Meta Muse has passed the ten-million-user mark in roughly six weeks, and its most popular job is shopping. Meta said during its third-quarter earnings call on October 21, 2026 that the AI agent's adoption confirms earlier estimates from the market intelligence firm Sensor Tower. Roughly one in four of the people who set up Meta Muse now open it daily, chief executive Mark Zuckerberg said — and among engaged users who shop through the agent, more than a quarter average purchases above two hundred dollars, a figure that surprised even industry insiders.

The growth figures are striking even by app-store standards. The research firm reported that the app accumulated ten point nine million downloads worldwide within its first forty days, with ten point two million of those in the United States alone, according to Crypto Briefing's summary of the data (Crypto Briefing). That puts the app in rare company: only about a dozen other apps in history have reached that milestone so quickly. Daily active users hit two point five million on October 16, 2026, growing fifteen percent week over week. Retention is the number analysts are watching most closely: the app's seven-day retention rate sits at thirty percent, against nineteen percent for rival AI apps, while fourteen-day retention stands at twenty-three percent versus sixteen percent for the competition.

"Now that Muse has achieved product-market fit," Zuckerberg said, the company's focus shifts to scaling the agent, building revenue streams, and optimizing compute costs. Meta expects the app to eventually generate more revenue per user than its family of apps does today — currently several hundred dollars a year across Facebook, Instagram, WhatsApp, and Messenger, according to TechCrunch's reporting on the call (TechCrunch). Meta also pointed to its newer Muse Wonder model, available to paid users, as a driver of better quality, engagement, and retention, and said it is developing a more efficient version to reach more people.

Shopping first, productivity second

Shopping is the standout behavior. Meta's chief financial officer Susan Li said the three leading use categories are shopping, personal productivity, and creativity, with shopping accounting for over twenty-five percent of weekly active users. "Plaid told us that they hadn't seen activity like they're seeing with their connector on any other service before," Zuckerberg said, referring to the fintech company whose connector links the agent to users' bank accounts for purchases. Beyond shopping, people commonly connect the agent to email and calendars, which Zuckerberg framed as a sign of trust in the app's security setup, including its Muse Secure VM. A forthcoming Muse Confidential VM would go further, preventing even Meta from seeing users' content.

What it means for your wallet — and your privacy

For users, an agent that shops is a double-edged convenience. On one side, it can compare prices, hunt down discounts, and handle returns without the tab-switching that eats an evening. Meta Muse has already shown it can find real money: earlier coverage on GenZ NewZ documented users saving hundreds through subscription cancellations, bill negotiations, and refunds. On the other side, the average purchase size above two hundred dollars means people are trusting the agent with serious money, not just coffee orders.

That trust is exactly what the rest of the industry is fighting over. Big banks have spent the month drawing lines around outside AI agents moving money, and retailers are deciding how much access agents get to their checkouts. Meta is pouring fuel on the race: the company directed thirty-nine percent of its United States advertising spend to the app in October 2026, up sharply from the prior month. The agent boom is also attracting serious capital elsewhere in AI, from a billion-dollar bet on leaner AI models to a blockbuster IPO filing showing huge growth and huge losses.

The counterweight is the money Meta is spending to get here. The company reported third-quarter revenue of sixty-four point six billion dollars, up twenty-six percent year over year, while guiding capital expenditure of one hundred thirty to one hundred forty-five billion dollars for the year, mostly aimed at AI infrastructure. Meta's bet is that an agent people trust with their shopping will eventually be worth more per person than a feed people scroll. A user base that large suggests the bet is at least plausible — whether it pays depends on what the agent does next.