Meta Muse is a consumer AI agent that Meta launched on 8 September 2026. It can find products and, with the user's approval, complete purchases, and brands cannot yet buy placement inside it. As of 9 October 2026, the practical job for a brand is to make its product data and checkout agent-readable, not to chase a Muse "channel" that does not exist yet.
Key Takeaways
- Muse launched 8 September 2026 in the US (and Canada, per one report) as an agent you message in Meta's app or WhatsApp.
- The usage numbers conflict: 1.8M daily users and 6.6M downloads come from Sensor Tower via Dealroom, while other reports cite roughly 700,000 daily users and over 1 million daily prompters.
- Meta has not said how brands can pay for prominence in Muse; Mediaweek reports monetisation is unresolved.
- Brands can prepare now: accurate feeds, live stock, clean checkout, and a Muse-readable web presence.
- Treat attribution as unsolved. Do not move budget on Muse numbers yet.
What Muse actually is
Muse is not a chatbot tab bolted onto Instagram. Engadget describes it as an agent that can shop, send emails and plan trips on your behalf. If you want a plain definition of the category, I wrote one in what an AI agent means for marketing.
Per Mediaweek's 24 September 2026 piece, it is US-only for now, reachable through Meta's app or WhatsApp, with plans for AI glasses, an agent email address and Mac computer use. Pricing, according to Crypto Briefing, is a limited free tier plus Power at $20 a month and Maximum at $100 a month.
The numbers, and why they disagree
This is the part I would not repeat blindly in a client deck.
| Claim | Figure | Source (as of 9 October 2026) |
|---|---|---|
| Downloads | 6.6M since 8 Sept | Sensor Tower, via Dealroom |
| Daily active users | 1.8M | Same; daily-user definition not given |
| Downloads | 5M+ in about 22 days (US) | Sensor Tower, per TNW and Crypto Briefing |
| Weekly users | 3M+ prompt weekly | The Information internal data, via TNW; TNW says it has not verified it |
| Daily users | 1M+ send a prompt daily (TNW); about 700,000 DAU (Crypto Briefing) | Conflicting |
My reading: all of these are third-party estimates or leaks, not audited Meta disclosures. Crypto Briefing also notes open rates are "significantly below" ChatGPT and Instagram, so downloads are a weak proxy for habit. The ones above that quote 1.8M should be hedged as "Sensor Tower estimate".
What brands can do there today
Honestly, not much directly. Mediaweek's reporting is that Meta has not announced brand-facing tools for the consumer agent, and that brands cannot yet buy prominence or see how recommendations are labelled. What you can influence is whether Muse can recommend and buy from you.
According to Mediaweek, Meta has outlined access to Shopify's catalogue, Shop Pay and PayPal and some major US retailers. If you sell via Shopify, you are plausibly closer to the path than a brand with a custom checkout. Verify this against your own store before promising anything.
The small-business side
Separately, Reuters (via The Star, 29 September 2026) reported a Muse for Small Business expansion. It connects to Shopify, QuickBooks, Stripe, Canva, Instagram professional analytics, Facebook Pages and Meta ad accounts, and Meta says it will not publish content, send messages or make purchases without approval. That is an owner-side tool, not a consumer discovery channel, and the report does not say where it is available.
What brands cannot do (yet)
- Pay to be the recommended brand. Meta has not said how Muse will be monetised; Zuckerberg has floated a small transaction fee, per Mediaweek. Digiday (25 September 2026) says Muse carries no ads, that Meta says Muse conversations are kept separate from its ad systems, and quotes an agency voice warning brands may get "just one shot (or very few)" with an agent.
- See attribution. Marketers are asking who gets credit for a Muse-assisted sale, and there is no answer in the sources I opened.
- Rely on Amazon. TNW and Mediaweek both report Amazon has blocked Muse from shopping on its site, so a brand on both Amazon and its own store may not get an equal path.
