How to Optimize for Muse: Getting Your Brand Picked by Meta’s AI Agent

Key Takeaways

• Muse is a separate product from Meta AI. It browses, compares, and checks out for the user, so it needs its own tracking and its own playbook.
• Muse picks in four stages: Find, Check, Rank, Buy. A brand has to clear every stage to reach the shopper.
• Find: Muse searches Meta’s catalog and up to three merchant sites on every shopping request. Catalog presence is the most reliable way in.
• Check: Your product page is a pass/fail test. Muse drops pages that aren’t a real product page, aren’t in stock and addable to cart, or lack a real product image.
• Rank: A reported instruction file calls it “imperative” that the top five products come from well-known merchants. Brand recognition is a direct ranking input.
• Buy: Only catalog products flagged for agentic checkout can be bought in chat. Everything else falls back to browser checkout on your site.
• Track pick rate rather than citations. Muse doesn’t cite the sites it reads.

More than 10 million people in the US and Canada set up Muse accounts within about six weeks of launch, and about 25% of them use it on any given day. Shopping is its most common use case: more than a quarter of Muse users shop with it, at an average purchase size over $200, according to Meta’s Q3 2026 earnings call.

Muse researches products, compares options, and builds the cart for the shopper, who sees a shortlist or a single pick to approve. Visibility now extends past AI answers to the agents that shop on a buyer’s behalf. Muse reads for the user, and brands it filters out never appear.

This guide covers how Muse differs from Meta AI and ChatGPT, the four stages it uses to pick brands, what to fix at each stage, and how to track where you stand.

What Is Muse and How Is It Different From Meta AI?

Muse is Meta’s personal AI agent. You give it a goal, and it carries out the steps: research, comparison, booking, or purchase. Meta launched Muse in the US on September 8, 2026, and it reached Canada on September 18. People use it in the Muse app on iOS and Android, on the web, or in WhatsApp.

Meta AI is Meta’s chat assistant. It’s built into Facebook, Instagram, WhatsApp, and Messenger, and it has its own app. It answers questions and, since early 2026, shows product cards. Muse is a different product built for delegation, and it runs inside its own environment.

Here are the key differences that marketers should know about:

Muse and Meta AI often surface the same products, which is why competitor content treats them as one. In Retailgentic’s tests, a product in Meta AI’s top three results also showed up in Muse 89% of the time. The overlap has limits: in 27% of prompts, the two shared no listing with the same title and seller. Retailgentic attributes part of that gap to the extra context Muse gets from user connectors like email and calendar, plus its broader product coverage.

That’s why Goodie tracks Muse as its own model, separate from Meta AI. A brand can look strong in Meta AI’s answers and still lose the final pick in Muse.

How Does Muse Decide Which Brands to Recommend?

Muse picks in four stages. It finds candidates in two pools, checks each product page, ranks what’s left with a strong bias toward well-known merchants, then routes the purchase through whichever checkout path the product qualifies for. A brand that fails any stage never reaches the shopper.

Meta hasn’t published its ranking method. What we know comes from Meta’s own documentation, an instruction file reported by State of Brand and PPC Land, and independent tests. The file was pulled in late September and may have changed since, and Meta hasn’t commented on it. Treat it as the best current evidence.

1. Find

On each shopping request, Muse searches Meta’s product catalog and runs a live browser search at the same time. The browser side is small by design: it visits no more than three merchant sites, returns no more than ten products, and stops after two sites if the list is already strong and seller-diverse. Shopify stores get a third route. Muse reads the Shopify Catalog through the Universal Commerce Protocol (UCP), the same standard Google uses for checkout in AI Mode and Gemini, according to Retailgentic’s analysis of Muse’s shopping instructions. Facebook Marketplace joins mainly for collectibles, furniture, and auto requests.

2. Check 

Muse reopens every link before showing it. It drops anything that isn’t a real product page (search and category pages are discarded), isn’t in stock with a working add-to-cart, or lacks a real product image.

