Problem
Agentic shopping only works if the catalog is legible to a model. Most Shopify product data is written for humans and Google: sparse attributes, messy titles, missing GTINs, and little structure an agent can trust when it compares SKUs and checks out.
Merchants who care about AI discovery need enrichment that does not break storefront copy they already like, plus a way to see when agents (not just browsers) hit product data.
Constraints
- Enrich fields for LLMs without silently overwriting merchant-approved storefront copy.
- Merchants must approve before anything ships to the live catalog.
- UPC / GTIN gaps must surface via the Catalog API so agent surfaces do not choke on incomplete identifiers.
- UI stays inside Admin (App Bridge + Polaris).
- Model output needs guardrails so generated fields stay reviewable.
What shipped
- A Shopify app that rewrites and enriches catalog fields into LLM-readable shapes while keeping human-facing copy intact until approval.
- UPC / GTIN compliance checks via the Catalog API.
- AI-traffic analytics so merchants can see agent hits on product data.
- Embedded admin UI with App Bridge + Polaris.
- Laravel backend; GraphQL Admin + Catalog API for reads/writes; Claude for structured enrichment passes.
For the broader problem framing, see agentic commerce.
Result
Catalogs that agents can shop against: structured fields, identifiers that validate, and a clear view of AI-driven discovery traffic. Built for real merchant Admin workflows, not a demo storefront. Outcomes here are structural readiness and reviewable enrichment, not invented GMV or traffic percentages.
Stack
Laravel, Shopify Catalog API, Shopify GraphQL Admin API, Anthropic Claude, App Bridge, Polaris.
Want agent-ready Shopify catalogs?
If you are shipping agentic commerce or AI enrichment on Shopify, see Shopify Plus Apps, the agentic commerce page, or start a Shopify package conversation.