On September 8th, 2026, Meta launched Muse, a personal AI agent that researches and compares options, books reservations, drafts and sends email, and checks out on retailer websites on a person’s behalf. Muse reached No. 1 on September 18th. September 24th download estimates ranged from 2.3 million to 4.3 million. Wells Fargo, citing Sensor Tower, estimated that Muse reached 448,000 US daily active users on day 10, a level ChatGPT’s mobile app took 49 days to approach. These are specific early mobile-app benchmarks, rather than a comparison of overall adoption. As of September 29, its US App Store listing showed a 4.9-star rating across roughly 55,000 ratings.

Muse is available in the US across iOS, Android, the web, WhatsApp and Mac. It expanded to Canada on September 18, initially through iOS, the web and WhatsApp, with Android still to come at launch. AI-glasses support is on the way, and plans include a free tier and $20 and $100 monthly options. At Meta Connect on September 23, Meta named 10 retail partners, added payment integrations and said it expects to monetize Muse over time through small transaction fees, although no rate, timing or fee structure has been disclosed. No advertising in Muse has been announced. For marketers, Muse opens another path from intent to purchase through the company many brands already rely on for paid social, while product ranking and detailed journey reporting remain open questions.

A Purchase Path That Never Touches an Ad

People describe a goal and Muse researches, compares and acts. Meta says it checks with the person before sensitive steps such as sending an email or making a purchase, although permissions vary by connector and selected defaults. Media reports indicate merchants would bear future transaction fees, but Meta has not disclosed the mechanics. Shopify currently lists no additional Meta direct-checkout fee beyond standard payment processing. Meta also says Muse conversations and data stored inside its virtual machine are not shared with its ad systems. This differs from its broader policy, effective December 16, 2025, of using interactions with generative AI features to personalize content and ads across connected Meta products, including WhatsApp when linked through Accounts Center.

Because Muse introduces no announced ad inventory and separates conversations and virtual-machine data from Meta’s ad systems, we expect limited direct near-term impact on the core Meta auction. Activity across connected services may still create signals under those services’ own tracking arrangements. Over time, as people delegate shopping tasks, more consideration could happen through an agent evaluating product details, price, availability, reviews and policies. The absence of ads today does not establish how Muse will monetize discovery in the future.

Separately, Meta announced Muse Image in July 2026, an image-generation model intended for Advantage+ creative. It is distinct from the personal Muse agent and should be tested within existing human review workflows when available in an advertiser’s account. For now, we recommend monitoring Muse as an emerging discovery channel. A practical first step is a repeatable prompt audit: ask the category questions your customers ask and track whether, and how accurately, your brand and products appear.

Agentic Checkout Has Retailers Choosing Sides

Muse supports payments through Link by Stripe, PayPal and Shop Pay, with Shopify-powered direct checkout available for eligible stores and products in supported markets. Meta direct checkout excludes subscriptions, bundles, customizable products, and B2B-only products. Named partners include Walmart, Best Buy, Sephora, Ulta, Wayfair, Gap and Instacart, with Expedia coming soon. Amazon blocked Muse from Amazon.com on September 20, citing authorization, transparency and security concerns and asking Meta to remove Amazon from the experience. Amazon operates its own agent, Alexa for Shopping, and generated $19.8 billion in advertising revenue in Q2 2026, up 26% year over year. That figure illustrates the scale of retail media; it does not establish Amazon’s motive for the block.

When an agent is doing the shopping, complete and accurate product data, including specs, sizing, price, inventory, shipping and returns, could carry greater weight in selection. Retail media may face greater exposure if agents reduce shoppers’ encounters with sponsored placements. We expect adoption to build gradually. In CNN’s test, a Target checkout required manual entry because the retailer blocked autonomous clicks. OpenAI reduced its emphasis on native Instant Checkout in March 2026 and shifted toward product discovery and merchant-controlled checkout. Walmart separately reported that its in-chat checkout converted at roughly one-third the rate of transactions completed on Walmart although Instant Checkout remains available for some eligible merchants and products.

