Meta's Muse agent had been installed more than 5mn times across the United States and Canada within 22 days of its September 8 release, according to estimates Sensor Tower published on September 30, 2026. The market intelligence firm also found that Meta had directed up to half of its daily house advertising impressions toward the app, which ranked among the 15 largest brands by advertising impressions in the US in the week to September 27.

In Short

Meta's new AI helper app, Muse, was installed more than 5mn times in the US and Canada in about three weeks, and an app-tracking firm expected it to pass that mark in the US alone on September 30, more than twice as fast as ChatGPT did. Meta helped it along with its own advertising, giving Muse up to half of the promotional space it normally uses for apps such as Facebook and WhatsApp. So the numbers show how much reach Meta can point at a new product, and they are an outside firm's estimates rather than figures from Meta.

Two counts, one milestone

The headline figure rests on two different measurements, and the report uses both. Across its two markets, Muse had already passed 5mn downloads when the report was written, according to Sensor Tower. The US-only number had not yet crossed the line. Sensor Tower projected it would do so on Wednesday, September 30 - the day of publication - which makes the American milestone an estimate of what was about to happen rather than an observation of what had.

The firm's chart, titled for a crossing of 5mn US downloads in 22 days, ends at 5.1mn on day 22. The report does not say whether the release date counts as day one. If it does, day 22 falls on September 29, a day before the date given in the text; counted from September 9, the two agree. The gap is small, but every comparison in the report is measured in days.

Then there is geography. Meta's announcement on September 8 described a US rollout on iOS, Android and muse.ai, as PPC Land reported when Muse first appeared, with no date for other markets. Canada followed ten days later. Meta posted on X on September 18 that the agent had reached the country, according to iPhone in Canada, which said iPhone and web access opened that afternoon while the Android version was still pending. The combined figure therefore blends 22 days of US availability with roughly 12 in Canada.

Earlier counts trace the shape of the curve. Digiday put US downloads at 730,000 in Muse's first five days, a figure recorded in PPC Land's weekly edition for the week to September 27. Reuters, citing Sensor Tower estimates, reported about 2.8mn downloads in the first two weeks, according to iPhone in Canada. The three figures may not share a geographic base. Read together, though, they describe an adoption rate that accelerated in the second and third weeks rather than tapering after a burst of curiosity.

Measured against ChatGPT, Grok and Claude

Sensor Tower chose the three most prominent standalone AI assistants as its yardstick. ChatGPT needed 56 days to reach 5mn US downloads, Grok 103 days and Claude 492 days, according to the firm. Muse's projected 22 days is less than half of ChatGPT's time. At the 22-day mark, the chart places ChatGPT at 2.3mn US downloads, Grok at 0.4mn and Claude at 0.1mn, against 5.1mn for Muse - about 2.2 times ChatGPT's early pace.

The conditions were not identical. A footnote in the report states that Grok and Claude were available only on iOS for their first five days, which depresses their early totals against an app released on both major stores and the web at once. The footnote is silent on ChatGPT, whose iPhone app reached US users in May 2023, roughly two months before an Android version followed in July. On that history, the 56 days in Sensor Tower's comparison were accumulated largely or entirely on one platform.

The broader set is small. Only 23 apps have reached more than 5mn US downloads within 22 days of release, according to Sensor Tower estimates, and about 85% of them are mobile games, with Fortnite, Pokemon GO and Candy Crush among them. Three were not games: Disney+, HBO Max and Threads, which drew about 20mn, 18mn and 14mn downloads respectively in that window. Twenty games out of 23 apps is 87%, consistent with the rounded share. Whether Muse itself is counted among the 23 is not stated; the report names only those three as non-gaming entrants.

That list cuts both ways. Muse's 5.1mn is about a quarter of the Disney+ total over the same span and a little over a third of Threads'. Threads, which Meta opened as a separate text-based app in July 2023 with sign-up through existing Instagram accounts, drew nearly three times as many US downloads in its first 22 days as Muse has. Meta has produced faster consumer adoption before, then, and from the same distribution base.

Twelve days at the top

Rankings moved faster than cumulative counts. Muse entered the US Apple App Store as the number two productivity app on release, according to Sensor Tower. By September 30 it was the top app overall on both the US App Store and Google Play, a streak of 12 days at number one on the App Store. In Canada it ranked first across both stores, with 11 consecutive days at the top of the Canadian App Store.

