Meta's Muse AI agent brings back 73% of new users — nearly double ChatGPT's rate

Meta's personal artificial intelligence agent Muse retained 73.3% of new mobile app users on the day after installation, compared with 41.3% for OpenAI's ChatGPT, according to September 2026 data from market intelligence firm Sensor Tower. The figures, shared by Menlo Ventures partner Deedy Das, give Meta an early advantage in user retention, although ChatGPT overtakes Muse later in the 14-day period.

Muse Leads ChatGPT in Day-One Retention

The September data, covering mobile applications across 81 countries, show that Muse brought back nearly three-quarters of its new users on day 1. ChatGPT's corresponding retention rate was 41.3%, a difference of 32 percentage points.

The comparison suggests that Muse is attracting users who are willing to return shortly after trying the service. Its early performance is notable because retaining users beyond their first session is a key challenge for consumer AI products, many of which can attract initial interest without becoming part of users' daily routines.

The figures were highlighted in an analysis shared by Das, with Sensor Tower data providing the underlying estimates. They measure whether users return to an application after installing it, rather than how many people complete a task, pay for a subscription or use the service regularly over several months.

That distinction matters when assessing whether Muse is establishing a durable position in the AI assistant market.

ChatGPT Overtakes Muse Within 2 Weeks

Muse's early lead does not persist throughout the measurement period. The same data show that ChatGPT overtakes Muse between day 8 and day 9, as Muse's retention rate declines more sharply.

By day 14, ChatGPT retained 26.8% of its initial mobile user cohort, compared with 22.5% for Muse. The reversal indicates that strong initial engagement does not necessarily translate into stronger long-term retention.

For Meta, the challenge will be to turn first-time users into people who repeatedly rely on Muse for useful tasks. For OpenAI, the figures suggest that ChatGPT's established usage patterns may help it maintain engagement after the initial discovery phase.

The comparison also illustrates why a single retention metric can give an incomplete picture of competition. Muse leads on the first day, while ChatGPT performs better by the end of the two-week period.

Meta Designed Muse to Complete Tasks, Not Just Answer Questions

Meta introduced Muse on September 8, 2026, positioning it as a personal AI agent that can take action on users' behalf rather than simply respond to prompts.

According to Meta's launch announcement, the assistant can help with tasks such as sending emails, booking travel, managing projects and working across connected applications. It is accessible through a dedicated application and WhatsApp, with a persistent virtual computer environment designed to let the agent interact with websites and services.

The product builds on Meta's broader strategy of integrating artificial intelligence into the company's consumer ecosystem. Its existing platforms give it opportunities to introduce AI services to users already familiar with its applications, although access to those audiences does not guarantee sustained use of a separate product.

Meta has also emphasised personalisation, with Muse designed to learn from conversations and users' goals. The idea is to make the assistant more useful over time by remembering context and helping manage recurring activities.

Whether that approach creates lasting engagement will depend on the quality of its results, the reliability of its actions and the extent to which users trust it with access to their accounts and personal information.

Early Downloads Show Strong Interest, but Retention Remains the Test

Muse's initial adoption has been rapid. TechCrunch reported in September, citing Sensor Tower estimates, that the application had surpassed 3.4 million downloads during its early rollout. Estimates from other analytics firms differed, highlighting the uncertainty involved in measuring a newly launched application.

The Information separately reported on September 30 that Muse had more than 3 million users prompting the agent at least once a week, based on internal company data it reviewed. More than 1 million people were sending prompts daily, according to that report.

These figures point to meaningful early interest, but downloads, active users and retention measure different aspects of adoption. Downloads indicate how many installations have taken place, active-user figures measure engagement over a specified period, and cohort retention tracks how many people return after first trying a product.

A service can perform strongly across one measure without matching that performance across the others. Meta's ability to convert its early audience into consistent users will be more revealing than download totals alone.

The Comparison Has Important Limitations

The retention figures should not be interpreted as a definitive ranking of the two products. The data compare mobile-app cohorts, and the services have different histories, distribution strategies and user bases. Muse launched much more recently, while ChatGPT has had years to develop established use cases and repeat-user habits.

Muse's initial availability was also more geographically limited than ChatGPT's. TechCrunch reported that Muse was initially available in the United States and Canada, making launch conditions and user-acquisition patterns important considerations when interpreting comparisons.

The figures do not establish why users returned or stopped using either application. They also do not measure the accuracy of responses, successful task completion, customer satisfaction or the proportion of users willing to pay.

Muse's agent capabilities introduce additional considerations. Giving an AI system permission to browse websites, access connected applications or perform actions can make it more useful, but it also raises questions about privacy, security and user control. Reuters reported in September that Meta had been testing a human-assisted calling feature for Muse before pausing it amid concerns over transparency and privacy.

Those issues could influence long-term trust, particularly as personal AI agents take on tasks involving communications, purchases and sensitive account information.

What Meta Needs to Prove Next

The next test is whether Muse can improve retention beyond its initial launch period. That will require evidence that users continue returning after the novelty fades and that the assistant can reliably complete tasks that would otherwise require manual work.

Meta will also need to demonstrate that its agent can provide consistent value without creating unacceptable privacy or security risks. The more deeply an assistant integrates with users' digital lives, the more important those safeguards become.

For the wider AI industry, the results show that early adoption and long-term engagement are separate challenges. Muse's day-one performance gives Meta an encouraging starting point, but ChatGPT's stronger day-14 retention suggests that the competition will depend on sustained usefulness rather than the initial excitement surrounding a new AI product.