Meta Muse Hits 3 Million Weekly Users, and Websites Cannot Block It by Name

By Tom Whitfield

More than 3 million people prompt Meta's personal AI agent Muse at least once a week, according to internal Meta data reviewed by The Information in early October. That number, already substantial before any public usage reporting exists, frames the problem now facing search teams: Muse and OpenAI's Dots browse websites in cloud browsers that identify themselves as the customer, and no robots.txt rule can stop them, because neither company publishes a user agent, request header, or signature for these browsers.

Muse is available free for most uses to anyone 18 or over in the U.S. and Canada. Meta says it launched the agent "with dozens of partners along with access to the entire Shopify catalogue," and is adding Walmart, Best Buy, Sephora, and Wayfair. OpenAI, which has published far less about Dots, says its plugin ecosystem lets agents "readily connect to over 4,000 apps." The parallel with mobile is hard to miss: StatCounter records computers falling from 98% of worldwide web traffic in January 2010 to under half in 2026. Agent-driven browsers may do to human browsing what the iPhone did to desktop.

What does this change for SEO teams?

The split is between two tiers. Connector-tier businesses — those with API integrations, including every Shopify store under the platform's default "Allow Shopify to manage for me" setting — hand their catalog over directly. Everyone else sits in the browser tier: Muse runs "a real up-to-date Chromium-based browser" on "a dedicated secure computer," in Meta's words, and works with whatever the page shows it.

For the browser tier, the familiar ranking work still applies, because agents find websites by name or by searching — though neither Meta nor OpenAI says which search engine its agent uses. What changes is what happens after arrival. Meta's documentation states 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 page's own scripts still run; the agent adds none.

That has a hard consequence: structured data in JSON-LD is invisible to Muse, because Chrome excludes script blocks from the accessibility tree. The author of the underlying analysis tested this on a real product page from a large Shopify store in October 2026. The structured data listed six of seven sizes as out of stock; the accessibility tree showed seven identical buttons, "Select size 8" to "Select size 14," with nothing marked unavailable. An agent learns what a shopper sees at a glance only by clicking each size, one by one.

Most of what the agent does read is standard on-page SEO: headings, link text, alt text, and facts in visible HTML. The transaction layer — labeled buttons, named form fields, checkout — belongs to front-end teams. Headings appear in the tree with their level; image alt text becomes the image's name; a price rendered in an image or canvas has no text to read. Collapsed sections depend on how they are hidden: content collapsed by height stays readable, while anything under display: none is out of reach.

Why robots.txt cannot reach these agents

Robots.txt works on names, and this visitor gives none. Meta's page for website owners lists five crawlers; Muse is not among them. OpenAI lists four; Dots are absent. Meta's description is blunt: "When Muse browses the internet, it will appear as your activity, so if you ask Muse to buy a shirt from a clothing designer's website, that designer might use your visit to show you an ad on Instagram." Amazon's robots.txt names 99 agents, and Muse and Dots appear on none of them.

When Amazon blocked Muse in September, it acted through its Conditions of Use, not robots.txt. GeekWire reported Amazon's three stated 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 users a workaround — "If a website blocks the cloud browser, ask your dot to try your connected computer" — meaning the decision effectively moves from robots.txt to bot-protection and firewall vendors. HUMAN Security has told PPC Land its bot-detection product can identify Muse.

Where agent activity shows up in reporting

Analytics is the quiet casualty. When Muse browses and buys on a site, GA4 counts it as a visit from the customer — there is no user agent, referrer, or IP range to filter by. When checkout happens inside an AI app through a connector, the site's pages never load, so analytics and pixels record nothing; the order appears only in the platform's order records. Shopify confirms this for stores on Meta and Google. GA4 added an AI Assistant channel on May 13 that sorts visits from ChatGPT, Gemini, and Claude by referrer, but it catches only referred link clicks. 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."

Several questions remain open, and the evidence cannot answer them yet: how many purchases these agents complete, how agents rank or choose between businesses in an answer, and how many website visits specifically come from agents. OpenAI has published no usage figures for Dots.

The leadership decisions

The analysis frames four decisions. Team: put SEO and front-end together on forms, bookings, and checkout, and give SEO a seat in bot-protection reviews. Risk: decide deliberately whether agents get in, rather than let an unreviewed scraper rule decide. Budget: WebAIM found 95.9% of the top million homepages had detected WCAG failures in 2026, so accessibility fixes on transaction pages were overdue before any agent arrived. Opportunity: when an agent compares ten stores, it can only buy from the ones whose pages it can interact with.

Markdown page versions and llms.txt files miss this visitor entirely — they strip out the buttons, forms, and checkout the agent came to use. Monitor whether Meta or OpenAI publishes agent identifiers, how GA4 and bot filters label these sessions, and whether Shopify's connector defaults expand to more platforms.