Counted back from publication, the US streak would have begun around September 18 or 19, two to three days after the advertising ramp that Sensor Tower dates to September 16. The Canadian streak, on the same arithmetic, began at or within a day of the country's September 18 opening. Muse was ranked first in Canada on Google Play as well, although Android access there had not been confirmed on opening day; the report does not say when it arrived.

Chart positions measure momentum over a short window. They carry no information about how many people keep using an app after installing it. Claude, for comparison, reached number one on the US App Store on March 1, 2026 during Anthropic's dispute with the Pentagon, a surge driven by news rather than by advertising.

The advertising push

September 16

According to Sensor Tower estimates, Meta began to ramp advertising spend for Muse on Wednesday, September 16, 2026, eight days after release, as downloads rose 73% day over day. The chart marks the moment between days 9 and 10. Before it, Muse's curve ran roughly level with ChatGPT's; after it, the two separate and the gap widens through day 22. The report's own summary is careful on cause, saying only that Muse's success "may also be boosted by Meta's aggressive advertising push".

Fifteen brands

Muse ranked within the top 15 largest brands by advertising impressions in the US over the week of September 21 to 27, per Sensor Tower data, a position the firm describes as "highlighting the aggressiveness of Meta's latest product launch". The report does not name the other 14 brands. Nor does it say which channels the impression ranking covers, whether those impressions were bought on outside media or served on Meta's own apps, or what the campaign cost.

House ads

The most unusual finding concerns house ads, the promotions a company runs for its own products. Muse received the largest share of Meta's house ads, accounting for up to 50% of daily house impressions between September 14 and September 27, according to Sensor Tower data (the report prints the end date as "927"). Over the same period, Facebook and WhatsApp ceded share of those impressions.

"Up to" marks a peak rather than an average across the fortnight. Even so, the reallocation is plain. Meta took promotional capacity that ordinarily supports its oldest apps and pointed it at a product three weeks old. Facebook, one of the two apps that gave up share, was among the Meta apps that Sensor Tower estimated had lost an average of 5% of downloads year over year in the second quarter.

The report does not reconcile its two advertising measures. The house ad window opens on September 14, two days before the date Sensor Tower gives for the spend ramp. Whether the spend estimate covers house impressions on Meta's own surfaces, purchases on outside media, or both, is left open.

What the report leaves out

Sensor Tower's disclaimer frames all of the above. Its mobile app and advertising estimates come from a proprietary panel and data science models, may not accurately represent metrics for any app or advertiser, and are meant for reading the direction of a trend, according to the firm. None of the figures in the report comes from Meta.

Several measures that would show whether the installs translate into use are absent. There are no daily or monthly active user counts, no retention curves, no session data and no estimate of how many users pay for a subscription. Is a download driven by a feed promotion worth the same as one driven by word of mouth? The report does not attempt to separate the two, and none of its comparisons with ChatGPT, Grok or Claude adjusts for promotion.

Sensor Tower's covering email described adoption as strong. The underlying data supports a narrower statement: Muse was installed quickly, ranked first quickly, and did so during a period in which Meta concentrated its own promotional inventory on it.

The product behind the numbers

Muse is Meta's consumer entry into agentic AI, software given a goal rather than a single instruction. Each person's agent runs on a dedicated cloud machine Meta calls Muse Secure VM, with its own browser, file system and terminal, and a separate program called Sentinel must approve anything it sends to the internet, according to the architecture PPC Land set out on September 9. Purchases go through Stripe's Link with a card generated for each transaction. People reach the agent through a dedicated app, muse.ai or a WhatsApp chat. Meta built it on Muse Spark, the model line that Meta Superintelligence Labs first presented on April 8, 2026.

Pricing is light so far. Meta described Muse as free for most uses, with subscription plans for heavier ones; Digiday's analysis on September 25 put those tiers at $20 and $100 a month, with merchant fees as the only other revenue, as PPC Land's weekly edition recorded. The same edition noted that Amazon had blocked Muse from its store after Meta declined to exclude Amazon from the agent's reach.

