★ FROM OUR SPONSOR
58% of Shoppers Ask AI What to Buy This BFCM
58% of shoppers now ask AI tools what to buy instead of searching, and about 85% of what those tools say about a brand comes from other people: reviews, best-of lists, videos and creator posts. When a shopper asks ChatGPT for the best gift this Black Friday, the answer is built from what creators say about you.
That makes BFCM the week your creator program matters most, and the week most programs break: briefs land late, codes leak to coupon sites, and gifts never turn into posts.
UpPromote, the data-backed affiliate platform built for Shopify, pulled 13 recipes from 65+ brands including Click & Grow, cfab and Vetain, each with the real setup and results. Click & Grow grew BFCM affiliate revenue 272% in a year. Here's how.
Guide: Make Your Catalogue Agent-Ready with a Product Data Audit
Understanding Agent-Ready Product Data
You don’t have to believe shoppers will hand their buying over to AI agents to justify this audit. Where AI already recommends products, weak product data is costing operators today through returns, wrong listings and missed shortlist spots.
Four developments make this a data problem now:
AI-recommended purchases come back as “not as described.” In a Blue Yonder survey of 2,000 US shoppers, nearly a quarter had bought on an AI recommendation, and 46% of those returned an item in the past year. The top reasons were “different than described” (36%) and wrong size or fit (23%). One caveat: this is survey data, and Blue Yonder notes it does not prove AI recommendations cause higher returns.
Assistants name a handful of products, not a page of results. Workflow Labs found Rufus narrowed discovery from 50 results to about five named products. Missing that shortlist costs more than slipping a few ranks.
Amazon may already show your website’s products, with the wrong details. Shop Direct includes merchants by default unless they opt out, and some listings show incorrect details or link to unauthorised resellers.
Checkout inside Google AI Mode has real limits. On Shopify it is on by default for eligible stores, but subscriptions, bundles, customisable and B2B-only products are not supported.
The cost lands on you, not the AI. Rithum’s US and UK research found more than half of shoppers blame the retailer or brand when an AI recommendation gets product facts wrong.
The right question to ask isn’t “how do I optimise for agents?” It’s: if a machine read only my structured product data, would it describe and recommend my product correctly?
The Mental Model
Your product record is the source. One governed set of facts per product.
Feeds and listings are the pipes. Merchant Center, Shopify Catalog, Amazon attributes, Shop Direct feeds.
AI assistants are the readers. They repeat what the pipes give them, mistakes included.
Return reasons are the error log. They show where the record failed a real customer.
You stay the accountable operator. Nothing gets published or switched off without a human check.
What You’ll Need
Scope
Your top 20 best-selling products. Clear every Merchant Center disapproval and warning on those first, and only expand once the method works.
Exports (copies, in one working folder)
Google Merchant Center: product feed, diagnostics, and the AI performance insights report with its attribute completeness score
Shopify: product export including metafields, store policies, Agentic channel settings, and the returns by return reason report
Amazon: attribute data, title, bullets, description and A+ for each top ASIN; the FBA Customer Returns report (or Return Reports for FBM) for the last 90 days; Voice of the Customer NCX rate and top return reason per ASIN
Step 1: Map Which Pipe Feeds Which Assistant
Fixing your product page does nothing for a channel that only reads your feed. Three records matter most:
Google Merchant Center feed: feeds Google AI Mode and Gemini, and (in the same Google format) Amazon Shop Direct if you submit a feed
Amazon listing: feeds Alexa for Shopping (renamed from Rufus in the US; Rufus still runs in Germany)
Shopify Catalog: sits behind ChatGPT and Shopify’s other agentic channels
Audit those three and keep them saying the same thing.
Step 2: Check Whether You’re in Amazon Shop Direct
Applies to any brand with a public website selling to the US, including brands already on Amazon.
How to Run the Check
From a US storefront, search Amazon for your brand name and each top product. Note every result labelled Shop Direct.
Open each one and compare price, stock status, variant, core specs and images against your own product page.
Check where the link goes. Some listings have pointed to unauthorised resellers.
If you also sell on Amazon, compare prices across both. DTC sale pricing or clearance stock can appear next to your Amazon listing.
Log every mismatch: product, field, what Amazon shows, what’s true, source URL, date checked.
Decide: Own Feed or Opt Out
Stay in with your own feed if you can keep price and stock accurate within hours. Use the Shop Direct Merchant Workspace (launched 4 Nov 2026) to verify your site and connect a feed via Feedonomics, Salsify, CEDCommerce, VersaFeed or SP-API. The SP-API route uses the Google product data format, so your Merchant Center feed is reusable.
Opt out if you can’t guarantee stock accuracy, run a strict MAP policy, or see sessions that never buy. Email [email protected], name every domain you run, keep Amazon’s reply, and search for your brand again a few days later.
Staying in passively isn’t recommended. Amazon keeps working out your details from your public site.
