ReFiBuy上线Meta Muse适配服务 瞄准AI代理购物流量

Key takeaways

1. On November 4, 2026, ReFiBuy launched adaptation support for Muse, Meta's personal AI agent, and opened early access, offering brands and retailers that sell through the Muse channel a closed-loop intelligent product information optimization service.

2. Within six weeks of launch, more than 10 million users in the U.S. and Canada had completed setup. Shopping is the most-used feature, with over a quarter of weekly active users shopping through Muse, and average order value exceeding US$200 including travel spending.

Practical guidance

1. When brand product data is not adapted, Muse prioritizes competing products with more complete data. Products in the top three Meta AI search results have an 89% chance of also appearing in Muse recommendations.

2. Merchants can complete consumer-scenario-level data optimization for 5% to 10% of their core SKUs before Black Friday, and use the monitoring and optimization loop to track impressions, rankings, and information accuracy.

Channel and traffic opportunity

1. Muse has become an important shopping entry point for Meta's personal AI agent. Within six weeks of launch, more than 10 million users in the U.S. and Canada completed setup. Shopping is the most-used feature, with over a quarter of weekly active users shopping through Muse and average order value exceeding US$200 including travel spending.

2. Black Friday is the first major promotional moment since Muse launched. Millions of consumers will use personal AI agents to shop during the holiday season, and brands can reach this group through the adaptation service.

Data and competition

1. Most brand product data has not been adapted. The audience, scenarios, and product fit that consumers care about remain largely with operations teams and in customer reviews, and are not synced to the data layer Muse can read. When data is insufficient, Muse prioritizes competing products.

2. Under the five-level product data richness framework, basic structured data is only Level 1. Precise matching requires at least Level 2 conversational attribute data, and ideally Level 3 scenario-matching data.

Product and operations

1. ReFiBuy uses content agents and attribute agents to fill in descriptions, Q&A, and structured details SKU by SKU. A review agent validates the output and attaches supporting evidence; merchants can set review rules, and for Shopify brands, approved content syncs to the storefront.

2. It monitors impressions, competitor rankings, and information accuracy across Muse and the Meta AI product catalog, with underperforming issues entering the next optimization cycle.

Growth market

1. Within six weeks of launch, more than 10 million users in the U.S. and Canada completed setup for Muse. Shopping is the most-used feature, with over a quarter of weekly active users shopping through Muse and average order value exceeding US$200 including travel spending. Black Friday is the first major promotional moment.

2. ReFiBuy has launched its Muse adaptation service and opened early access to brands and retailers in the U.S. and Canada. Merchants can work with ReFiBuy to identify product data gaps, review the first batch of optimized content, and set up Muse-side monitoring for priority products.

Opportunities and risks

1. The opportunity lies in reaching millions of consumers who will use personal AI agents during the holiday shopping season through the Muse channel; adapted product data can increase the likelihood of being recommended by AI agents.

2. The risk is that most brand product data is not adapted. Information on target audience, scenarios, and fit is not synced to the data layer Muse can read, and when the available data cannot answer user queries, Muse prioritizes competing products with complete data.

Response measures

1. Under the five-level data richness framework, precise matching by personal AI agents requires at least Level 2 conversational attribute data, and ideally Level 3 scenario-matching data.

2. In practice, merchants should complete consumer-scenario-level data optimization for 5% to 10% of core SKUs before Black Friday, and use the monitoring and optimization loop to track impressions, rankings, and information accuracy.

Product data adaptation requirements

1. For a personal AI agent to match products precisely to specific users, at least Level 2 conversational attribute data is required, and Level 3 consumer-scenario matching data is the ideal; basic structured data alone is Level 1 and is unlikely to support precise matching.

2. ReFiBuy's content agents generate product descriptions and FAQs, while attribute agents fill in the structured details needed for shopping agent matching across all product variants; a review agent validates the output and attaches supporting evidence.

Business opportunity and digitalization implications

1. Within six weeks of launch, more than 10 million users in the U.S. and Canada completed setup for Muse. Shopping is the most-used feature, with over a quarter of weekly active users shopping through Muse and average order value exceeding US$200 including travel spending. Black Friday is the first major promotional moment.

2. Factory-type merchants should focus on building their product data layer, structuring information such as target audience, use scenarios, and fit so AI agents can read it; for Shopify brands, approved content syncs to the storefront, ensuring consistency between product pages and the information AI agents crawl across the web.

