Meta Muse: How Meta's AI Shopping Agent Picks Products

On September 8, 2026, Meta launched Meta Muse, a personal AI agent that can browse the web, fill out forms and buy things for the person using it. Meta's launch announcement positions it as an agent that does the work rather than answering questions, and on the same day Shopify added Meta as an AI channel with merchant products shared by default.

For a consumer brand, that means a new reader for your catalog and product pages. Muse picks a handful of products before a person sees your site, sometimes completes the purchase itself, and decides based on the product data Meta already holds and how readable your pages are to a non-human browser.

What is Meta Muse?

Meta Muse is a personal AI agent that runs tasks on a person's behalf, including shopping, using its own cloud browser and saved preferences. Meta says Muse runs on a dedicated cloud machine it calls Muse Secure VM, which has its own browser, and that it can open a browser, fill out forms and negotiate for the person who owns it. It remembers preferences and checks with the user before sensitive actions such as making a purchase.

Shopping is only one of its jobs, but it is the one that touches brands most directly. Meta has also launched Muse for Small Business (September 29), which connects the agent to Meta ad accounts, Facebook Pages and tools such as Shopify and Klaviyo. This article is about the consumer side: Muse as the shopper's agent.

Where is Muse available, and how fast is it growing?

Muse launched in the US as a standalone app on iOS, Android and the web at muse.ai, plus WhatsApp, and has since expanded. Meta's launch post says Muse is rolling out in the US on iOS, Android and muse.ai and that people can message it in WhatsApp. It is not built into the Instagram or Facebook feed, though it can work from Instagram content such as a saved Reel.

By the time Meta published its small business update, it described the consumer agent as available in the US and Canada. Japanese coverage reports a rollout in Japan starting October 23, though Meta's English newsroom had not posted it at the time of writing. We found no announcement of a UK launch.

Adoption has been fast. Sensor Tower estimated that Muse crossed 5 million downloads around September 30, about three weeks after launch. On October 20 it estimated 10.2 million US downloads in 40 days. These are third-party estimates; Meta has not published download or user figures.

How does Muse shop?

Muse clarifies what the person wants, searches Meta's product catalog and merchant websites in parallel, then returns a short list of product cards.

1. Conversation. When a request has shopping intent, Muse may ask follow-up questions to narrow it down, and it checks a stored shopping profile that records the person's tastes by product category.

2. Discovery. Muse then runs two searches at once. One is a query to Meta's product catalog, the same catalog that powers shopping on Facebook and Instagram. The other is a browser session that visits merchant sites directly.

3. Recommendation. Candidates from both sources are filtered and ranked, and Muse shows a small set of product cards. In Profound's test it typically showed six.

The split between the two sources is the important part. In Profound's test, the catalog query returned around 60 products from 30 merchants per prompt, while the browser visited about three domains and returned four products from two merchants. Yet the final six cards were usually two from the catalog and four from the browser. The catalog fills the candidate pool; the browser has an outsized say in what the shopper sees.

As PPC Land summarized them, the file tells Muse to fill its top five picks with well-known merchants, to drop listings priced outside a category's normal range, and to rank a brand's own store first when the shopper names it. Meta has not confirmed the file or said whether it is still in use.

Does Muse check out inside the agent or hand shoppers to your site?

Both happen: some purchases complete inside Muse, while others send the shopper to the merchant's site, depending on where the product came from. At launch, Meta said Muse can pay with Link by Stripe, which issues a one time use card, and that Shop Pay was coming soon.

Shopify's side moved quickly. Its September 8 changelog says products are shared with Meta through Shopify Catalog by default and that merchants can manage catalog access and direct checkout separately. On September 21, Shopify CEO Tobias Lütke said Muse could complete purchases on Shopify stores through Shop Pay, as The Keyword reported, citing The Wall Street Journal. According to Shopify's Meta channel help page, direct checkout is on by default for eligible stores and is shown to shoppers in the US, Canada and Mexico.

That gives two paths, which we will call AI-led and AI-assisted:

• AI-led checkout. The product comes through Meta's catalog with direct checkout enabled, and the shopper pays inside Muse. Your site never loads.

• AI-assisted hand-off. The product was found by Muse's browser on the open web, or you switched direct checkout off. Muse sends the shopper to your product page to finish buying. The leaked instructions limit in-chat checkout to catalog products flagged for it.

One more detail matters for measurement. Shopify's help page notes that browser pixels such as Google Analytics do not fire in Meta's direct checkout; only server-side checkout events are sent. Amazon took a harder line: it blocked Muse from Amazon.com on September 21, citing an agent that does not identify itself.

Are your products already in Muse?

If you sell on Shopify and have Meta's sales channel connected, your products are probably already eligible, because Shopify shares them with Meta by default. Shopify's help page lists the store-level conditions: you sell to customers in the US, Canada or Mexico, your products are eligible for Shopify Catalog, you have accepted the Agentic Storefronts supplemental terms, and your terms of service, privacy policy and refund policy are filled in.

