Meta says Muse has 10 million users. The shopping pitch comes next.
The company is exploring commerce fees and premium AI, while its audience, security and profitability claims need careful separation.
Meta says its Muse AI assistant has passed 10 million users, with shopping emerging as one of the activities it most wants to turn into a business. The milestone is striking, but the more revealing story is how much of the company’s pitch still concerns what comes next.
In comments covered by TechCrunch on October 21, Meta described an assistant that can help with purchases as well as email, calendars and other tasks. It is also exploring different ways to charge for that activity, rather than relying only on the advertising model associated with its largest social platforms.
The figures come from the company and related reporting, not an independent audit conducted for this report. A total user figure is not the same thing as daily active users, and neither is the same thing as the number of people making purchases through an AI agent.
Ten million does not mean ten million daily shoppers
Mark Zuckerberg said that roughly a quarter of people who had set up Muse were now daily users. That is a different denominator from a claim that every one of the 10 million users returns each day.
TechCrunch also describes earlier Sensor Tower estimates of download growth. Downloads can help indicate demand, but downloading an app, creating an account and continuing to use a service measure different things. Combining those figures into one larger-sounding activity claim would overstate the available evidence.
Shopping is important because it offers a potential route from an assistant’s recommendation to a commercial transaction. The report describes a Plaid connection supporting agentic shopping and company comments about purchase sizes among an engaged shopping subset. Those observations do not establish the spending behaviour of the entire Muse audience.
The proposed business model has several parts
Meta expects Muse to remain free for most people while exploring commerce-related income and premium options for heavy users. More advanced models or additional usage could become part of a paid offering.
That is a set of monetisation plans, not proof that the product has already reached the profitability or revenue per user executives hope to achieve. The same distinction applies to suggestions that Muse could eventually earn more per user than Meta’s family of apps.
The company has also discussed serving more than a billion people over the next few years and sharply improving efficiency. Those are ambitions about future capacity and economics. They should not be reported as current audience totals or savings already realised.
An assistant that acts needs more than a good answer
An AI system that summarises a message has a different risk profile from one connected to a calendar, account or purchasing process. Once an assistant can take action, the practical questions include what it is allowed to access, what requires approval and how a user can stop or reverse a task.
Meta discusses a SecureVM approach and a forthcoming ConfidentialVM system that it says would prevent even the company from seeing certain content. The forthcoming system is a company promise about a future capability, not an independently established guarantee covering every interaction today.
The report does not test those protections or establish that errors and unauthorised purchases are impossible. Readers considering an action-taking assistant should distinguish a demonstrated feature from a proposed safeguard, and check the actual permissions offered in the version they use.
The next contest is about usefulness after the download
Meta’s plans reach beyond a standalone app, including a role for smart glasses and ideas for networks of agents interacting with people and businesses. Not every idea discussed on an earnings call is already a public feature.
The 10 million milestone shows why competitors and investors are watching Muse. Whether it becomes a durable business will depend on repeat use, reliable actions and a clear exchange of value. A large download chart can open that conversation; it cannot answer it on its own.
