 as the company also launched its first paid frontier models, Muse Wonder and Muse Wonder Full. But the retention story has a twist: Muse’s early lead fades fast.
The numbers behind the buzz
Sensor Tower tracked new users of the major AI apps through their first 14 days. Here’s the scoreboard:
- Day 1: Muse 73.3%, ChatGPT 41.3%
- Day 7: Muse still ahead, with more than 30% retention
- Days 8–9: ChatGPT overtakes Muse
- Day 14: ChatGPT 26.8%, Muse 22.5%, Threads 21.2%, Google Maps 18.5%, Grok Bot 10.4%
For context, neither app is close to the most established apps at day 14: WhatsApp keeps 67%, YouTube 59%, Instagram 54%, and TikTok and Snapchat hold 37%. Still, only 23 apps have ever crossed 5 million US downloads within 22 days of launch, per Sensor Tower — Muse is one of them.
Similarweb puts Muse at 2.18 million daily active users as of October 5, with Sensor Tower reporting more than 5.1 million downloads by early October.
Why day-one retention is a money metric
Retention is the raw material of revenue. An app that brings users back on day one has a far cheaper path to subscriptions and ads. Meta’s own advertising machine helped: Sensor Tower found Muse captured up to 50% of Meta’s daily house-ad impressions across Facebook and WhatsApp in mid-September.
Wall Street noticed. CNBC host Jim Cramer pointed to the retention data as evidence Meta is a “monster long” position. Deutsche Bank analyst Benjamin Black estimates Muse could contribute as much as 8% of Meta’s 2030 revenue — roughly $36 billion at the high end, or $2.4 billion at the low end of his forecast.
The timing helped: the data landed the same day Meta unveiled Muse Wonder, its first closed and paid frontier model, with $20–$100/month Muse tiers and a $50/$200 Muse Code plan. Investors read hot app plus paid models as a monetization engine coming together.
The catch: Muse loses users fast
The impressive first-day figure tells only part of the story: Muse’s retention curve drops steeper, and by day 14 it sits in second place.
Why ChatGPT wins the long game
ChatGPT benefits from years of habit formation. Muse is an agent — it does errands like shopping and booking — which creates spectacular first impressions but also more chances to disappoint when a task fails.
The Amazon problem
Meta’s biggest structural obstacle is the open web itself: sites including Amazon have blocked AI agents from browsing and buying, cutting off exactly the errands Muse was built for.
What this means for you
If you’re an investor
Day-one retention of 73% is elite-tier app data. But the fade by day nine is the real signal: the investment case rests on Meta solving the agent-blockade problem and converting curiosity into paid Wonder subscriptions.
If you’re a creator or marketer
An app with 2+ million daily users and Meta’s distribution muscle is a new attention channel forming in real time. Early movers inside agent-driven shopping get the same advantage early Instagram advertisers got.
If you run a business
Watch which websites unblock agents. When major retailers open their doors to AI shoppers, agentic commerce flips from demo to revenue channel — and agent-ready catalogs go first.
The bigger picture: the agent wars just got real
Meta’s Muse overtook ChatGPT as the top free US App Store app within days of its September 8 launch. OpenAI has answered with Dots, its own always-on agent, and Meta with paid frontier models. This retention data is the first hard evidence of how the two compare in users’ hands.
FAQ
What is Muse’s day-one retention rate?
73.3%, according to Sensor Tower data for September 2026 shared by Menlo Ventures partner Deedy Das — nearly double ChatGPT’s 41.3%.
Is Muse beating ChatGPT?
On day one, yes — by a wide margin. But ChatGPT overtakes Muse between day eight and day nine, and leads 26.8% to 22.5% by day 14.
Why does Muse lose users after the first week?
Agents create great first impressions but face real-world friction: tasks fail, and major websites like Amazon block AI agents from browsing and buying. ChatGPT’s simpler chat habit proves stickier over time.
How many people actually use Muse?
Similarweb estimated 2.18 million daily active users on October 5, 2026, while Sensor Tower reported more than 5.1 million downloads by early October.
Could Muse make serious money for Meta?
Deutsche Bank analyst Benjamin Black estimates Muse could drive up to 8% of Meta’s 2030 revenue — about $36 billion at the high end, $2.4 billion at the low end. Meta just launched $20–$100/month Muse subscriptions to start converting users into revenue.
Sources: Sensor Tower data via Benzinga’s report on the retention numbers and Jim Cramer’s reaction, TipRanks’ breakdown of the day-by-day figures, IA en un minuto’s coverage of the Sensor Tower data, and Meta’s official Muse product site.