Two further steps came in the final days of September. On September 28, Meta set up a division to sell its AI stack to companies and named the Muse agent among its first four products, with no prices or dates. A day later it extended Muse to small businesses in the US and Canada, with connectors to Instagram professional account analytics, Facebook Pages and Meta ad accounts, and skills for tasks including reviewing performance and managing campaigns, according to iPhone in Canada.

Why this matters for the marketing community

The first point concerns Meta as an advertiser. A product that did not exist on September 7 was, within three weeks, one of the 15 brands with the most US advertising impressions, and it absorbed up to half of the house inventory Meta uses to promote its own apps. Meta's advertising business was already selling more slots at higher prices before Muse arrived: in the second quarter, ad impressions rose 14% and average price per ad 12%. If a large part of the Muse campaign ran on Meta's own surfaces, those impressions occupied placements that the auction could otherwise have sold. The report does not say where they ran, and Meta has not published how house promotions are weighed against paid demand.

The second concerns money. Meta reported second-quarter advertising revenue of $59.36bn, up 27%, against capital expenditure of $31.08bn and free cash flow of $784m, and narrowed full-year capital expenditure guidance to $130bn to $145bn. Advertising supplied more than 97% of second-quarter revenue. Digiday's September 25 analysis argued that advertising inside Muse looks likely, pointing to the thin subscription revenue and to AI spending that had absorbed 98% of Meta's operating cash flow by the second quarter. Against that sits Meta's own commitment, made on September 8, that Muse conversations and the data on each person's virtual machine stay out of its ad systems. Meta AI works under a different rule. Since December 16, 2025, conversations with it have fed Meta's content recommendations and its ad personalization across more than 1bn monthly users. The faster Muse's user base grows, the more valuable the data Meta has promised to keep apart becomes.

The third concerns where advertising budgets go next. Sensor Tower is also the firm whose estimates put ChatGPT's advertising business above a $1bn annualised run rate, with financial services rising from 2% to 13% of its US ad spend between April and August, and found ads shown per user per hour on the ChatGPT mobile app up 163% over the same months. At the same time, ChatGPT's share of US AI prompt volume fell from 70% in January to 50% in June in Comscore panel data. Assistant audiences are dividing among several companies, and Meta's entry adds one whose owner already runs Advantage+, the automated campaign system that passed an annual revenue run rate above $75bn in the second quarter. An agent that buys on a user's behalf, inside an app promoted by the same company that sells the ads, places Meta on both sides of a purchase decision.

The fourth concerns evidence. Every figure here is a third-party estimate built from a panel and models, and Sensor Tower says as much. PPC Land's weekly edition for the week to September 27 argued that the most reliable numbers in advertising increasingly arrive through litigation rather than through platform disclosure. Meta has published no install, usage or subscriber figures for Muse. Until it does, the record of the agent's first three weeks rests on outside estimates whose authors describe them as directional.

Timeline

Summary

Who: Sensor Tower, a market intelligence firm, published estimates on Meta Platforms' Muse agent, built by Meta Superintelligence Labs. The comparisons cover OpenAI's ChatGPT, xAI's Grok, Anthropic's Claude, and Disney+, HBO Max and Meta's own Threads.

What: Muse passed 5mn downloads across the US and Canada within 22 days of release, with a US-only crossing of 5mn projected for September 30, against 56 days for ChatGPT, 103 for Grok and 492 for Claude. It held the top position on the US App Store for 12 days and on the Canadian App Store for 11. Meta ramped Muse advertising on September 16, gave the app up to 50% of its daily house ad impressions between September 14 and 27, and placed it among the top 15 US brands by advertising impressions in the week to September 27.

When: Sensor Tower published the report on September 30, 2026, covering the period since Muse's US release on September 8, 2026; Canadian availability began on September 18.

Where: The United States and Canada, the only two markets where Muse is available, measured on the Apple App Store and Google Play.

Why: The figures show Meta concentrating its own promotional inventory on a new AI agent while its advertising business supplies nearly all of its revenue and its AI spending absorbs most of its cash flow. For advertisers, the questions are where those impressions ran, whether Muse's separation from Meta's ad systems holds as the user base grows, and what the installs amount to once the promotion eases - none of which the panel-based estimates can yet answer.