3. In practice, prioritize scenario-level data for 5% to 10% of core SKUs before Black Friday, and keep refining based on monitoring of Muse and Meta AI product catalog results.

Industry trends

1. AI agent shopping is emerging as a new traffic entry point. Within six weeks of launch, more than 10 million users in the U.S. and Canada completed setup for Muse. Shopping is the most-used feature, with over a quarter of weekly active users shopping through Muse and average order value exceeding US$200 including travel spending.

2. Product data richness is divided into five levels: basic structured data is only Level 1, precise matching requires at least Level 2 conversational attribute data, and ideally Level 3 scenario-matching data. Products in the top three Meta AI search results have an 89% chance of also appearing in Muse recommendations.

Client pain points

1. Most brand product data has not been adapted. The audience, scenarios, and fit that consumers care about mostly sit with operations teams and in user reviews, and are rarely synced to the data layer Muse can read.

2. When existing data cannot answer user queries, Muse prioritizes competing products with complete data. Brands also need to monitor results on both Muse and the Meta AI product catalog.

Solution

1. ReFiBuy's commerce intelligence engine uses content agents and attribute agents to fill data gaps SKU by SKU, and a review agent to validate output with supporting evidence. Merchants can set review rules, and the system records approval and rejection preferences; for Shopify brands, optimized content syncs to the storefront.

2. The monitoring component directly monitors Muse search results and Meta AI product catalog results, tracking impressions, competitor rankings, and information accuracy, with underperforming issues entering the next optimization cycle. The closed loop already covers ChatGPT, Microsoft Copilot, Google AI Mode, and Amazon Alexa shopping channels.

Platform requirements and problems

1. AI agent shopping requires product data that can be read and matched, but basic structured data is only Level 1. Precise matching by personal AI agents requires at least Level 2 conversational attribute data, and ideally Level 3 consumer-scenario matching data.

2. Most brand product data is not adapted, and information on target audience, scenarios, and fit is not synced to the data layer Muse can read. When data is insufficient, Muse prioritizes competing products with complete data, so platforms need to push merchants to complete their product data.

Operations and risk control

1. ReFiBuy monitors Muse search results as well as Meta AI product catalog results, which Muse relies on heavily, tracking whether brand products are shown, how they rank against competitors, and whether displayed information is accurate.

2. Underperforming issues are added to the next content optimization cycle for merchant review. Merchants can set review rules to determine content publishing standards, and the agents record preferences for approved and rejected content.

Merchant recruitment and ecosystem

1. The Muse adaptation service is now open for early access to brands and retailers in the U.S. and Canada. Approved merchants can work with ReFiBuy to identify data gaps, review the first batch of optimized content, and set up Muse-side monitoring for priority products.

2. For Shopify brands, approved optimized content syncs to the storefront, ensuring consistency between product pages and the structured data crawled by AI agents across the web. Merchants are advised to complete scenario-level data optimization for 5% to 10% of core SKUs before Black Friday.

New industry developments

1. ReFiBuy's launch of an adaptation service for Muse, Meta's personal AI agent, marks the entry of AI agent shopping into the commercial services stage. Within six weeks of launch, more than 10 million users in the U.S. and Canada completed setup for Muse. Shopping is the most-used feature, with over a quarter of weekly active users shopping through Muse and average order value exceeding US$200 including travel spending.

2. Product information is sourced from Meta's own product catalog, Shopify storefronts, and product pages across the web. Products ranked in the top three of Meta AI search results have an 89% chance of also appearing in Muse recommendations.

New questions and business models

1. Brand product data is not adapted. The audience, scenarios, and fit that consumers care about mostly remain with operations teams and in reviews, and are not synced to the data layer Muse can read. When data is insufficient, Muse prioritizes competing products, creating a competition over information completeness.

2. The five-level product data richness framework proposes Level 1 basic structured data, Level 2 conversational attributes, and Level 3 consumer-scenario matching, providing an analytical framework for AI agent product matching.

3. Service providers use content agents, attribute agents, and review agents to fill data gaps SKU by SKU, and form a business model through a monitoring and optimization loop. Monitoring already covers ChatGPT, Microsoft Copilot, Google AI Mode, and Amazon Alexa, making cross-platform AI visibility management a new topic.