Meta's Muse AI Shopping Agent Launch

Twelve days after Muse launched, Amazon shut it out of Amazon.com, saying Meta refused a request to remove the bot and alleging Muse accessed the site and captured login credentials without disclosing that it was an agent, not a person^[1]. The stakes go beyond politeness: Amazon controls roughly 40.5% of US retail e-commerce, and sponsored product listings are a major profit center for its ad business^[2]. An agent that picks a product algorithmically, skips the search results page, and checks out on a user's behalf can route purchases around exactly the paid placements that fund that business, which is the core reason Amazon cited for blocking Muse^[1]. Shopify took the opposite stance, opening its merchant catalog and Shop Pay checkout to Muse with no separate opt-in required from individual merchants, betting that an open agentic-commerce layer will win merchants who don't want to depend on Amazon's marketplace^[3].

Meta is not charging users a subscription for Muse's shopping function; instead it expects to profit over time by taking a small fee from completed transactions, which gives it a direct incentive to maximize how many retailers it signs and how much purchase volume flows through the app^[2]. That explains the rapid sequence of retail deals - Walmart, Gap, Sephora, Wayfair and Best Buy within Muse's first three weeks, followed by a 'Muse for Small Business' push bundling Shopify, Dropbox, Slack, Klaviyo and Canva^[2][4]. The mechanism underneath those purchases is easy to miss: Muse pays through Stripe's Link and Shopify using one-time-use virtual payment cards, so a user's real card number is never shared with the merchant it buys from^[5]. It's a deliberate privacy design choice aimed at the exact anxiety critics raise about agentic shopping - handing an AI your card details - even as the agent gains deep access to email, calendars and health data elsewhere in its feature set.

Walmart is the clearest illustration of how warily retailers are treating Meta's agent. It joined Muse's retail partner list for in-agent purchases, but it is simultaneously embedding its own assistant, 'Sparky,' directly inside ChatGPT and Gemini, and it co-developed an open 'Universal Commerce Protocol' with Google, Shopify, Etsy, Wayfair and Target^[6]. Walmart had already dropped OpenAI's Instant Checkout in March 2026 before this latest round of hedging^[6]. Google is playing a similar multi-front game: its Gemini integration with Walmart predates Muse by months, and Google's own shopping agents have also been blocked by Amazon, meaning Google and Meta are now effectively on the same side of the access fight even as they compete for the same agent-layer business^[7]. The pattern across retailers is consistent - work with Muse for distribution, but don't let any single AI platform become the sole interface to your customers.

Meta's decision to ship Muse for free, at scale, immediately is widely read as a response to mounting competitive pressure. Instinct, a horizontal consumer-agent startup, had just raised a $250 million Series B at a $2.5 billion valuation weeks before Muse launched^[8]. Venture investors describe Meta's move in stark terms: Mayfield's Navin Chaddha called it a 'ChatGPT moment for agents' that could mean 'death or consolidation of other horizontal consumer agents' unable to match Meta's free pricing, compute and distribution, while Boldstart's Ed Sim summed up the dynamic more bluntly - 'he who has the gold that gets the compute makes the rules'^[8].

Muse's growth curve is genuinely fast: it moved from No. 4 to No. 2 on the US App Store within two days of its September 8 launch^[9], and by early October had surpassed 5 million downloads with more than 3 million weekly active users^[10]. Investors have reacted sharply in both directions - Muse-linked partners like Shopify, PayPal and Expedia were reported higher, while travel and media stocks tied to intermediaries that agents might route around fell^[11]. But the experts closest to retail are more cautious than the download chart suggests. GlobalData's Neil Saunders called Muse 'an important part of the journey' rather than a tipping point, citing continued consumer unease about letting AI make purchases^[12], and broader coverage has flagged unresolved questions over fraud and merchant liability when the 'shopper' is software rather than a person^[13]. That caution shows up in public reaction too: broadcast coverage has framed Muse as a potential 'Trojan horse' for data collection even as it ships a friendly, approachable brand, and community discussion skews skeptical of trusting an ad-driven company with purchasing authority and payment access - even as at least one firsthand account describes Muse successfully navigating a retailer's site end-to-end before asking for purchase confirmation, a concrete sign that the underlying agentic behavior works roughly as advertised, trust concerns aside.