Meta Q3 Earnings Call: Over 10 Million People Set Up Muse Within Six Weeks, Coming to Meta Glasses in November

Download the earnings call here.

Download the follow-up call here.

Meta is positioning Muse as a shopping assistant, a tool for businesses and an interface for its smart glasses. On its October 21, 2026 earnings call, CEO Mark Zuckerberg said more than 10 million people had set up Muse in the US and Canada. The company also outlined AI changes to recommendations and advertising, security features and ongoing youth-related litigation.

How Meta will scale Muse: More than 10 million people in the US and Canada had set up Muse in its first six weeks, Zuckerberg said, with about 25% of users accessing it daily. Shopping was its most common use case, with “more than a quarter of people who are using Muse” engaging in it and an average purchase size “over $200”.

“We want to help every person have an extremely capable personal agent that understands your goals and can help you achieve them and improve your life in whatever ways you want,” Zuckerberg said.

Meta also plans to connect Muse to users’ content and advertising experiences. Zuckerberg said users would be able to connect it to “help improve the personalization and quality of their content and ads across Meta”.

During a follow-up discussion, CFO Susan Li said Meta needed to expand its user base and business partnerships before monetising transactions. “Today, we are not monetizing Muse outside of subscriptions,” Li said in the follow-up call. Zuckerberg said Meta expected to improve Muse’s compute efficiency “by more than 10x in the next year”.

Meta brings its own models into recommendations: Meta said it was transitioning its content-understanding systems to models from its Muse family. In offline evaluations, these models outperformed previous production models and reduced per-unit content-processing costs by more than 80% across Instagram, Threads, Facebook and advertising.

On Instagram Reels, recommendation changes increased time spent by more than 4% in the third quarter. CFO Susan Li said the company had upgraded its architecture so that it “ranks sequences of Reels at a time instead of evaluating each Reel independently”, while also adding richer watch-time signals.

Meta also completed the global rollout of a generative retrieval architecture across Facebook Feed and Reels. Inspired by large language models, the system labels content with semantic representations and generates potential recommendations. Meta said the system helps surface relevant new posts before they accumulate significant engagement.

The architecture now covers roughly one-quarter of unconnected Feed views and half of Reels views, increasing Reels watch time by 1%. Meanwhile, “nearly 45% of recommended Instagram Reels are now less than one day old”, Li said, up more than 11 percentage points from the previous quarter.

Meta also uses large language models to understand public Instagram posts. Li said those signals had “improved the engagement with Instagram search results by 1% through the end of this quarter”.

AI models are changing advertising systems: Meta is applying generative AI to ad retrieval and ranking. Its Meta Generative Recommender uses large language models to reason across ad content and user preferences. In the third quarter, the company launched generative retrieval on Facebook Feed, driving “a nearly 2% increase in conversions across targeted segments”, Li said.

Meta also expanded its GEM ad-ranking model to additional surfaces. The changes drove a 1.7% increase in conversions for offsite conversion campaigns and a 9.6% increase in app ad conversions. Improvements to Andromeda, its ad-retrieval system, increased conversions by 1.4% on Facebook.

In the follow-up call, Li said Meta’s LLM-inspired modelling techniques helped the company “select better ad candidates using a generative approach rather than scoring every ad individually”. She said the approach helped it predict “not just whether someone will convert, but what the conversion is worth”.

Meta is also using AI to help advertisers manage campaigns and produce creative content. Nearly 10 million advertisers used at least one of its generative AI creative tools in the third quarter. Adoption of its video-generation features grew 25% quarter-on-quarter, partly because tools can generate videos in “just a few seconds from a single image upload”, Li said.

The AI Business Assistant is gaining “creative insights, stronger business context and more agentic actions”. These features will allow marketers to plan, create and optimise campaigns, schedule tasks and alerts, and retain control through approvals and previews.

AI tools expanded for businesses and internal work: Meta is extending Muse to small businesses. In the follow-up call, Li said the company launched Muse for small businesses in late September, with connectors for Instagram accounts, Facebook Pages and advertising accounts, as well as third-party services such as Shopify, QuickBooks and Canva.

The tool can analyse advertising performance, draft campaigns and customer responses, organise direct messages and help businesses update their websites. In the follow-up call, Li said Meta believed Muse could make it easier for businesses to “manage their business, manage their marketing campaigns, engage with customers” and identify ways to improve ad performance and customer responses.

Meta is also developing an enterprise platform offering an application programming interface (API), Muse Code and other services. Zuckerberg said the company was training larger models beyond Muse Wonder, while developing smaller models that would deliver “Muse Wonder-level intelligence more efficiently”.

Meta also reported internal productivity gains. In the follow-up call, finance vice-president Chad Heaton said: “total software output per engineer has more than doubled year-on-year”. Employees produced 10% to 20% more analytics output than a year earlier, while “more than 70% of measured analytics output is now AI-assisted”, compared with effectively none at the beginning of 2026.

Muse is coming to Meta glasses: Zuckerberg said Muse would come to Meta Glasses in November 2026. Users already value features such as “music, calling, capturing moments, hearing assistance”, he said, but Meta expects Muse to become “the most important part of the experience”. The company also plans to ship Muse Charm, a hardware device, less than a year after its conception.

In the follow-up call, Li said security and privacy were critical because an agent can act on someone’s behalf: “Muse books and buys and sends and manages accounts on someone’s behalf. So, mistakes have real-world consequences, and trust is one of the most important factors for adoption.”

She said each user’s Muse runs on “its own dedicated isolated computer in the cloud”, with its memory, recurring tasks and artefacts stored in that person’s virtual machine. A separate system called Sentinel checks actions, connected apps and anything Muse sends to the internet. “Muse proposes actions, but only Sentinel can grant permission to perform them,” Li said in the follow-up call.

Li added that Muse “has no visibility into your passwords, logins or payment methods”. It asks users before sensitive actions, such as sending an email or making a purchase, and keeps an audit trail of actions taken and planned. Zuckerberg said the forthcoming Muse Confidential VM would ensure that ‘even Meta cannot see your content’.

Youth litigation and legal costs: CFO Susan Li said Meta continued to monitor “active legal and regulatory matters that could significantly impact our business and financial results”. She added: “We continue to see scrutiny on youth-related issues in several markets and have a number of youth-related trials scheduled in the coming months in the United States.”

Meta’s third-quarter expenses rose 68% year-on-year to $51.6 billion, including a $9.9 billion charge related to its August 2026 multistate attorneys general agreement. Li said expense growth was “primarily driven by legal-related costs, infrastructure costs, and third-party AI token costs”.

Meta Q3 Earnings Call: Over 10 Million People Set Up Muse Within Six Weeks, Coming to Meta Glasses in November

Why read this

Topics

Tags

Body