
Amazon blocked Meta’s Muse shopping agent from Amazon.com after its US launch, drawing a hard line between helping a shopper choose and letting an outside agent transact. Agentic commerce will move only as fast as retailers permit it through the checkout.
News
1️⃣ Amazon locks the agentic checkout
2️⃣ Shopify resets the conversion baseline
3️⃣ Temu finds Europe’s marketplace gaps
Insights
4️⃣ Social wins Gen Z’s decision
5️⃣ Social and LLMs shape choice
6️⃣ Discovery becomes a data problem
LET’S ENTER THE MAZE!
1️⃣ News

The Maze: Amazon blocked Meta’s Muse shopping agent on Amazon.com after its US launch, exposing a gap between a shopper approving a purchase and a retailer accepting the agent.

Source: Amazon via GeekWire. Amazon’s notice warns that access by an unauthorized AI agent violates its Conditions of Use.
Amazon’s September 20 block followed objections to unauthorized access and the handling of customer accounts.
Meta’s safety design keeps credentials separate from Muse and asks users to approve purchases, but cannot supply retailer consent.
Sellers cannot assume an outside assistant can complete an Amazon order simply because the customer wants it to.
Why it matters: For sellers, a listing is not a guaranteed AI sales channel: an assistant must be allowed to reach checkout, not just recommend the product to a shopper.
2️⃣ News

The Maze: Shopify is changing how store visits are counted from September 21–23, so traffic and conversion rates may shift even when orders, sales and customers do not.
Shopify now filters bots from session reports by default, while total orders, sales and customer counts remain unchanged by the update.
Revised session counting keeps visits together across midnight and includes some direct-to-checkout journeys, so the count can rise or fall.
Treat September 21 as a measurement break and compare results against a new baseline before judging ads or store changes.
Why it matters: A higher conversion rate can come from counting fewer visits, not selling more: check sales and reset the baseline before giving a campaign the credit.
3️⃣ News

The Maze: Temu ranked first in four European markets in ECDB’s 2025 ecommerce data. Each lacks a local Amazon storefront or a dominant domestic marketplace.

Source: ECDB. The map shows Temu’s position in European ecommerce retailer rankings for 2025.
Croatia, Finland, Hungary and Slovenia take the first-place list, while Temu ranks second in eight more countries, including France.
Low prices and broad choice meet less entrenched competition in the four leaders, but no local Amazon storefront does not mean Amazon cannot deliver there.
These 2025 positions predate later tariff effects, so they are a baseline for testing whether Temu can defend its lead as costs change.
Why it matters: Temu’s threat depends on local competition. Retailers need a clear reason to win the basket beyond price; the 2025 ranking cannot prove what happens next.
4️⃣ Insight

The Maze: Stores still introduce Gen Z to brands, but social media becomes its leading evaluation channel when a shopper weighs the purchase and looks for proof.
Gen Z moves from 23% social influence at discovery to 34% at evaluation, while in-store falls from 28% to 21%.
Boomers reverse the pattern: stores lead discovery at 52% and evaluation at 31%, while social remains secondary at 7% and 16%.
Brands need different proof by stage: store presence can earn attention, but social validation often does more work when younger shoppers decide.
Why it matters: A blended channel budget misses the handoff. Measure discovery and evaluation separately, then fund the channel that helps each generation move to a decision.
5️⃣ Insight

The Maze: In BCG’s November 2025 survey, social media influences 90% of reported decisions among users; LLM search reaches 85%, putting both near the final choice.
Social edges LLM search in every displayed category, from beauty care (93% versus 87%) to automotive (86% versus 85%).
The smallest gaps are household appliances and travel, where the paired figures differ by only one point.
These are conditional influence measures, not population usage rates or proof that either channel independently caused a sale.
Why it matters: Brands need emotional proof for social and accurate product truth for LLM comparison. One broken detail can lose a recommendation before checkout.
6️⃣ Insight

The Maze: BCG’s Global Consumer Radar shows digital sources average 17% of journeys, versus 8% for traditional media; at brand choice, the gap is 51% versus 38%.
Internet research (22%), social media (20%) and brand or retailer websites (20%) are the leading discovery routes in the November 2025 survey.
Social media (56%), LLM search (55%) and first-party sites (51%) lead reported influence on the brand ultimately chosen.
The implication is not to abandon mass media; it is to treat product truth, retailer data and answers as marketing infrastructure.
Why it matters: Research-led buyers compare across interfaces. Brands win consideration by being clear, current and credible wherever shoppers’ questions arise.
🗞️ Quick hits
Everything else you should know
🌍 Retail expansion needs local leverage
JD Sports signed a franchise partnership with Grupo Axo to launch stores and ecommerce in Mexico from 2027, using a local operator to accelerate a rollout planned to exceed 140 premises.
Wegmans committed $110 million to refrigerated distribution and returns facilities in the US, trading upfront capital for more control over availability and third-party logistics costs.
📏 Platforms redraw measurement and liability
Google Ads added multi-source conversion measurement so offline uploads can fill gaps left by website tags, improving coverage while raising the cost of weak matching and duplicate governance.
A US appeals ruling let Shein revive a copyright claim over product photos posted by Temu sellers, keeping marketplace content controls—not only first-party publishing—inside the liability debate.
🤖 Retail AI moves into operations and hardware
Thri5 raised $5.4 million USD to expand its AI platform for retail planning and operating decisions across North America and Europe, pushing automation closer to inventory and execution.
Google introduced Googlebook, a US laptop built around integrated Gemini features and on-device assistance, turning hardware distribution into another route for model usage.
⚖️ Frontier AI gets a standards proposal
OpenAI proposed international frontier-AI standards for evaluation, reporting and governance, seeking less fragmented rules while potentially raising the compliance bar for smaller model builders.
THAT’S IT FOR TODAY!
You’re the reason our team spends hundreds of hours every week researching and writing this email.
What do you think of this issue?
See you next time in the maze!
MarketMaze team



Social Wins Gen Z’s Decision 📱