- Meta vs. Google in the Agentic Commerce Race: Meta’s Muse directly challenges Google’s AI shopping agent, betting on simulated human browsing against Google’s open standard.
- The Incumbents’ Dilemma: As AI shopping agents arrive, existing shopping platforms face a dilemma: open their systems or build defensive walls, with Amazon locking out agents while online travel platforms facing disintermediation risks.
- The China Experience: Alibaba and Tencent demonstrates that frictionless execution relies on owning the entire loop—from product discovery to digital payments.
“Find me a hotel in Lisbon under €150, near the trams.”
An AI agent searches, compares, and pays. The booking is confirmed.
This is Meta’s Muse, its consumer AI app that surged to #1 on the US App Store within days of launch. The debut marked Meta’s challenge to Google in the race to become the default AI shopping agent for consumers.
The central question is not which company builds the smarter frontier model, but who controls the end-to-end transaction loop.
Can Google and Meta bypass existing marketplaces—or will the incumbents defend their territory?
As Google and Meta attempt to put together a fragmented ecosystem, does China’s closed-loop model reveal the end-state of agentic commerce?
The Consumer Agentic Race: Google vs. Meta
Both platforms want to be the default shopping assistant for consumers—handling discovery, comparison, and payment inside their own apps. But neither owns warehouses, delivery, or payment infrastructure.
To turn search into actual purchases, they chose different approaches:
Google is setting up an open industry standard for agentic shopping.
Meta is using AI agents that browses web stores just like a human.
The Race Timeline
January 2026: Google introduces the Universal Commerce Protocol (UCP) with Shopify, Etsy, Wayfair, Target, and Walmart, setting open data and checkout standards for agents.
May 2026: Google rolls out Universal Cart across Search and AI Mode, pairing it with AP2 (Agent Payments Protocol) for native, on-page checkout.
September 2026: Meta launches Muse as a standalone app and inside WhatsApp, giving each user a dedicated cloud virtual machine with its own browser, terminal, and file system to perform agentic AI shopping. Muse shortly overtook ChatGPT to claim #1 on the US App Store.
How They Solve the Agentic Wall
There are two obstacles for agentic commerce: fragile web scraping for product information, and anti-fraud engines blocking autonomous card payments.
Google and Meta addressed these problems through different strategies:
Google: Building open standard. Google is building an open commerce standard with partners like Shopify, Walmart, and Target, backed by payment giants like Visa and Mastercard. It introduces Universal Commerce Protocol (UCP) and Agent Payments Protocol (AP2) to links AI search directly to store backends to check inventory, create carts, and handle payments.
However, this model depends entirely on retailer cooperation. Stores must adopt Google’s protocol while handing over product discovery and checkout to Google Search and Gemini—a trade-off many brands may resist to protect their direct customer relationships.
Meta: Dedicated AI agent per user: Meta takes the opposite path by simulating human shopping. Its AI agent Muse operates inside cloud-hosted virtual machines, browsing store pages, selecting options, and filling checkout forms for users. Payments are handled via Stripe Link, either checking out directly or generating single-use virtual cards.
To boost reliability, Meta is also building direct integrations with partners like Shopify, PayPal, Expedia, and Instacart for real-time inventory and instant checkout.
However, running a virtual computer for every shopper demands massive compute. Meta raised its 2026 capital expenditure guidance to $130–$145 billion, nearly doubling the $72 billion spent in 2025. In addition, merchants could actively deploy anti-bot defenses to block simulated traffic.
| Meta | ||
| Core Architecture | Server-Side Protocol (UCP): Standardizes product feeds, live inventory queries, and checkout schemas across merchant APIs. | Cloud Virtual Machine (VM): Provisions a sandboxed Linux container in Meta’s data center running a headless browser and AI agent. |
| Payment | AP2 (Agent Payments Protocol): Exits to merchant gateways via cryptographically verified payment tokens. | Sentinel + Stripe Link: Meta’s Sentinel gatekeeper halts execution for user sign-off, then issues a single-use virtual card. |
| Merchant Buy-in Required | Yes | No |
| User Interface | Embedded natively inside Google Search and AI Mode. | Direct messaging inside WhatsApp and standalone mobile apps. No PC or local compute required. |
| Vulnerability | Retailers refusing to adopt UCP due to fear of commoditization. | Cloud compute cost and aggressive merchant bot-blocking. |
The Incumbents’ Dilemma: Partner or Build the Wall?
As autonomous AI agents arrive, existing shopping platforms face a dilemma: open their systems or build defensive walls.
