For a Shopify merchant, the priority is to distinguish four states: the announcement of a channel, access to catalog data, the ability to pay, and the recording of an order. For an SEO or acquisition team, a fifth state matters: being able to link that order to the journey that produced it. A presence in a catalog provides none of these proofs on its own.
The companies behind the Muse ecosystem
Shopify
1PasswordCompanies and services cited in Meta’s official announcement of September 8, 2026. Source [1]
What Muse changes in the shopping journey
Meta describes Muse as a personal agent able to work over time, use a browser and continue a task after the app is closed. Its commercial relevance therefore goes beyond answering a product search. The agent can prepare an action, come back when a condition changes and ask for approval before the purchase. The announcement provides for the Muse app and WhatsApp as conversation interfaces. It places the initial rollout in the United States, on iOS, Android and the web. [1]
The technical documentation specifies the payment flow described at launch: Link provides a single-use card tied to a merchant, an amount and a validity period. The agent asks for human approval. Shop Pay also appears there as an upcoming integration. These are capabilities declared by Meta, with no purchase test carried out for this analysis. [2]
Official demonstration · Muse by MetaStripe provides a useful cross-check in its press release of September 8: US consumers can connect Link to Muse. It describes payment with the saved payment method at merchants that accept Link, and a single-use virtual card for other supported purchases. The amount must be approved in the conversation. This confirmation covers Link, without establishing the availability of Shop Pay in Muse. [17]
The implication for a store is concrete: an agent that browses may encounter the product page, the shipping terms and the site's checkout. A channel that offers an integrated checkout may present a different interface. Data, customizations and measurement do not necessarily follow the same path. The journey must be identified before concluding that a catalog, theme or tracking change will solve the problem.
| State | What can be established | What remains to be proven |
|---|---|---|
| Announcement | A platform describes a capability and a timeline | Access for a given user |
| Catalog | Data is accessible to a channel | Its selection and display for a query |
| Payment | A flow is documented for an eligible scope | Whether it works for the product, the country and the store |
| Order | An order appears in the merchant system | Its precise origin, its payment and its outcome after returns |
Four states, four proofs
Announcement
A channel is presented by the platform.
Catalog
Product data is accessible.
Payment
A payment flow is available.
Order
A purchase is actually recorded.
Articles by Verity ScoreThis grid is a reading method, not a measure of Muse's performance. In particular, it avoids turning “available products” into “completed sales”.
Shop Pay: a discrepancy that requires naming the surface
Meta's page of September 8 explicitly states that Shop Pay is coming soon. The technical publication on safety repeats this caveat. In parallel, the Shopify help page “Selling on Meta” describes a direct checkout enabled by default for eligible stores. It refers to Meta surfaces, without establishing a timeline specific to Muse and Shop Pay. [1][2][3]
Shop Pay: two scopes to keep apart
Meta · Muse
Meta’s announcement states that Shop Pay is coming soon.
Shopify · Meta surfaces
The Help Center documents a direct checkout for eligible stores.
Articles by Verity ScoreA difference in timeline, journey or population could explain the gap. These are hypotheses. The sources reviewed do not make it possible to choose between them. The defensible wording is therefore the following: Shopify documents Meta direct checkout; the general availability of Shop Pay in Muse is not confirmed by the Meta announcement reviewed.
| Dated milestone | Source | Actual scope |
|---|---|---|
| January 11, 2026 | Shopify announces UCP, new integrations and the Agentic plan | Infrastructure context; no proof of the Muse launch |
| May 11, 2026 | Shopify Changelog: new Agentic page in the admin | Centralization of settings and measurement |
| July 7, 2026 | Displayed date of the Agentic Storefronts terms reviewed | Contractual framework; not the availability date of each channel |
| September 8, 2026 | Meta presents Muse | US rollout, Link described, Shop Pay coming soon |
| September 9, 2026 | Review of the Shopify and Meta documents | Documentary snapshot, with no test order |
Sources for the timeline: Shopify, announcements and terms; Meta, launch announcement. Documentation with no visible update date is not artificially dated to the day it was accessed. [1][8][9][10]
Who is eligible, and for what?
