Meta’s AI companion app Muse launched on September 8. In its first two and a half weeks on the market, downloads have crossed 3.4 million, according to estimates from Sensor Tower that TechCrunch reported Thursday, September 24. That is up from 2.5 million just a week earlier. Two rival tracking firms put the number even higher or lower: Apptopia estimates 4.3 million total downloads, while Appfigures puts it closer to 2.3 million, a reminder that third party install estimates vary widely and should be read as ranges, not counts.
The growth curve is the more striking number. Sensor Tower’s data shows downloads climbing at a 55 percent daily clip through Muse’s first ten days on the market, compared with a 24 percent daily rate for ChatGPT across its own opening ten days. Downloads for Claude and Grok, by contrast, declined during their respective launch windows, per the same firm. Muse took the top spot on the U.S. App Store on September 18 and the top spot on Google Play the next day, and has not given up either ranking since.
Much of that momentum traces to Meta Connect, the company’s annual developer conference held this week, where Meta laid out what’s coming next for Muse: an avatar you can talk to over video, computer use on Mac, its own dedicated email address, more third party connectors, and eventual availability on Meta’s smart glasses. Sensor Tower recorded a 27 percent jump in daily active users the day after the conference wrapped.
What makes Muse’s climb different from a typical AI app launch is how little of it Meta is paying for directly. Ads promoting Muse accounted for only 6 percent of the app’s impressions between launch and September 19, according to Sensor Tower. Instead, Meta is running the same playbook that pushed Instagram Threads past 500 million monthly active users: cross-promotion to the billions of people already inside Facebook and Instagram. Meta began serving house ads for Muse on September 9, and within ten days those house ads outpaced internal promotion for Facebook, Instagram and WhatsApp combined. By September 22, Sensor Tower ranked Muse among the top 10 brands in the U.S. by advertising spend, with additional buys running on Reddit, TikTok and YouTube.
Connor Hayes, Meta’s head of Threads, told Business Insider that Muse’s marketing differs from the Threads playbook in a specific way: Threads promotions surface user-generated content matched to a person’s interests, while Muse promotions are built to demonstrate a concrete task the assistant can do for that specific user, based on what Meta already knows about them.
That distinction is the real story here. Every AI lab building a consumer chat app, from OpenAI to Anthropic to xAI, has to buy or earn its way to a first download. Meta does not. It can route Muse in front of roughly 3 billion existing Facebook and Instagram accounts and let its ad targeting engine, the same one that sells shoes and mattresses, decide who sees which capability demo. No other AI company has that distribution asset, and it is the reason Muse’s growth curve looks less like a startup’s viral hit and more like a feature rollout to an existing user base.
For competing consumer AI apps, the near term question is not whether their models are as capable as Muse’s. It is whether they can afford customer acquisition costs anywhere close to what Meta is paying, which on these numbers is close to zero.
Reported by TechCrunch (Sarah Perez) on 25 September 2026, citing data from Sensor Tower, Apptopia and Appfigures.