Amazon chose the wall: Amazon blocked Meta’s Muse and rejected Google’s open standard. Its built-in shopping assistant, Rufus, already reaches over 300 million users and drives $12 billion in incremental annual sales, lifting purchase likelihood by more than 60%. In addition, its “Buy for Me” feature reverses the dynamic and allows customers to buy from external brand sites directly within Amazon’s app.
The travel intermediary trap: Online travel platforms face a higher risk. Booking and Expedia spend billions each year on search engine ads to acquire customers. If AI agents from Google and Meta book directly into hotel reservation systems, their intermediary business model would be challenged.
Travel platforms are defending themselves on three fronts:
In-App Assistants: Rolling out their own AI tools (such as Booking’s Penny and Expedia’s Romie) to keep users from searching elsewhere.
Locked Tiers: Hiding member discounts and loyalty points behind logins to block external scrapers.
Wholesale Supply: Acting as backend fulfillment partners, such as Expedia supplying inventory to Meta’s Muse, to monetize transactions even when discovery starts upstream.
Shopify: Integration by default: While marketplaces are protecting traffic against AI agents, Shopify operates on a different business model. It does not run a shopping app; it powers merchant backends and monetises on transaction volume.
As Shopify monetises on sales regardless of where the transaction take places, it does not need to block AI agents. Instead, it co-developed Google’s Universal Commerce Protocol (UCP) and partnered with Meta to let shoppers check out via Shop Pay inside WhatsApp. For Shopify, AI agents aren’t a threat—they are just another sales channel.
The China Experience: One Company, Every Step
In China, the agentic loop didn’t need to be negotiated among search engines, retail stores, and payment networks. It was built inside vertically integrated ecosystems from day one.
Alibaba: The Full-Stack Closed Loop
Alibaba owns every component in the loop and turned its AI model, Qwen, into an autonomous shopping agent.
Products: Direct access to Taobao and Tmall.
Services: Travel booking through Fliggy, and food delivery via Ele.me.
Payment: Frictionless settlement through Alipay.
As inventory, pricing, delivery logistics, and payments all live on Alibaba’s ecosystem, transactions execute cleanly end-to-end inside the app.
In August 2026, Alibaba expanded its ecosystem by opening Qwen to outside partners across logistics (SF Express), mobility (Hello TransTech), and housing (Ziroom), while embedding the transaction engine directly into Honor smartphones and Lenovo PCs.
Tencent & WeChat: The Blueprint Meta Wants to Build
If Alibaba mirrors Amazon, WeChat is the ecosystem Meta is trying to replicate.
A decade ago, WeChat solved the problem of fragmented apps by inventing Mini-Programs—lightweight services that run directly inside its messaging app alongside WeChat Pay.
Today, Tencent simply opened that infrastructure to its AI agent, Xiaowei, alongside launch partners like JD.com, Meituan, Trip.com, and KFC China. Instead of simulating clicks like Meta’s Muse, Xiaowei talks directly to digital storefronts already running inside the chat window and performs tasks like ordering bubble tea or Meituan takeout, checking store hours, etc.
In addition, Tencent has signed agent-to-agent (A2A) deals with major phone OEMs so a voice assistant can hand a task to WeChat directly.
To secure the final transaction step, WeChat offers an “AI exclusive card” that isolates agent payments from the main account with balance limits and per-transaction confirmation.
The Commerce Loop Scorecard
| Transaction Stage | Alibaba | Tencent | Meta | Amazon | |
| Intent & Discovery | Owns | Owns | Owns | Owns | Owns |
| Storefront & Cart | Owns | Owns | Partners | Scrapes / Partners | Owns |
| Payment | Owns | Owns | Partners | Partners | Owns |
| Fulfilment | Owns | Partners | Partners | Partners | Owns |
Takeaway from China
The Chinese market shows that agentic shopping is easier to implement when the entire transaction pipeline is connected under one roof.
Alibaba achieves it because it owns everything from the product catalog and delivery network to the payment system. Tencent succeeded because WeChat created a standardized mini-program ecosystem a decade before modern AI models existed.
Google and Meta face a far tougher task. They have to connect a fragmented web of independent merchants, payment networks, and protective platforms like Amazon. The winner will not be the company with the best AI model, but the one that makes the shopping process completely seamless.
This article was originally published by Asia Pulse.
This article is a “periodical publication” for information only and is not investment advice or a solicitation to buy or sell securities. This article does not constitute a “personal recommendation” or “investment advice” under UK FCA regulations. Investing in equities involves significant risk. The author holds NO position in the securities mentioned. There is no warranty as to completeness or correctness. Please do your own due diligence or consult a licensed financial adviser. Please read the Full Disclaimer before acting on any information. Images created with the assistance of AI.