Shopify states that Meta direct checkout is shown to customers located in the United States. The stated condition is selling to US customers. This makes it possible neither to announce availability for French shoppers, nor to conclude that all merchants established in France are excluded. The shopper's country and the seller's country of establishment are two different pieces of information. [3]
Catalog eligibility is another filter. Shopify notably requires a Starter plan or higher, with a documented exception for the Agentic plan, a store that is not in private mode, a title, at least one image and a price greater than zero. Publication on compatible channels and an identifiable product URL also count. Products that are unlisted or hidden from search engines are excluded. Shopify also reviews the account's standing and history; a correctly completed product listing therefore does not guarantee the store's admission. [4]
| Situation | Documented state as of September 9 | Consequence for the merchant |
|---|---|---|
| Shopper in the United States | Muse rollout announced; Meta direct checkout planned for US customers | Verify access and the exact journey |
| Shopper in France | The announcement reviewed does not confirm a French launch | Do not present Muse as an available local channel |
| Eligible store with eligible products | Meta direct checkout described as active by default | Read the channel's actual state in the admin |
| Subscription, bundle, customizable or B2B-only product | Not supported in Meta direct checkout | Do not extend catalog eligibility to this checkout |
| Agentic plan without a main Shopify store | Specific procedure; direct checkout disabled by default | Do not apply the automations of the other plans |
| Meta shoppable ads | Facebook and Instagram channel with feed sync required | Distinguish this feed from discovery via Catalog |
The payment restrictions come from Shopify's Meta page, the catalog rules from its eligibility page, and the Agentic plan exception from its dedicated guide. This table is not a verdict on any particular store. [3][4][7]
The limits also affect the experience. Shopify excludes certain checkout blocks from Meta direct checkout, as well as local delivery, in-store pickup and pickup points. Automatic discounts and discount codes are described as supported. Shopify Functions relating to, among other things, shipping and validation may work, but this does not prove that every site customization carries over into Meta. [3]
Catalog: the integration reduces work, it does not erase the checks
Shopify's page dedicated to Meta states that discovery relies on Shopify Catalog. It explicitly ties the Facebook and Instagram feed requirement to shoppable ads. Another page, dedicated to product discovery, more broadly describes Meta surfaces as powered by that Facebook and Instagram channel. The documentation is therefore not perfectly aligned in its wording. [3][5]
To handle this discrepancy, the Meta-specific page is the most precise reference on the discovery/advertising distinction. It does not justify installing an advertising channel simply because an email mentions Muse. For a given store, the right check is to examine the intended surface, its state in Agentic and its current terms. An existing integration should not be modified on the basis of a generalization drawn from another page.
Shopify describes Catalog as a structured source of titles, descriptions, options, images, prices and availability, updated continuously. This promise does not provide a measured maximum delay between a stock change and its reflection in Muse. Nor does it demonstrate that the right variant will always be selected for an ambiguous request. [5]
An ecommerce team can prepare a verification without changing anything: choose a SKU and a variant, note their price and currency, record availability for the target market, then compare the information at each accessible step of the journey. If a Muse observation becomes possible, it must carry the same timestamp and the same context. A discrepancy should be kept as an observed discrepancy; its cause is not automatically a faulty Shopify sync.
For catalogs that use custom fields, the issue may be mapping: the data exists in the store but is not carried into the field expected by Catalog. Shopify documents this possibility. In SEO, the useful work then consists of making the attributes actually available and understandable, with the usage constraints, dimensions or compatibilities that make a decision possible. Adding adjectives or lengthening a description without information does not provide that proof. [5]
The Verity method on product representation gaps provides a complementary framework for comparing the approved fact with what is displayed. Its field is ChatGPT; its findings specific to that platform are not results measured on Muse. [13]
A Meta order does not measure Muse’s contribution on its own
Shopify states that orders placed on Meta surfaces are attributed to Meta in the admin. The documentation reviewed does not demonstrate the existence of a field that isolates Muse across all of these orders. An increase in the Meta channel therefore cannot automatically become an increase in Muse sales. [3]
Direct checkout introduces another measurement break. Shopify specifies that Google Analytics and custom pixels do not fire in that checkout. The events described are sent server to server for the start and the end of checkout. The acquisition lead should avoid comparing this journey, without precaution, to a web conversion measured by browser-side pixels. [3]
The Agentic page presents reports on sales, orders, store sessions and conversions from those sessions. These measures do not all share the same scope. Dividing all orders attributed to a channel by store sessions alone can produce a meaningless rate if some orders were completed off-site. [6]
From visibility to order
Visibility
The product appears in an answer or a catalog.
