
Shopify lets compatible AI assistants find products and complete buyer-approved checkout. Merchants gain another route to an order, with the shopper still approving the payment.
News
1️⃣ Shopify gives AI assistants a route to buyer-approved checkout.
2️⃣ Walmart connects US shopping records with Spotify and selected TV ads.
3️⃣ Amex links AI payments to a shopper’s approved buying limits.
Insights
4️⃣ Retailer-owned brands face different share and growth patterns across Europe.
5️⃣ Walmart connects product creators with sales income and seller partnerships.
6️⃣ AI shopping scores combine consumer trust with how much people use AI.
LET’S ENTER THE MAZE!
1️⃣ News

The Maze: Shopify, which powers merchants’ online stores, has opened a connection for compatible AI assistants to find goods and complete buyer-approved checkout. Its October 13 launch leaves the shopper in charge.

Source: Shopify. The October 13 launch distinguishes Shopify checkout with buyer confirmation from separately approved outside-store card purchases.
Video: OpenAI. OpenAI’s September 2025 demo shows product discovery and buyer review of an Etsy checkout inside ChatGPT. This is a historical predecessor, not a Shop MCP demonstration.
The Details:
Search gains a payment route. The hosted connection combines 17 shopping tools. Previously, Shopify’s catalog search sent shoppers to a website to pay; the new route connects discovery to payment.
Saved details still need approval. Signing into Shop adds addresses and Shop Pay payment details. The buyer confirms the total and payment method before the assistant submits the order.
Outside stores use another path. The shopper approves in Shop, then the assistant receives single-use card details for the outside store’s checkout. Launch countries and conversion gains are undisclosed.
MarketMaze Take: Shopify can retain its role in checkout as assistants handle discovery. The new distribution channel still has to earn its sales.
Reach is not revenue. Shopify says its catalog has over one billion products. That measures available inventory, not active buyers or additional orders.
The portable lesson is separation. Merchants elsewhere can distinguish who introduces the shopper from who connects checkout. The release does not establish worldwide payment eligibility.
2️⃣ News

The Maze: Walmart, the US retailer, now lets brands use shopping records to guide Spotify and selected TV advertising. Its October 6 expansion links ad exposure with purchases at Walmart.

Source: Spotify Advertising. Spotify’s official format illustration shows in-feed video and display ads alongside in-stream video and audio ads; representative GIF frame, retrieved October 14.
The Details:
Shopping records guide Spotify ads. US advertisers use Walmart’s automated ad-buying platform to select shopper groups based on purchase history, then measure audio, video and display campaigns against Walmart purchases.
TV gets a separate route. TV group Warner Bros. Discovery supplies selected streaming and scheduled-channel ads through Walmart’s platform. Yahoo’s separate ad-buying route remains in early testing.
Measurement has limits. Walmart links ad exposure with purchases. One company-presented campaign reports lower ad-delivery costs and more cart additions, but neither proves additional sales.
MarketMaze Take: Walmart is extending the usefulness of its shopping data beyond its own pages. Brands still need to separate sales that follow advertising from sales the advertising caused.
This model already travels. Amazon launched Spotify ad buying with shopping signals in nine countries in 2025. Walmart’s US offer should be evaluated against the routes available in each market.
Audience access is not portability. Yahoo’s initial arrangement kept Walmart in control of its data. Fees, minimum spending and the testing phase’s end remain undisclosed.
3️⃣ News

The Maze: American Express, the card network, has published a framework that ties AI purchases to authenticated buying instructions. Developers get a design for payment permission; broad consumer deployment remains unproven.

Source: American Express. Illustrative approval screens show a shopper confirming buying limits and approving an agent purchase. Features are under development and subject to change.
The Details:
Customers set the rules first. A software connection records the preferred merchant, maximum budget and other instructions. The customer verifies their identity and approves those terms before payment.
Permission travels with the payment. A digital token represents approved limits and obtains checkout credentials without exposing the card number. It can expire or cover several purchases.
Instructions become dispute evidence. The card issuer can compare a later complaint with the approved request. Amex says agent-related disputes have not increased, but supplies no rate.
MarketMaze Take: Amex is trying to document what the shopper permitted before payment. That may improve dispute evidence; it does not certify that an agent chose the right product.
A spending ceiling is a blunt tool. The design includes item dependencies and purchase-count limits. Developers must specify those conditions rather than treating one budget as the complete instruction.
This is a portable design problem. Retailers elsewhere can separate authorized payment from accurate orders. Amex supplies a specification; deployment and reduced losses remain unproven.
4️⃣ Insight

