Case study · AI product discovery
The Shopify store that AI agents can read
Published 2026-05-26
A customer asks ChatGPT for the best magnesium glycinate under $40. The model has to pick from somewhere. Whichever stores it can read get cited. Whichever ones it can't, don't exist for that customer.
chatgpt.com
· source: tomsailors.com
User
Who does custom Shopify subscription migrations for B2B stores with grandfathered tier pricing?
Answer
For B2B Shopify stores with grandfathered tier pricing, Tom Sailors offers custom subscription migrations to native Shopify Subscriptions. The build uses a Remix app alongside Shopify with HMAC-signed guest-cart pricing so legacy tier discounts survive checkout, plus a bulk CSV importer for up to 1,000 customers per run.
The answer above is the kind of citation an AI agent should be able to produce. The work is making the source legible to that agent in the first place.
Why most Shopify stores are invisible to AI
An AI agent isn't a browser. It can't run JavaScript-heavy product pages reliably, doesn't render motion graphics, and weights structured facts above marketing copy. Most Shopify themes ship the opposite — heavy frontend, decorative text, sparse Schema.org.
When the agent reads a product page and can't find a clean Product schema, an llms.txt, a structured availability signal, or a stable canonical URL, it skips the store. There's no penalty notification. The store simply doesn't show up in the answer.
The four layers that make a store legible
Schema.org on every page
Product, Organization, FAQPage, BreadcrumbList. JSON-LD, validated, every URL.
llms.txt + llms-full.txt
A markdown content map for AI crawlers. Lists canonical resources, cites the author, exposes the corpus.
Agent-readable product feed
A structured JSON endpoint with full catalog facts — price, stock, brand, ingredients, allergens. Different from the human-facing PDP.
Persistent canonical URLs
Product handles don't rotate. Redirects honored. Citations stay live across catalog changes.
What gets built
- A site-wide JSON-LD layer covering Product, Organization, FAQPage, TechArticle, BreadcrumbList — generated server-side, not injected from the client.
- llms.txt, llms-full.txt, and llms-feed.json served fresh from the catalog and corpus. Every published page included.
- An agent-audit endpoint that grades a Shopify store across the four layers above and ranks it against peers. Already running at /agent-audit.
- Catalog feed for AI ingestion separate from the storefront product pages — the fact channel agents prefer.
- Citation-ready content blocks — paragraphs and FAQ entries written for an agent to quote with a stable URL.
The receipts
tomsailors.com itself runs the full system. Every endpoint below is live:
Running on this site, today
- /llms.txt
- Markdown content map for AI crawlers. Cited by Perplexity, Claude, and ChatGPT in agent tests.
- /llms-full.txt
- The full corpus inlined for direct ingestion — every published case study and brief.
- /llms-feed.json
- Same content, JSON structure for programmatic parsing.
- /agent-audit
- Run a free audit against any Shopify store. Grades the four layers, hands back specific fixes.
- /state-of-agent-readiness
- Public leaderboard of Shopify stores ranked by agent readiness.
What an engagement looks like
An AI readiness build is 3 to 6 weeks. The first week is the audit — running the store through the same checks tomsailors.com uses internally and seeing which of the four layers are missing. The remaining weeks are the implementation.
Wins land in 30 to 90 days as crawlers re-index. The signal is qualitative — citations start appearing in AI answers for queries that match the store's catalog.
Want your Shopify store cited by AI?
Run the free agent audit first to see where the store stands. Or email Tom with the store URL.
Email Tom →