Visit
A session is measured on the store.
Attributed order
A sale is linked to a source based on the available data.
Articles by Verity ScoreFor each analysis, keep the period, the available channel or referrer, the known place of completion, and the cancelled or refunded orders. If the Muse origin is not exposed, the category remains “Meta, surface not isolated”. This discipline leaves an unknown visible rather than fabricating a result.
The Verity guide to Share of AI Voice is useful for defining visibility as a distinct indicator. A presence in an answer, a visit and an order have different denominators. No Muse traffic volume, conversion rate or revenue was measured in this analysis. [14]
Fees, returns and data: what the merchant keeps
Shopify's Meta page states that no specific fees are added for selling through its direct checkout, apart from the usual payment processing fees. “No additional fees” therefore does not mean “free transaction”. The Agentic Storefronts terms also provide for the possibility of fees depending on the channel, presented in the admin. The current rule documented for Meta does not constitute a guarantee of permanent free use for all channels. [3][8]
The merchant remains the merchant of record. The merchant manages orders, their fulfillment, returns and customer service. A purchase initiated in a conversation does not automatically transfer these responsibilities to the agent. The store's policies must be present and up to date; they are part of the terms published by Shopify. [3]
Meta also mentions Link protections. The Link page reviewed reserves them for eligible purchases and specifies that Cover Genius assesses coverage. It would be wrong to promise them for every Shopify order or for a future Shop Pay checkout. Customer service must first identify the payment method, the journey and the terms applicable to the case. [1][11]
On the data side, Shopify describes channel access to product information and, when an order is placed through a direct checkout, to the necessary data such as the customer's name, email or address. The documentation states that the full order history and customer base are not opened to the channel. This remains different from an absence of sharing: the data needed for the transaction does flow. [12]
Muse’s privacy limits are already documented
Meta describes a virtual machine dedicated to each user, isolated credential storage and a Sentinel agent that controls outbound actions. The same publication acknowledges that Muse can make mistakes and be subject to prompt injection attacks. These descriptions constitute the architecture declared by its publisher; they do not amount to independent certification in this analysis. [2]
The distinction between the virtual machine versions is essential. Meta states that the current architecture may allow access to certain data in order to operate, secure or support the service. The Confidential VM version, meant to cryptographically prevent that access, is announced for later in the year. Presenting this future property as achieved at launch would be inaccurate. [2]
The advertising policy calls for the same precision. Meta says it does not pass conversations or VM data to its advertising systems. But its documentation specifies that visits made by Muse to websites may be treated as the user's activity and influence the ads they see afterward. The absence of direct sharing of conversations does not mean the absence of any advertising consequence of the journey. [2]
Finally, Meta explains that it uses interaction trajectories scrubbed of certain personal information to train its models, with an opt-out option in the settings. This mechanism must be distinguished from credential storage, advertising sharing and the order data accessible to partners. These are different processing operations, subject to different controls. [2]
The five decisions to make before increasing acquisition effort
| Proposed priority | Owner | Useful verification | Decision the evidence allows |
|---|---|---|---|
| Qualify the scope | Ecommerce | Shopper country, product, channel and enabled payment | Identify a candidate journey, without announcing universal availability |
| Check one SKU | Catalog and SEO | Variant, price, currency, stock, attributes and policies | Correct an identified missing or contradictory data point |
| Examine the exceptions | Ecommerce and technical | Bundles, customization, subscriptions, shipping, checkout functions | Avoid a promise that the direct journey cannot keep |
| Define the measurement | Acquisition and analytics | Available attribution, place of payment, sessions, cancellations and returns | Distinguish channel observation from incremental revenue |
| Review the controls | Store manager | Catalog, direct checkout, future enrollment and terms | Choose channels with knowledge of their effects |
This plan is a recommendation. It starts with reversible readings and comparisons. It does not set a universal timeline or an advertising budget: those depend on the catalog, on access to the product and on the results actually observed.