Subscriptions, bundles, customizable products and B2B-only items are excluded from direct checkout, and every product needs a title, an image, a price above zero and a reachable URL.

Opting out is possible but partial. Merchants can switch catalog access or direct checkout off under Sales channels, then Agentic, then Meta. Shopify notes that turning off catalog access does not stop Meta from finding your public pages through crawling. If you are not on Shopify, your route into the catalog is Meta's Commerce Manager catalog, the same feed that runs Facebook and Instagram shopping. Meta has not published Muse-specific eligibility rules for non-Shopify catalogs.

What product data does Muse read?

Muse reads two things: structured attributes in Meta's product catalog, and the content of your product pages as rendered in a real browser session.

Catalog attributes

The catalog side is the structured feed: title, description, price, availability, brand, images, product category and variant fields such as size, color, material, pattern, gender and age group, as defined in Meta's product data specifications. This data decides whether you enter the candidate pool and whether in-chat checkout is possible. If your category is wrong or the variant fields are empty, a request like "waterproof hiking boots for wide feet under $200" has nothing to match on.

On-site PDP content

The browser side reads your page the way a screen reader would. Profound found that Muse navigates through the browser's accessibility tree and recommends semantic HTML with clear labels. What the agent can use from the page is the copy: who the product is for, intended uses, dimensions, compatibility, materials and limitations. Specs locked in images or vague lifestyle copy give it little to compare.

How is Muse different from ChatGPT, Google and Amazon shopping?

Muse differs mainly because it browses merchant sites itself and leans on Meta's existing commerce catalog, rather than a feed built for an AI assistant. Here is how the major platforms compare as of late October 2026.

For a deeper look at the two main protocols, see our breakdown of OpenAI ACP vs Google UCP, and for the broader shift, our explainer on what agentic commerce is.

What should you fix first?

Start with whether you are in Meta's catalog at all, then make the catalog and the product page say the same specific things. In order:

1. Confirm your catalog connection. On Shopify, open Sales channels, then Agentic, then Meta, and check that catalog access is on and decide whether you want direct checkout. Off Shopify, check your Commerce Manager catalog for errors.

2. Fix product categories. Map every product to the most specific category available, in Shopify's taxonomy and in Meta's catalog.

3. Fill variant and attribute fields. Size, color, material, pattern, gender and age group belong in their own fields, not only in the title.

4. Make price and availability match everywhere. The feed, the page and checkout should agree to the cent.

5. Rewrite PDP copy for comparison. State who the product is for, the main use cases, dimensions, compatibility, materials and limitations in plain text near the top of the page.

6. Check that a browser agent can read the page. Use semantic headings and labeled buttons, and do not hide key specs inside images or tabs that only load on click.

7. Fix measurement. Make sure server-side purchase events are set up, because pixels will not see direct checkout orders. Our guide to AI search visibility metrics covers what else to track.

What hasn't Meta said yet?

Meta has not published how Muse ranks products, which merchants qualify for in-chat checkout outside Shopify, or what it will charge merchants for agent sales. Most of what is known comes from Shopify's documentation, third-party tests and an unconfirmed leaked file.

• Ranking. No official documentation on how catalog and browser results are weighted, or whether brand recognition really shapes the top picks.

• Fees. Mark Zuckerberg said in September that Meta expects to earn revenue from Muse through a small fee on transactions, but Meta has not published who pays it or how much.

• Non-Shopify checkout. Meta has not said how merchants on other platforms enable in-chat checkout through Commerce Manager.

• Agent identification. Muse browses without naming itself, so merchants cannot reliably separate its visits from human ones in analytics.

• Geography. No UK date has been announced, and Shopify's direct checkout currently covers the US, Canada and Mexico.

How Nudge helps brands get ready for Meta Muse

Nudge works on both paths Muse uses: the catalog data it queries, and the product page shoppers reach when Muse hands off. On the catalog side, Nudge Catalog Enrichment analyzes every SKU, fills missing attributes, maps products to Shopify taxonomy categories and category metafields, and keeps Google Merchant Center and protocol feeds consistent, so the same specific facts reach Meta, Google and OpenAI. See our guide to catalog enrichment for AI shopping agents.

On the page side, Nudge Funnels adapt what the shopper and the agent see. Organic funnels change the existing product URL in place on your own domain, so a shopper arriving from a Muse hand-off lands on a version that answers the use case, fit and comparison questions the conversation raised. Agent funnels serve a machine-legible version of the page to AI shopping agents that identify themselves, fact-identical to what humans see. Because Muse often browses without identifying itself, the organic page has to stand on its own, which is why both modes matter. Every variant runs against a live control, and nothing ships without your approval. Nudge Shopper Insights then tracks how often your products appear across AI assistants and which pages they cite.

Brands like Mosh & OluKai are using Nudge to enrich their catalogs & grow agentic revenue by 46% and 8x product mentions. For how agent traffic converts once it reaches your site, read how funnels convert AI agent traffic.