Muse Hit 10M Users. Did the Avatar Actually Matter?
Meta’s Muse passed 10 million users with shopping as a breakout use case. Muse Realtime Avatar adds presence, but public numbers still point to the agent, not the face, as the growth engine.
When Meta launched Muse, the pitch was bigger than another chatbot. Muse is a personal AI agent: shopping, inbox triage, reservations, multi-step chores. Alongside that agent story, Meta introduced Muse Realtime Avatar, a customizable digital face, body, and voice meant to give the assistant presence in a live conversation.
So did the avatar actually matter? The honest short answer: probably yes for engagement and memorability, but it is not what the public numbers prove. Muse’s reported momentum, 10 million-plus users and strong shopping behavior, is attributed to agent work getting done. The avatar looks more like packaging and social fuel than the engine.
What the Numbers Actually Say
Users
10M+
Top task
Shopping
Avg purchase*
>$200
Avatar split
Not public
On Meta’s Q3 2026 earnings call, Mark Zuckerberg said Muse had more than 10 million users after its September launch in the US and Canada. Shopping stood out. Coverage of the call reported that a large share of engaged users were shopping through the agent, with average purchase sizes over $200 for the shopping cohort Meta described. Productivity hooks, email and calendar connections, were the other headline behaviors.
That is a strong product-market-fit signal for an agent that can spend money and touch private accounts. It is not an avatar A/B test. Meta has not published a breakdown showing whether those shoppers were in a realtime avatar session, on voice only, on text, or letting Muse run chores in the background. You cannot defend a claim like “the avatar lifted retention X%” from public data.
What you can defend: people appear to use Muse to outsource annoying work, shopping, scheduling, refunds, inbox cleanup. Presence may help. Proof of avatar-driven growth is still missing.
Key features
- 10M+ users reported by Meta on the Q3 2026 call
- Shopping highlighted with >$200 average purchase size in Meta’s framing
- No public avatar vs text/voice attribution for those metrics
- Trust signals (connectors, secure VM story) matter as much as the face
Where Avatars Help Muse
Avatars are not visual fluff when they work. A face with realtime expression turns a chore list into a conversation. Meta’s research post on Muse Realtime Avatar describes an audio-driven system that keeps voice, lips, and mannerisms in sync from the same speech-token stream as Muse Realtime Voice, with roughly subsecond response latency in Meta’s published measurements.
For Muse specifically, the avatar case rests on three ideas. First, presence: multi-step tasks feel less like issuing commands to a void. Second, personalization: users who customize a face talk about the assistant as “theirs.” Third, distribution: Meta’s social graph is a natural home for showing off a character, and reports have described avatar-sharing UI in progress on Muse web, even when it is not fully live for everyone.
That third point is strategic for Meta even if it is early. An avatar you can display or share is growth packaging. An agent that buys the right product is retention.
Key features
- Presence turns chores into conversation
- Customization builds emotional ownership
- Shareable faces can become social distribution, when shipping catches up
- Meta’s own research frames avatar as embodiment on top of realtime voice
Where Avatars Fall Short
People trust Muse with shopping carts and inboxes because connectors, payments, and task completion work, not because the mouth shapes look right. Reliability, safety, context, and time saved decide whether someone hands over Plaid or their calendar.
An avatar cannot rescue a bot that books the wrong restaurant or stalls mid-refund. A too-real face can even raise the bar: if it looks and talks like a person, users expect it to think and act like one. Uncanny competence gaps feel worse when the UI implies a human stand-in.
There is also a product-availability caveat. Muse Realtime Avatar was introduced as a realtime embodiment technology with a rolling / early-access path in public coverage. Treating “Muse users” and “avatar users” as the same population overstates what the face has already done in the wild.
Key features
- Agent reliability beats facial realism for high-stakes tasks
- Hyper-real faces raise expectations the model must meet
- Avatar rollout maturity is not the same as Muse’s overall user count
- No public evidence that avatars drove the shopping numbers
Avatar as Packaging, Agent as Engine
Task execution
Core adoption
Personalization
Retention
Avatar
Engagement
Commerce
Revenue
The cleanest way to read Muse is as stacked layers. Task execution handles shopping, bookings, and problem-solving: the core reason people open the app. Personalization, memory and context, keeps them returning. The avatar adds personality and presence: engagement and memorability. Commerce turns successful commands into purchases and, in Meta’s stated model, platform take-rate.