The settings are found under Sales channels > Agentic. Shopify allows catalog access and direct checkout to be managed separately when automatic management is disabled. Removing Catalog can take up to seven days for sharing to stop, according to its documentation, and does not necessarily prevent discovery through web crawling. Hiding a product entirely may affect its visibility in search engines; this is not a simple Muse preference. [6]
The launch justifies preparing the data and the measurement. It does not yet justify presenting Muse as a proven revenue source for a store. The next useful result is a verifiable trace linking a product, an eligible shopper, a payment flow and an attributable order. As long as one of these elements is missing, it must remain named.
Sources and methodology
Documentary analysis closed on September 9, 2026. Official sources searched and then re-read, claims cross-checked by surface and date. No store, messaging or customer data consulted; no purchase, setting change or field test. Hypotheses and recommendations are distinguished from the platforms' statements. Local appendices retain the extractions, contradictions and checks.
[1] Meta. Introducing Muse: The World’s First Personal AI Agent Built for Everyone. September 8, 2026. Rollout announcement, not proof of universal access.
[2] Meta. How We Built Safety Into Muse. September 8, 2026 according to the metadata; accessed September 9. Architecture, purchases, data and limits declared by the publisher.
[3] Shopify Help Center. Selling on Meta. Accessed September 9, 2026; update date not displayed in the body reviewed. Meta scope, with no Muse-specific timeline.
[4] Shopify Help Center. Requirements for being included in Shopify Catalog. Accessed September 9, 2026. Catalog conditions, not a guarantee of recommendation.
[5] Shopify Help Center. Shopify Catalog and product discovery for agentic storefronts. Accessed September 9, 2026. Wording discrepancy on Meta surfaces retained.
[6] Shopify Help Center. Managing agentic storefronts. Accessed September 9, 2026. Settings and reports described, without verification of an account.
[7] Shopify Help Center. Setting up agentic storefronts for stores on the Agentic plan. Accessed September 9, 2026. Procedure for a separate plan.
[8] Shopify. Agentic Storefronts Supplemental Terms of Service. Displayed update: July 7, 2026, fresh read on September 9. An earlier extraction returning May 25 was discarded for the date and version.
[9] Shopify. The agentic commerce platform. January 11, 2026. Historical context, without projecting UCP capabilities onto Muse.
[10] Shopify Changelog. A new home for Agentic Storefronts in your admin. May 11, 2026.
[11] Link. What’s covered with protections?. Accessed September 9, 2026. Conditional coverage; no individual contract verified.
[12] Shopify Help Center. Data sharing and privacy for selling with agentic storefronts. Accessed September 9, 2026.
[13] Verity Score. ChatGPT and Shopify: Audit the Product Representation Gap. August 31, 2026; re-read September 9. Methodological complement, not Muse proof.
[14] Verity Score. Share of AI Voice: 5 Engines, Not 1. Accessed September 9, 2026. Definitional complement, without reuse of its metrics.
[15] Meta. How We Designed Muse. September 2026. Design reference read for context; team anecdotes not reused as independent observations.
[16] Shopify. Sell in AI chats with Shopify. Accessed September 9, 2026. Commercial page checked against the more precise terms of the Help Center.
[17] Stripe. Stripe helps Muse shop across the internet with Link. September 8, 2026; accessed September 9. Press release from the payment platform, with no independent purchase.
[18] Universal Commerce Protocol Blog. Guide des catalogues produits pour les agents IA. Accessed September 9, 2026. Methodological complement on catalog structure; does not prove a UCP integration in Muse.