Step 3: Audit the Attributes Agents Rely On
An assistant can’t infer what you didn’t write down. Shopify’s own example: a person can see a down jacket looks warm, but an AI needs the fill power, shell material and temperature rating stated in the feed.
Build the Audit Grid
One row per product, one column per attribute group, one cell per channel (Amazon, Shopify, Merchant Center). Score each cell Complete and consistent, Conflicting or Missing across:
Identity: brand, GTIN, variant grouping
Specs and dimensions
Compatibility
Sizing and fit
Materials
Price and availability
Shipping and returns policy
Identity comes first because OpenAI requires stable item, group and offer IDs, and Google warns that products without a GTIN may have limited visibility.
Channel-Specific Fixes
Merchant Center: run AI performance insights first, then add the six optional conversational attributes launched in May 2026 via a supplemental data source: question_and_answer (up to 30 pairs), related_product, document_link, item_group_title, variant_option and popularity_rank.
Amazon: fill every applicable attribute for each top ASIN, and answer frequent customer questions in the description. Amazon has been replacing customer Q&A with AI-generated answers on many pages, so those answers need to live in your bullets, description, attributes and A+.
Shopify: confirm Catalog eligibility, fill Terms of service, Privacy policy and Return and refund policy, and keep the Google & YouTube channel’s automatic sync on.
Where channels disagree, fix the record at the source and push it out. Don’t patch one channel.
Cross-Channel Attribute Audit Prompt
Run this on copies of your exports, never on live systems:
GOAL
Find every product where the attributes AI shopping assistants rely on are missing
or conflict across Amazon, Shopify and Google Merchant Center, and produce a fix sheet.
CONTEXT
We sell [PRODUCT CATEGORY] through [Shopify store / Amazon US / other] in [COUNTRIES].
Scope: our top [20] products by revenue, listed in the attached SKU list.
INPUTS & EVIDENCE
Use the attached Merchant Center feed export, Shopify product export and Amazon
listing/attribute export. Treat [SOURCE, e.g. our PIM / Shopify] as the source of truth.
Match products by GTIN first, then SKU. If a product cannot be matched, flag it.
CONSTRAINTS & AUTHORITY
Analyse copies only. Do not invent specs, certifications or compatibility.
If a value is missing everywhere, mark it Missing; do not fill it.
OUTPUT
1. A grid: one row per product; columns for identity (brand, GTIN, variant grouping),
specs and dimensions, compatibility, sizing, materials, price and availability,
shipping and returns policy. Each cell: Complete and consistent / Conflicting / Missing.
2. A fix sheet: product, attribute, channel, current value, correct value (from source of
truth), evidence.
3. Draft question_and_answer pairs for Merchant Center, only from facts in the inputs.
FINISH LINE & CHECKS
Done when every in-scope product is scored, every Conflicting cell shows both values,
and every drafted answer cites the input it came from.For the rest of the guide, including the returns-to-data loop, the Google AI Mode checkout review, the shopper-prompt and returns analysis prompts, common mistakes and a 30-day rollout plan, here’s the full SOP. My gift to you :)
WANT TO REACH 70,000+ ECOMMERCE READERS?
Put your brand in front of an engaged audience of 70,000+ subscribers, with 75% based in the US.
Interested in sponsoring our newsletter? Get our media kit & rates.
The Quick Read:
OpenAI launches Images 2.6, the newest version of its image model.
DoorDash has received only several hundred orders from AI agents in total, against roughly 3 million deliveries a day. Why it keeps building for agents anyway is the telling part.
AI traffic now brings US retailers 83% more revenue per visit and converts 90% better than other traffic. Adobe expects that gap to double this holiday season, yet over a third of retail homepage content is unreadable to AI.
TikTok launches a standalone TikTok Shop app in four Southeast Asian markets. The bet: people don’t want to shop inside a video app, and Douyin’s near $500B in GMV shows what’s at stake.
AI shopping agents could threaten Amazon and Walmart’s ad business.
US organic traffic falls 13% in two years, yet email now adds more visits than paid search. New data shows which channels actually pick up the slack, and why AI isn’t the only culprit.
The Tools List v1:
📚 SciSpace - Do hours worth of reading in minutes. Data extraction, AI Copilot, and paraphraser in one tool.
🔎 Exa Categories - Search filters that get only the type of content you want.
🤖 Envato - Create better AI images with beautiful, free and endlessly reusable style prompts for Midjourney, Clipdrop, Firefly and DALL-E
📝 Fanfuel - A tool to create YouTube content, thumbnails, scripts and analyze audience.
ABOUT THE WRITER:
Meet Joanna (Jo) Lambadjieva
Jo Lambadjieva is an entrepreneur and AI expert in the e-commerce industry. She is the founder and CEO of Amazing Wave, an agency specializing in AI-driven solutions for e-commerce businesses. With over 13 years of experience in digital marketing, agency work, and e-commerce, Joanna has established herself as a thought leader in integrating AI technologies for business growth.
For Team and Agency AI training book an intro call here.