On that map, the avatar is packaging. It brings people in, gives a reason to open a conversation, and makes Muse feel less like a form. People stay if the underlying agent finishes the job. That is the same split conversational twin builders should internalize: face for presence, agent for value.
Key features
- Task execution = why people use Muse
- Personalization = why they return
- Avatar = why it feels alive and shareable
- Commerce = how Meta wants to monetize success
What’s Missing: Real Measurement
To know how much avatars help, Meta would need controlled comparisons: text-only vs voice-only vs static avatar vs full animated realtime avatar. The metrics that matter are day-7 and day-30 retention, tasks completed, chores outsourced, purchase rate, and revenue per user, sliced by modality.
Until those cuts are public, attributing Muse’s 10 million users to the avatar is guesswork. The defensible claim is narrower: avatars plausibly boost presence and engagement; Muse’s reported momentum aligns more tightly with agent capabilities, especially shopping and account-connected productivity.
Key features
- Need modality A/B tests, not vibe checks
- Watch retention, task completion, and revenue per user by interface
- Public narrative today favors agent utility over avatar causation
What Twin Builders Should Take From Muse
For Ai Twin readersMuse is a reminder that a customizable face can differentiate an assistant in a crowded market, especially for a company that lives on social and visual surfaces. It is also a reminder that the chart Meta bragged about was shopping and trust in connectors, not lip-sync preference scores.
If you are building a conversational twin, copy the useful half of the lesson. Use the avatar as packaging that makes the agent present and memorable. Put roadmap weight on the engine: accurate answers, tools, handoffs, and tasks that finish. That is the Ai Twin thesis in product form, a browser-rendered conversational twin that talks, plus an agent that works. Presence without competence is a demo. Competence without presence is a chatbot. You want both, in that priority order.
For the category framing, see what an AI twin is. For why the face layer should stay economical while the agent scales, see browser-rendered AI avatars.
Key features
- Avatar gets attention; agent earns retention
- Measure modality impact instead of assuming face = growth
- aitwin.me: conversational twin + your existing agent stack
- Ship presence and disclosure together if you operate in regulated markets
My take
Want Meta’s avatar tech write-up?
Bringing Your Muse to Life (Meta AI Research)
Want the earnings numbers?
TechCrunch / earnings coverage of Muse’s 10M users
Want the twin definition?
What Is an AI Twin
Want to ship the combo?
Create a twin on aitwin.me
Avatars made Muse stand out. They make the product more memorable, and potentially more shareable as Meta finishes social distribution features. Did they make Muse dramatically more useful on the evidence we have? Not yet.
In a crowded AI assistant market, a customizable face is a sharp touch, especially for Meta. Muse’s true value still comes from whether it does real work: the right product, the right booking, the calendar that stays correct, purchases that complete without babysitting.
Simply put: the avatar gets your attention; the agent keeps you coming back. If Meta nails both, daily engagement from presence plus reliable agent results, that is a potent combination. Based on what is public today, avatars look like rocket fuel. The workhorse is still Muse’s ability to get stuff done. Builders should take the same split to production twins on aitwin.me.
Frequently Asked Questions
What is Meta Muse?
Muse is Meta’s personal AI agent app, positioned to handle multi-step tasks such as shopping, email, calendars, and reservations. Meta reported more than 10 million users after its September 2026 launch in early markets.
What is Muse Realtime Avatar?
Muse Realtime Avatar is Meta’s embodiment technology that turns Muse Realtime Voice into an expressive interactive avatar with synchronized lip motion and mannerisms. Meta detailed it in a September 2026 research post and described rolling or early-access availability in product coverage.
Did the Muse avatar cause the 10 million user milestone?
There is no public breakdown proving that. Meta’s reported highlights emphasize overall users and shopping/productivity usage. Avatar contribution to retention or spend remains unverified in public data.
How should AI twin products think about avatars after Muse?
Treat the avatar as packaging for presence and engagement, and the agent as the engine for retention and revenue. Measure both. That is the same model conversational twins on aitwin.me are built around.