- Assume safe handling. Reuters-reported tests of a "human concierge" feature, rolled back after Meta's executive called the undisclosed start "a miss", are a reminder that agent behaviour is still moving. I could not open the NYT report on safety objections, so I am not characterising it.
What I would do this week
- Audit product data. Price, stock, shipping, returns and variants should match between your site, feed and checkout. An agent that quotes a wrong price loses you the sale and the trust.
- Test your own checkout as a machine would. Guest checkout, no forced account creation, no pop-up that blocks a script. See my AI marketing stack audit for how I review tool dependencies.
- Make claims quotable. Short, factual product descriptions with specs outrank clever copy for an agent.
- Set a monitoring routine. Search your brand in Muse monthly, screenshot, log. That is the only "reporting" available.
- Do not shift paid budget. Wait until Meta publishes formats and measurement.
Risks I would flag to a client
Privacy is the largest. TNW reports setup asks for Gmail and calendar access, Meta uses Muse data to train its AI by default (with personal details removed, and an opt-out), and a security researcher found a critical flaw soon after launch that Meta says it fixed. If your brand's customers are in the audience, expect questions. I would also avoid any Muse-specific promise in contracts until terms exist, because the product is weeks old.
A product-data checklist an agent can use
Agents do not browse like people. They read structured information and act on it, so the following gaps matter more than your homepage banner:
- Price and currency consistent across page, feed and checkout.
- Availability updated often enough that an out-of-stock item is not recommended.
- Delivery and returns stated in one plain sentence on the product page.
- Product identifiers such as brand, model and size, so an agent can match a request to the right variant.
- Reviews that are real and recent. Agents, like answer engines, lean on third-party corroboration.
- Schema markup for Product and Offer, which also helps search engines. The same hygiene serves Google, ChatGPT and Muse.
None of this is Muse-specific, which is the point. I would not build anything for a channel with no published formats, but I would fix anything on this list that is already broken.
Who should care most
Ecommerce brands with Shopify stores in the US, WhatsApp-heavy businesses and anyone selling impulse or replenishment products are closest to the agent-shopping use case. Service businesses and B2B should watch but not act.
Where this fits in the bigger agent picture
Muse is one of several agents (OpenAI also launched an Agents API; see my model choice guide). The shared lesson: discovery is shifting from a results page a human scans to an answer an agent acts on. The brands that win have clean data and trusted reviews.
FAQ
What is Meta Muse?
Muse is Meta's personal AI agent, launched on 8 September 2026. It can search, plan and, with approval, buy things, and it works in Meta's app and WhatsApp.
How many people use Muse?
It depends on who you ask. Sensor Tower figures relayed by Dealroom say 6.6M downloads and 1.8M daily users; other reports cite about 700,000 daily users. None are audited Meta numbers.
Can brands advertise inside Muse?
Not as of 9 October 2026. Mediaweek reports Meta has not said how Muse will be monetised, and brands cannot yet buy prominence.
Can Muse buy from my store?
Reports say Muse can complete purchases, with access outlined for Shopify's catalogue, Shop Pay and PayPal. Test your own flow rather than assuming coverage.
Does Amazon allow Muse to shop?
No. TNW and Mediaweek both report Amazon has blocked Muse from shopping on its site.
Is Muse available in India?
I found reports of US availability and the app topping charts in the US and Canada. I did not find an India launch date, so check Meta's official page.
Is my customer data safe?
Muse requests broad access (Gmail, calendar) at setup and uses data for training by default with an opt-out, per TNW. Treat privacy as a live issue.
Should I change my social budget because of Muse?
No. There are no ad formats or attribution yet. Fix product data and checkout first.
How is Muse different from Meta AI?
I did not find a source that cleanly compares them, so I will not guess. Check Meta's help pages.
Work with me
I am Younus Fardeen, an India-based organic growth, SEO and AEO strategist with 4+ years of marketing experience across edtech and startups, including social-led growth at Masai School. If you want help getting your product data and content ready for AI agents and answer engines, see my work and reach me through the contact form at younusfardeen.in.