How AI shoppers find D2C brands
ChatGPT Shopping, Perplexity, Google AI Mode, Claude, and the custom shopping agents underneath them — five surfaces with different mechanics, one underlying signal hierarchy. This is the mental model, the per-surface playbook, and the 5K-brand cohort underneath it. Built around xpay’s agentic commerce suite for Shopify, WooCommerce and every other major platform.
0
LLM citations observed last 30 daysFour surfaces, one underlying mechanism
AI shopping discovery happens across four surfaces with distinct mechanics. Switch tabs to see how each one finds and ranks brands, what it needs from your storefront, and where to dig deeper.
OpenAI's shopping mode surfaces product recommendations inline with ChatGPT answers. Eligibility is partly opt-in (feeds), partly earned (schema cleanness + brand signal density).
What it needs from your storeProduct + Offer JSON-LD on every PDP
AggregateRating + Review schema (not behind JS)
Returns policy explicit and discoverable
Feed surfaced to OpenAI when invited or via partners
“best D2C skincare brands for sensitive skin”
For sensitive skin, the brands that come up most consistently across product comparisons are Tower 28, ILIA Beauty, Drunk Elephant, Beautycounter, and Versed. Each has transparent ingredient lists and accessible review aggregations.
Crawl → Extract → Rank → Cite → Checkout
Crawl
Agents and their indexers fetch your storefront. Static HTML matters here — JS-injected content is less reliable.
Identify agent traffic →Extract
JSON-LD parsing pulls Product, Offer, Review, AggregateRating, Breadcrumb, FAQ. Missing schema = missing brand in the answer.
See the schema patterns →Rank
Brands ranked against the query — review density, recency, brand entity strength, price clarity, returns risk.
See the leaderboard →Cite
Top-ranked brands named in the answer. Most queries cite 3–5 brands. Number-7 is invisible.
Reviews apps that get cited →Checkout
The cited brands convert when the agent can complete checkout via UCP / ACP / MCP. Without it, you stay in the consideration set.
Protocols overview →What agents weight when picking which brand to name
Weights are estimates from our 5K-brand cohort observations — actuals vary by engine and query, but the rank order is stable.
AggregateRating + Review schema
Open-schema reviews app (Judge.me / Okendo / Junip) emitting per-product JSON-LD.
Product + Offer JSON-LD on PDP
Theme-level schema. Verify in Google Rich Results test on every PDP template.
Brand entity consistency
Consistent Brand schema + matching OG / Twitter tags. Avoid name drift across listings.
Catalogue breadth + image coverage
Image arrays with absolute URLs. Variants surfaced with offers[] or AggregateOffer.
Pricing clarity (currency, sale validity)
priceCurrency required. priceValidUntil distinguishes real sales from always-on promos.
Returns policy in Offer
hasMerchantReturnPolicy with category + window + return method.
BreadcrumbList + category mapping
Three-level breadcrumbs from home → category → subcategory.
Agent checkout surface (UCP/ACP/MCP)
A real endpoint, not just a button. Agents need to complete a transaction, not just see one.
Brands that score well — and why they show up in answers
5,575 D2C brands scored across the seven dimensions. Top of the board, today:
Terms you’ll see across this cluster
Agentic commerce
Buying flows where an AI agent (acting for a human) discovers, compares, and transacts with a merchant — replacing or augmenting the browser-shopper.
Agent-readiness
How well your storefront supports AI shopping agents — measured across catalogue, schema, reviews, pricing clarity, checkout endpoints, inventory and returns visibility.
AI Overview citation
A named mention of your brand inside an AI-generated answer (Google AI Mode, ChatGPT, Perplexity, Claude). The 2026 equivalent of a top-3 organic ranking — except only 3–5 brands get named per query.
JSON-LD
Linked-data JSON format used to describe entities (Product, Offer, Review, FAQ) for machines. The lingua franca of agent-readable storefronts.
ACP
Agentic Commerce Protocol — Stripe-led standard for letting agents complete checkout on the merchant's behalf. Implementations vary by vendor.
MCP feed
Model Context Protocol feed — Anthropic-originated framing for exposing structured commerce data to agents. Increasingly adopted as a vendor-neutral product feed shape.
UCP
Universal Commerce Protocol — the umbrella xpay uses for agent-compatible checkout endpoints across ACP, MCP and others.
Brand entity
How agents identify your brand across the web — the unique signature formed by Brand schema, social profiles, OG metadata, and consistent naming.
Pick the path that matches your moment
ChatGPT named a competitor
Run the live diagnostic. Get a real score + the three fixes that lift you most.
Run diagnosticMy PDPs need fixing
Schema-by-schema teardowns of agent-readable product pages. Copy what works.
See PDP patternsMy reviews app is closed-schema
The reviews-app verdict matrix + an alternative if you can't migrate yet.
See alternatives