Most Shopify agencies and merchants assume they're ready for agentic commerce. They're on the platform. The theme generates JSON-LD. Products are appearing in AI results. Their brand monitoring tools show citations and mentions. The conclusion is that structured data is handling itself and the channel is working.
Usually it isn't. The products are getting recommended via brand mentions on third-party sites, editorial coverage, manufacturer spec pages, or data baked into the model during training. The citations the monitoring tools are picking up are coming from those sources — not from the merchant page or the catalogue. And critically: those signals can't be managed. You can't update them when a product changes, correct them when the price is wrong, or retire them when a line is discontinued. You're dependent on a layer of data you don't own and can't touch.
That's not a reason to relax. It's a reason to understand what's actually at stake.
Conversational AI vs agentic commerce — why the distinction matters
These are not the same thing, and most current thinking about AI visibility conflates them.
Conversational AI — ChatGPT, Perplexity, Gemini in chat mode — answers questions. It draws on training data, web crawls, brand mentions, and editorial coverage. It recommends. The human then decides what to do next. The merchant page is optional in that journey, which is why a brand with strong third-party coverage can get recommended despite weak structured data.
Agentic commerce is transactional. An AI agent receives a goal — "buy me a dairy-free protein powder under £40, at least 20g of protein per serving" — and executes it. It compares options, selects a product, and completes checkout without the human visiting a single product page. To do that, it reads structured data directly to evaluate candidates against the buyer's criteria. If the attributes aren't machine-readable, the product isn't evaluated. There's no human in the loop to compensate for the gap.
The latitude that exists in conversational AI — where brand signals and training data can carry a weak catalogue — does not exist in agentic commerce.
Shopify has already shipped the infrastructure
Shopify has been building for this directly. The Universal Commerce Protocol is an open standard co-developed with Google covering the full commerce journey from discovery through checkout. Agentic Storefronts went live for all US Shopify merchants on March 24, 2026, connecting eligible catalogues to ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini. From January 2025 to January 2026, orders originating from AI searches increased 15x.
The infrastructure is live. The eligibility gate is product data quality — not a manual opt-in. The quality and completeness of that data determines whether a product surfaces in a conversation or gets passed over.
Agentic Storefronts being activated is not the same as products being recommended. That distinction is everything.
The layer you can actually control
AI agents don't browse a store the way humans do. They rely on structured data. Product information like title, price, material, and dimensions must be organised in standard, machine-readable fields rather than embedded in marketing copy or page layouts.
The additionalProperty layer — machine-readable attributes that sit inside the Product schema block — is the one AI-facing signal that belongs entirely to the merchant. Material composition. Dimensions. Dietary flags. Size system. Form factor. The exact signals an agent evaluates when someone submits a filter-style query or delegates a purchase.
Here's the difference between what passes schema validation and what an agent actually needs:
// Passes validation. Agent cannot match to a filter query.
{
"@type": "Product",
"name": "Merino Crew Neck Jumper",
"description": "A fine-knit merino wool jumper in natural colours.",
"offers": {
"@type": "Offer",
"price": "95.00",
"priceCurrency": "GBP",
"availability": "InStock"
}
}
// Agent can evaluate against "wool jumper, machine washable, under £100"
{
"@type": "Product",
"name": "Merino Crew Neck Jumper",
"additionalProperty": [
{ "@type": "PropertyValue", "name": "material", "value": "100% merino wool" },
{ "@type": "PropertyValue", "name": "care", "value": "machine washable" },
{ "@type": "PropertyValue", "name": "fit", "value": "regular fit" },
{ "@type": "PropertyValue", "name": "weight", "value": "lightweight" }
]
}The first passes every schema checker. The second is what agentic commerce actually reads.
The same gap applies across every category. "Dairy-free protein powder, 20g protein, under £40." "Linen duvet cover, 200 thread count, king size." "Reef-safe SPF50 sunscreen, fragrance-free." Each of those queries requires structured attributes to evaluate. Prose descriptions don't reliably substitute.
What four cohorts of real scans show
We've run Product Rank across four categories of Shopify brands — beauty, homeware, food and supplements, and fashion. Across all four, the structural foundation was generally solid. JSON-LD present, price and availability readable, brand and identifier signals in place.
The attribute layer failed universally.
- Beauty (9 brands, 1,588 products with schema):
additionalPropertyfailed on every store, including the 82-scoring Q+A Skin - Homeware (5 brands, 1,827 products): same finding from Boll & Branch at 84 down to Snowe and Year & Day at 40
- Food & supplements (8 brands, 422 products): Trip at 81 through Cymbiotika at 40 — attribute coverage absent throughout
- Ramraj Cotton (3,440 products):
material_missing,core_spec_signals, andadditionalPropertyfailing on 99.7% of the catalogue — despite valid JSON-LD and brand identifiers on nearly every product
Valid schema. Missing attributes. Every cohort, every score band.
Shopify's default theme generates a Product JSON-LD block. It passes validation. But it misses critical fields: GTIN/EAN identifiers, aggregate ratings, return policies, and shipping details. The schema is present. The eligibility signals are not.
One additional factor: Amazon has broadly locked its catalogue out of the open AI protocols — protecting its advertising business by keeping shoppers on Amazon.com rather than buying through ChatGPT or Copilot. That means the ChatGPT, Perplexity, and Google AI Mode surfaces are filled primarily by merchants who have opted in. Brands with clean Shopify catalogue data are competing in a channel Amazon cannot reach. The opportunity is real — but only for catalogues that are actually eligible.
What this means for agencies
The catalogue going into Shopify's agentic infrastructure is your clients' product data. If the attribute layer is empty, the infrastructure routes around it — or surfaces a competitor who has it.
The audit deliverable that holds up to client scrutiny isn't "your schema is valid." Every Shopify theme ships with that. It's: here are the products missing the attribute signals that agentic commerce reads for filter queries and purchase execution, here's where a competitor with better attribute data is positioned to take the sale, and here's what needs to change before the channel penalises you for it.
What Product Rank measures
Product Rank scans Shopify catalogues at scale and scores every product against a deterministic rule set covering structural schema, attribute completeness, offer integrity, and content quality. The same catalogue scanned twice gets the same score.
The goal is not to predict whether a product will be cited in a conversational response — that depends on signals inside the model that aren't observable from outside. The goal is to surface the gaps in the signals you own: the ones you can fix, verify, and show a client.
If you manage Shopify stores and want to see where your clients' catalogues stand, you can run a free scan at productrank.io.