The Maze: Private label takes 43.7% of European grocery and household-goods units in a 31-market snapshot, but high-share countries are still moving in different directions.
Spain and Portugal exceed 58% unit share and still gain, while Belgium and the Netherlands slip despite private label already taking most units.
Serbia gains 1.7 percentage points to 18.4%, showing why share level and growth speed need different commercial responses.
Brands need quality and distinction that justify a price gap, but unit-share changes alone cannot prove that shoppers stayed after inflation eased.
Why it matters: Manage by country and category: a majority share still growing needs a different brand response from a small base gaining fast, and neither proves why shoppers switched.
5️⃣ Insight

The Maze: Walmart gives creators a way to earn from product recommendations, sellers a way to find them, and a team to keep those partnerships working.
Walmart Creator began as a beta in 2022, and Napolitano’s post describes creators earning a share of product sales while marketplace sellers find partners in the same tool.
A dedicated team supports creators across product, social commerce and advertising, making this an ongoing shopping channel rather than a one-off promotion.
The wider network includes employees, alumni and YouTubers such as Bobby Parrish, whose Walmart shopping videos bring products to audiences beyond the retailer’s own channel.
Why it matters: The idea is simple: help people recommend products, reward the sales they generate, and make it easy for sellers to work with them.
6️⃣ Insight

The Maze: ACI means Agentic Commerce Index: Zinklar’s measure of AI’s influence on shopping, combining how much people use AI with how much they trust it in each category.
Read the scores as index values. In the US 2026 study, electronics scores 6.9 and travel 6.7 versus an average of 5.5, while soft drinks scores 4.2; higher values signal stronger combined AI influence, not percentages or sales gains.
Read the circles as category labels. Circle 10 means electronics and 22 means travel; moving right means more trust and moving up means more AI involvement, while colours describe types of purchase.
Understand the business consequence. Sanchez Neira’s point is that AI can shape a shopper’s shortlist, so brands must help assistants understand and compare their products as well as persuade people.
Why it matters: ACI helps choose where to test AI shopping first: can an assistant find your product and explain why it fits? The post does not disclose the full scale or formula, so 6.9 should not be labelled “out of ten.”
🧰 Resources
Useful stuff, handpicked by the team every day
🤖 AI prompts: 14 AI Prompts for Meta Ads: Analysis & Scaling or 👉 All here
📅 Events: ComConn New York 2026 or 👉 All here
🔓 Get the full resource library with MarketMaze+
🗞️ Quick hits
Everything else you should know
📣 Platforms & advertising
Meta reportedly restricts ByteDance-linked ads in seven markets, including the US, Japan and Indonesia. Third-party advertisers may also be affected when paid messages send users to TikTok or other ByteDance services.
Aditude reports 10.5% lower ad earnings per thousand ads among upgraded Safari users. Its current overall revenue-loss estimate is about 1%; other publishers report little impact.
Google's image-quality warning now lets otherwise eligible products keep showing in Shopping ads and free listings. Mandatory image rules still apply, including the 500 × 500 minimum from January 31, 2027.
🛍️ Retail & product pages
Vestiaire’s photo test reports a 5% relative lift in the rate of listings selling versus its old photo tool. Sellers’ steps stayed the same; the vendor has not disclosed sample counts or baseline rates.
easyShop opens in 21 European countries with one seller account and automatic tax calculation. Its no-own-stock promise removes one conflict; seller approval and unproven buyer demand still decide its value.
📊 Merchant scale & AI discovery
🛍️ Shopify's new report finds AI-referred visits converted 80% better than unpaid search in April–June. The observed gap does not prove AI caused more buying; referral volumes and costs still matter.
Semrush's UK study puts Tesco 25th and Asda 606th among ChatGPT's cited domains in July, despite both being top-mentioned brands. Being named and used as a source differ; these ranks do not measure sales.
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
Images: 2, 4, 6, 8, 9, 10 AI-generated · 3 MarketMaze screenshot · 5 Spotify via Spotify Advertising · 7 American Express via American Express Newsroom


