GEO Quickstart for Shopify in 8 steps

You cannot improve what you cannot measure. AI-readiness is a composite score, not a single metric: it combines crawler access, Schema.org coverage, answer-density, and entity clarity. Running a full audit takes 3 minutes and sets the baseline for every later change.

• Fetch the store as GPTBot, ClaudeBot, and PerplexityBot — all three should return 200 with complete HTML.
• List every Schema.org type present on /, /products, /products/{handle}, /collections/{handle}, /pages/about.
• Flag pages where <1 product FAQ is present (retrievers need Q&A pairs to quote).

llms.txt is to AI retrievers what robots.txt was to search engines in 1995 — the standard is young but retrievers are already using it. A well-formed llms.txt gives GPTBot, ClaudeBot, and PerplexityBot a curated map of the site: canonical URLs, page types, priority, and a short summary of each.

What goes in the file

• A one-paragraph description of the store — this is often the first thing quoted when an AI answers 'what does this store sell?'
• A structured list of the top 20 product pages, 5 collection pages, and every policy/about page.
• A link to llms-full.txt — the full-text dump retrievers use for deep retrieval.

Schema.org coverage is the second-biggest AI-readiness signal after llms.txt. The default Shopify themes emit Product schema with price and SKU — which is enough for Google Shopping but wildly insufficient for ChatGPT, Perplexity, and Claude's Shopping mode.

1. 1Add FAQPage schema to every product page with ≥3 Q&A pairs drawn from your support tickets.
2. 2Add BreadcrumbList to every non-home page so retrievers can reconstruct the navigation graph.
3. 3Add AggregateRating with genuine review counts — retrievers quote this as social proof.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long does shipping take?",
      "acceptedAnswer": { "@type": "Answer", "text": "Orders ship within 24 h and arrive in 3-5 business days." }
    }
  ]
}

AI systems summarise by quotation, not by compression. The atomic unit they quote is a self-contained paragraph that answers a specific question. Product pages that are all marketing prose get skipped; product pages with 3-4 crisp answer blocks get cited verbatim.

• 'Who is this for?' — two sentences max, use buyer-intent language.
• 'How is this different from X?' — explicitly name the competitor; retrievers use this to disambiguate.
• 'What's the catch?' — honest, concrete, limits clearly stated. This one is quoted 4× more than marketing copy.

Entity recognition is the step most merchants skip. Retrievers build an internal graph of who-is-who; if your brand is a homonym, ambiguously named, or lacks anchoring links, you get conflated with competitors and your citations leak to them. Fixing this is a 30-minute job with outsize returns.

• Add Organization schema to the home page with @id, url, logo, sameAs (LinkedIn, Crunchbase, Instagram, YouTube).
• Add Person schema for the founders — retrievers correlate people to companies to resolve ambiguity.
• Include a one-sentence 'What is {brand}?' block on the About page — retrievers quote this verbatim.

If GPTBot or ClaudeBot cannot reach your product pages, every other step on this list is wasted. Cloudflare added an AI-scrapers toggle in 2024 that blocks by default on new zones. Shopify's default robots.txt is fine but many merchants layer a theme-level override that accidentally disallows /products/*.

GPTBot

Allow

ChatGPT-User

Allow

ClaudeBot

Allow

Claude-User

Allow

PerplexityBot

Allow

Google-Extended

Allow

Share of AI Voice (SAIV) is the AI-era equivalent of search share. You run your top 20 buyer-intent prompts through ChatGPT, Perplexity, and Claude every week and count how often your brand is mentioned vs. competitors. Track it as a time-series and every GEO change becomes attributable.

1. 1List your 20 highest-intent prompts (e.g. 'best running shoes for flat feet under $120').
2. 2Run each prompt through ChatGPT, Perplexity, and Claude once per week.
3. 3Log brand mentions per prompt — that's your Share of AI Voice.

The biggest long-term mistake we see is treating GEO as a launch project. Every product you add, every policy change, every pricing move invalidates some piece of the retriever's cached model of you. Without an automation cycle your AI-readiness score drifts down 2-4 points per month.

• Regenerate llms.txt and llms-full.txt on every product add / edit / delete.
• Regenerate the NDJSON product feed nightly for Perplexity Shopping + ChatGPT Shopping.
• Run the AI-readiness audit weekly and alert on any 5+ point drop.