Meta Muse Hits 3 Million Weekly Users, and Your Analytics Can't See It

By Tom Whitfield

More than 3 million people prompt Meta's Muse at least once a week, according to internal Meta data reviewed by The Information in early October. Your analytics cannot tell any of them apart from ordinary customers.

Muse and OpenAI's Dots split the commercial web into two tiers: sites with a connector deal and sites the agent reaches through a cloud browser. Meta says Muse "launched with dozens of partners along with access to the entire Shopify catalogue," adding Walmart, Best Buy, Sephora and Wayfair. OpenAI says its dots "can readily connect to over 4,000 apps" through plugins. Everyone else gets the browser tier: Muse runs, in Meta's words, "a real up-to-date Chromium-based browser" on "a dedicated secure computer."

Muse is already available free to anyone 18 or over in the U.S. and Canada. The shift mirrors mobile: StatCounter puts desktop computers' share of worldwide web traffic at 98% in January 2010, falling to under half by 2026. Agent-driven browsers now threaten the human browser the same way.

Why robots.txt can't stop Muse or Dots

Neither Meta nor OpenAI publishes a user agent, request header or signature for these browsers. There is no name to put in robots.txt. Meta's page for website owners lists five crawlers; Muse is not among them. OpenAI lists four; dots are absent.

Meta describes the problem plainly: "When Muse browses the internet, it will appear as your activity." The agent walks in wearing your customer's identity, and the site that hosted the visit may retarget that customer on Instagram.

Amazon set the precedent in September. It blocked Muse through its Conditions of Use, not robots.txt. GeekWire reported Amazon's three reasons: Meta never told Amazon that Muse would access its store, the agent does not identify itself when browsing, and it appears to capture and store customer credentials.

OpenAI offers a workaround when sites push back. Its documentation tells users: "If a website blocks the cloud browser, ask your dot to try your connected computer." The decision on blocking moves from robots.txt to bot protection. HUMAN Security, as reported by PPC Land, says its bot-detection product can identify Muse. Ask your vendor whether it can tell a human-operated Chromium browser from an agent-operated one.

What do the agents actually read on a page?

Meta says Muse's browsing sub-agent "sees an accessibility tree snapshot of the page, not the raw DOM" and "has no ability to run JavaScript in the page context." The accessibility tree is what screen readers use. Most of it is on-page SEO — headings, link text, alt text, prices in HTML text.

Two findings from a live Shopify product page show the gaps. The page's structured data listed six of seven sizes in one color as out of stock. In the accessibility tree Chrome built, seven identical buttons appeared, "Select size eight" to "Select size 14," with nothing marked unavailable. An agent learns stock status only by clicking each size. On the same page, shipping eligibility and the free-shipping threshold appeared only after scripts ran, so raw-HTML crawlers missed them.

Structured data itself is invisible to Muse: Chrome leaves script blocks out of the accessibility tree. Keep it for Google; put the same facts in visible text. Buttons need labels, form fields need labels, and icon-only add-to-cart controls without accessible names are unnamed objects.

Where agent orders show up — and where they don't

• In-app checkout (Meta or Google direct): the order appears only in platform order records. Analytics and pixels never run. Shopify confirms this for its stores.
• Agent sent the person via a link: a referred visit, or a direct visit with no referrer.
• Agent browsed and bought on your site: a visit counted as your customer. No user agent, referrer or IP range exists to filter by.

Google Analytics 4 added an AI Assistant channel on May 13 that sorts visits from ChatGPT, Gemini and Claude by referrer. It catches only the second case, and only when the app sends one. Shopify says its bot classification is "designed to be conservative because it's better to miss some bots than incorrectly label real customers as bots." Agent sessions land in the human count.

What leaders must decide

The author of the analysis, writing for Search Engine Journal, frames this as four decisions:

• Team: SEO and front-end jointly own forms, bookings, signups and checkout. Give SEO a seat in bot-protection reviews.
• Risk: Decide deliberately whether agents get in. Amazon blocked Muse on purpose. An unreviewed default is deciding for you.
• Budget: WebAIM found 95.9% of the top million homepages had detected WCAG failures in 2026 — the fix was overdue before any agent arrived.
• ROI: Nobody can measure browser-tier agent revenue yet. The accessibility work pays today through screen-reader users.

Shopify stores should check their tier first: the default setting "Allow Shopify to manage for me," under Sales channels then Agentic, shares products "with Meta by default through Shopify Catalog."

The citations race still pays today. But OpenAI, Google and Meta — the companies whose AI answers cite your site — are each building agents that can also use it and buy from it. Stop at the citation, and the agent can quote your website but cannot buy from it.