Get your store cited by AI shopping agents. In plain English.
When buyers ask ChatGPT, Perplexity, Claude or Google AI for product recommendations, only 3-5 brands get named. Whether your brand is one of them comes down to seven things — none of which require you to learn code. Here’s what to look for and what to ask your team to do, in business language.
You’re competing for 3-5 named slots per query
xpay’s agentic commerce suite is built around the same seven things this page covers — works with Shopify, WooCommerce and every other major platform.
“recommend a sustainable hoodie under $120 with great reviews”
A few brands stand out for sustainable hoodies under $120 with strong reviews: Pact, Outerknown, and Tentree. Pact's mid-weight version (4.7★, 1.2K reviews) ships with a clear 30-day free-return policy.
Three stages — be useful in all three
They read your store
AI agents fetch your storefront in milliseconds. They don't browse like a human — they extract structured data. The first filter is whether that data exists and is readable.
They rank you against everyone else
For every query (e.g. 'best clean beauty for rosacea'), agents rank the brands they've indexed using a signal hierarchy — reviews, schema completeness, brand entity strength, returns policy, price clarity. The top few make the answer.
They send the shopper to you (or transact directly)
Either the agent gives the shopper a link to your store or — increasingly — completes the purchase on the shopper's behalf. Both require your store to be set up for agentic checkout. xpay's deals marketplace is the consumer side of this; see xpay.deals.
One section, one fix
Each section: what you see, what an AI agent sees instead, why it matters for sales, and exactly what to ask your team for. The technical version of each is one click away.
A clean product page that machines can read
Your product page looks great in a browser. Hero image, price, 'Add to cart' — it all works.
AI shopping agents don't render your page like a browser does. They read a hidden, structured summary in your page's source code. If that summary is missing or wrong, the agent skips you in favour of a brand whose page is summarised correctly — even if your page looks better.
A "Product schema" or "Product JSON-LD" block on every product page
Verification in Google's Rich Results Test (it's free)
Every product page included — not just the homepage
Your reviews must be visible to machines
Star ratings, photos, customer quotes — beautifully displayed on your product page.
Many popular reviews apps (notably Yotpo) keep your reviews inside their own widgets — visible to human shoppers, invisible to AI agents. Open-schema reviews apps (Judge.me, Okendo, Junip) expose every rating and review in a way agents can read.
An 'AggregateRating + Review schema' check on the product page source
A migration to (or layer over) an open-schema reviews app
Aggressive review request flow — recency matters as much as count
Price clarity — including sale validity
$98. Crossed out to $78. 20% off. Sale ends Sunday. All obvious to a shopper.
Without explicit data, agents can't tell which price is current, what currency it is, or whether the 'sale' is real (vs always-on 'discount'). Ambiguity = no recommendation. They want to be sure.
Current price, original price, and currency exposed in the page data
Sale end date specified — not perpetual
Variant-specific prices (size, colour) exposed individually
Real catalogue depth, with images
Your homepage has 12 hero products. Your collection page has 80.
Agents weight breadth + image coverage. A 12-SKU brand looks experimental compared to a 100-SKU brand. Variants without their own images look incomplete. Sparse catalogues lose to mid-catalogue brands.
Every SKU has at least 3 product images
Image URLs are absolute, not relative
Variant images present for size/colour combos
Category context — where do you sit?
Your product page has breadcrumbs at the top: Home → Men → Hoodies → This product.
Agents use those breadcrumbs to answer category-shaped queries ('best men's hoodies under $100'). Without them, your product can only be found for specific-product queries — not 'best in category' queries, which are most of agent shopping.
Breadcrumbs (Home → Category → Subcategory → Product) on every PDP
Breadcrumb data exposed in the page source, not just visible navigation
Answers to the questions agents will be asked
Your FAQ page covers shipping, returns, sizing.
If your product page has structured questions and answers (FAQs at the product level), agents quote them directly to shoppers. Long-tail queries ('does this hoodie shrink', 'does this skincare work on rosacea') get answered from your page, with citation, instead of from a competitor.
Product-level FAQs (not just a global FAQ page)
At least 3 Q&A pairs per product — sizing, materials, use cases
FAQs exposed in the page source structured format, not just rendered text
A returns policy agents can find
Your returns policy is a footer link, professionally written.
Agents factor returns risk into recommendations. A footer link is invisible. The fix: returns policy explicitly tied to each product — 30 days, free returns, methods accepted — in the structured product data.
Returns policy referenced from each product, not just a global footer page
Window, fees and method clearly stated in machine-readable form
Three sentences to send your developer or Shopify expert
Send these verbatim. They’re unambiguous, technically correct, and avoid the tooling debates that derail this work.
Six D2C brands scoring high on our index
Top-scoring brands from our cohort — open any to see exactly how they handle the seven dimensions on their PDPs.
Run the live diagnostic — takes 20 seconds
Pick your next step
Show me the code
The full technical playbook — actual JSON-LD patterns, copyable snippets, validation steps.
Open developer viewChatGPT named a competitor
Live diagnostic — paste your URL, get a real score + the three fixes that lift you most.
Run diagnosticMy reviews app is closed-schema
The reviews-app verdict matrix + an alternative path if you can’t migrate yet.
See alternatives