3. Rank

The instruction file calls it “imperative” that the top five products come from well-known merchants or websites, and it doesn’t define “well-known.” When a shopper names a brand, that brand’s own store goes first, and direct sellers beat resellers by default. Listings priced well above or below the category’s normal range are dropped, and Muse prefers the country version of a site that matches the shopper’s location. Meta also says Muse “remembers what matters to a person,” so stated preferences like size and fit shape the shortlist.

4. Buy

Catalog products flagged for agentic checkout can be carted and bought in chat. Everything else goes through browser checkout on the merchant’s own site.

Why Doesn’t ChatGPT Visibility Carry Over to Muse?

Strong ChatGPT visibility helps a brand at one stage, Rank, and does nothing for Find, Check, or Buy.

In answer engines, citations drive visibility. Results from the AEO Periodic Table V4 study found that earned citations and social and community citations together carry 22% of total weight, more than any single on-page factor. Our expert vs. customer reviews study found expert outlets out-cite customer review platforms roughly 15 to 1 in consumer categories, winning 17 of 21 industries. Those signals carry over to Muse because they build the brand recognition its ranking rule rewards.

The other three stages run on different rules: feeds, catalog data, and whether an agent can act on the page. That’s why the Periodic Table study left shopping surfaces out on purpose. Muse also skips the citation step entirely: it doesn’t identify itself when it browses, and its answers don’t cite the sites it reads. In ChatGPT, a citation is the win. In Muse, the pick is.

ChatGPT finds and compares products, then hands checkout to the merchant. Muse takes the purchase through to approval in chat, so checkout eligibility shapes visibility in a way it doesn’t in ChatGPT. Muse’s rules also change on Meta’s schedule, separate from any answer engine. Our social citations study logged 16 unannounced sourcing changes across five AI models in seven months, and Muse adds its own moving parts on top: catalog rules, browser limits, and checkout partners.

Why Doesn’t ChatGPT Visibility Carry Over to Muse?

Strong ChatGPT visibility helps a brand at one stage, Rank, and does nothing for Find, Check, or Buy.

In answer engines, citations drive visibility. Results from the AEO Periodic Table V4 study found that earned citations and social and community citations together carry 22% of total weight, more than any single on-page factor. Our expert vs. customer reviews study found expert outlets out-cite customer review platforms roughly 15 to 1 in consumer categories, winning 17 of 21 industries. Those signals carry over to Muse because they build the brand recognition its ranking rule rewards.

The other three stages run on different rules: feeds, catalog data, and whether an agent can act on the page. That’s why the Periodic Table study left shopping surfaces out on purpose. Muse also skips the citation step entirely: it doesn’t identify itself when it browses, and its answers don’t cite the sites it reads. In ChatGPT, a citation is the win. In Muse, the pick is.

ChatGPT finds and compares products, then hands checkout to the merchant. Muse takes the purchase through to approval in chat, so checkout eligibility shapes visibility in a way it doesn’t in ChatGPT. Muse’s rules also change on Meta’s schedule, separate from any answer engine. Our social citations study logged 16 unannounced sourcing changes across five AI models in seven months, and Muse adds its own moving parts on top: catalog rules, browser limits, and checkout partners.

How Do You Optimize for Muse?

Optimizing for Muse means working each stage: get into the catalog so Muse can find you, make product pages pass its checks, and build the signals it ranks on. Catalog and page work sits with your ecommerce and dev teams. Ranking signals sit with PR, social, and brand, and they build over quarters. The fourth stage, checkout, gets its own section below.

Get Into Meta’s Catalog With Agentic Checkout On

The catalog is the most reliable way into Find, and the only route to in-chat checkout.