Since Meta has not disclosed how Muse ranks products, the strongest immediate investments are fundamentals that support multiple agents: complete and consistent feeds, structured site data, accurate inventory and pricing, and clear shipping and return policies. Shopify merchants should confirm eligibility and how direct checkout is configured for their store. DTC brands should align ecommerce, legal and security teams on whether to allow, partner with or restrict agents, so their access reflects a deliberate business decision.

Agent Traffic Is Measurement’s Newest Blind Spot

Muse runs inside a dedicated cloud virtual machine with its own browser and keeps working after the person closes the app. Granular Muse-specific referral and journey reporting remains limited. Shopify merchants do receive Meta sales-channel attribution for orders completed on Meta surfaces, but Google Analytics and third-party client-side pixels do not fire during direct checkout. Browser-based Muse visits outside supported commerce integrations may still appear as direct, unattributed, or automated traffic, or be blocked by bot-management systems. These gaps could obscure paid media’s role in building the awareness that leads someone to ask an agent for a brand. Media mix modeling and incrementality testing can help complement click-based attribution. The practical move now is to reconcile sales-channel reporting with site analytics, segment identifiable agent traffic, review bot-management rules, and establish a baseline.

Where the Impact Lands First

Based on information available as of September 29, our early read is that Muse itself is likely to bring limited direct change over the next 6 to 12 months to Meta paid social delivery and targeting, Google Search and Performance Max, B2B lead generation, and CTV and programmatic. Shopify and DTC brands, subscription and service businesses, web analytics, and organic search and GEO may feel an effect depending on category, access and data readiness. Higher potential exposure sits with retail media networks, on-site ad monetization, and customer service and retention workflows that agents can interact or negotiate with. Product feed and catalog quality are immediate readiness priorities across these groups. These are hypotheses to revisit as adoption and reporting develop.

Consumer Trust Will Set the Pace

Early reviewers praised Muse’s handling of multi-step tasks while documenting limits, including recommendations for restaurants that had closed years earlier. Researchers also disclosed a serious Mac vulnerability that Meta subsequently patched. A separate report about access to a user’s own virtual-machine filesystem raised concerns that Meta disputed, describing the access as intended behavior. A more recent Marketplace incident involved Muse sharing a user’s address after an “Allow Always” permission was enabled, highlighting questions about permission clarity and user expectations. In Coveo’s 2026 study of 4,000 US and UK shoppers, 16% were very comfortable letting an AI assistant make purchases, 33% somewhat comfortable, and 45% uncomfortable. Forecasts that predate Muse are substantial: Bain projects $300 billion to $500 billion in US agentic commerce by 2030, and Morgan Stanley projects $190 billion to $385 billion. The firms use different definitions, so those estimates are not directly comparable. We expect reliability, clear permissions, and consumer confidence to help determine how quickly delegated purchasing reaches meaningful scale.

Muse also joins Google’s UCP-enabled shopping experiences, shopping and checkout inside ChatGPT, and Amazon’s Alexa for Shopping and Buy for Me features. Google’s Universal Commerce Protocol is infrastructure supporting commerce experiences, rather than a consumer agent itself. Preparing for AI agents therefore requires work across platforms. Muse’s early growth is notable, but comparisons with ChatGPT should remain specific to the mobile-app download and daily-user benchmarks measured.

What to Do in the Next 30 Days

With Muse on AI glasses, a pocket device called Muse Charm, and additional connectors on the way, platforms and retailers are still deciding how to respond. Now is a useful time to evaluate exposure and establish baselines. We do not recommend changing media plans or budgets based on Muse’s early adoption alone:

• Product data audit: Review feeds and structured data for completeness, accuracy, and current pricing and inventory.

• Agent access decision: Align ecommerce, legal and security on access, participation and checkout settings.

• Analytics baseline: Reconcile sales-channel reporting with site analytics and segment identifiable agent traffic.

• Prompt audit: Track how Muse describes and recommends your brand and products.

• Creative readiness: Plan a governed Muse Image test once it is available in your account.

Muse is still early, and important details, including future fees, product ranking, and granular journey reporting, remain uncertain. Brands can use this period to improve product data, define agent access and establish measurement baselines. A follow-up in 60 to 90 days should assess observed traffic, sales, and customer behavior, alongside new platform reporting, before those findings inform changes to media plans.¡¡™