• Shopify stores: Meta became an AI channel in Shopify’s Agentic Storefronts on September 8, 2026, and products are shared with Meta by default through Shopify Catalog. Meta direct checkout covers eligible stores selling to customers in the US, Canada, or Mexico, per Shopify’s Selling on Meta documentation. In your admin, go to Sales channels > Agentic, open Meta, and confirm that Shopify Catalog access and direct checkout are both on.
• Other platforms: Keep a complete, current product feed in Meta Commerce Manager. Muse searches the same catalog that has fed Meta’s shopping surfaces for years.
• Watch for hidden exclusions: On Shopify, products set to Unlisted or tagged with the seo.hidden metafield are hidden from all agentic storefronts, along with sitemaps, search engines, and on-site search. Audit both before assuming a product is in Meta’s catalog.
• Prepare for UCP if you’re not on Shopify. It isn’t a confirmed route into Muse for other platforms yet, but the same integration already powers Google’s AI Mode and Gemini checkout, so the work pays off on Google even if Muse access comes later.

For smaller brands, this is the highest-return step. Catalog presence doesn’t depend on winning one of the three browser visits, where the well-known-merchant rule gives established names first claim.

Make Every Product Page Pass Muse’s Checks

Muse reads your site through a real Chromium browser and sees an accessibility tree snapshot of each page, according to Meta’s security write-up. A page that fails one check is dropped.

Build the Signals Muse Ranks On

The top five slots go to well-known merchants, so brand recognition is an input you have to build on purpose. The rest of the ranking rules reward clean pricing and clear product data.

• Earned media and expert reviews. Expert outlets dominate consumer-category citations in Goodie’s data. Pitch the publications your category’s buyers trust.
• Social proof on Meta’s own platforms. Tag products in Instagram posts and keep Facebook and Instagram shop content current. If a shopper shares an Instagram post with tagged products, Muse skips search and goes straight to those items.
• Branded demand. When a shopper names your brand, Muse puts your own store first. Every branded request is a guaranteed slot.
• Pricing inside the category’s normal range. A steep launch discount reads as an outlier and gets dropped. If you price at a premium, make the reason visible on the page: materials, warranty, certifications.
• Product facts in visible text. Muse filters by stated constraints like size, color, features, and location. Write dimensions, fit notes, compatibility, materials, shipping times, and return terms into the page. Products with missing data are easier to skip.

What Doesn’t Move Muse

Skip the tactics with no evidence behind them. Meta documents no Muse user agent, robots.txt token or special file, so Muse-specific crawler rules and “agent instructions” files do nothing. Hidden text aimed at AI assistants is worse than useless: Muse’s classifiers scan page content for prompt injection and block or flag what they find. And there’s no paid placement to buy. Muse has no ads today.

How Does Checkout Affect Whether Muse Picks You?

Checkout eligibility decides how far Muse can take a purchase on your store. Catalog products flagged for agentic checkout can be added to a Muse cart and bought in chat. Open-web products go through browser checkout, with in-chat checkout switched off, according to the reported instruction file.

PayPal checkout isn’t available in Muse yet, per CNBC, so today every purchase runs through Shop Pay or Link by Stripe. Every path stops for the shopper to approve the total in chat. Merchants should also expect a cost eventually: Zuckerberg has said Meta will take a small fee on transactions Muse facilitates for businesses, but Meta hasn’t published the rate, the terms, or a start date.

Browser checkout is where purchases break. Muse fills forms like a shopper would, so anything that blocks a guest buyer blocks the agent too:

• Forced account creation. Offer guest checkout.
• CAPTCHAs and aggressive bot challenges on cart and checkout pages.
• Unlabeled form fields that an accessibility tree can’t identify.
• Price or stock mismatches between your catalog feed and the live checkout page.

Muse is one of several AI shopping agents with its own checkout rules. Our guide to how AI personal shoppers pick products breaks down what ChatGPT, Perplexity, Google, and Amazon each require, and the product data fixes that carry across all of them.

Optimizing for Muse: FAQs