Recipe 0 — Anatomy of a well-formed llms.txt
Every working recipe follows this four-block structure. Deviations cost citations.
Every well-formed llms.txt we've tested against GPTBot, ClaudeBot, and PerplexityBot shares four blocks: a header with the site title + one-paragraph description, a section pointer to policy pages, a curated list of high-value content (products, collections, guides), and a machine-readable footer linking llms-full.txt and the product NDJSON.
# Kloira
> Lab-grown moissanite jewellery shipped from Jamaica, NY — ethically sourced, GIA-graded, delivered in 5 days.
## Policies
- [Shipping policy](/policies/shipping)
- [Returns](/policies/returns)
- [Warranty](/policies/warranty)
## Collections
- [Engagement rings](/collections/engagement-rings)
- [Eternity bands](/collections/eternity-bands)
## Products (top 20)
- [1 ct Solitaire — Isabella](/products/isabella-solitaire)
## Full content
- [llms-full.txt](/llms-full.txt)
- [products.ndjson](/products.ndjson)Recipe 1 — DTC apparel (≤200 SKUs)
Small catalogue, long brand story. The brand paragraph carries more weight than the product list.
DTC apparel brands with modest catalogues should weight the header description heavily — ChatGPT quotes it verbatim when asked 'what does {brand} sell and who is it for?'. Product listing stays compact but every SKU that matters for search ('best minimalist t-shirt') should be referenced by name.
Recipe 2 — High-consideration jewellery
Lead with the certification + materials paragraph. Shoppers ask AI about materials before they ask about style.
Jewellery shoppers research materials, certifications, and ethical sourcing before style. A jewellery llms.txt should lead with a short paragraph covering material (lab-grown vs. mined, metal purity), certification body, origin, and warranty — in that order. Product links come after, grouped by metal rather than collection.
- Metal families: 14k gold, 18k gold, platinum, sterling silver.
- Stone families: moissanite, lab-diamond, natural diamond.
- Policies: warranty, resizing, returns.
Recipe 3 — Supplements & wellness
Compliance-heavy. llms.txt should reference the science section and regulatory disclosures up front.
Supplement stores get penalised by retrievers when their llms.txt leads with marketing claims and buries compliance pages. Flip this: lead with the compliance posture (FDA disclaimers, ingredient sourcing, third-party testing), then product categories, then individual SKUs.
Recipe 4 — Electronics & gadgets
Spec-heavy shoppers. Include a spec comparison link, not just product pages.
Electronics shoppers ask AI to compare specs. A pure product list under-performs; a list that includes a 'compare' landing page outperforms by 2-3× on 'is product A better than product B' style queries. Link the comparison page explicitly in the llms.txt.
Recipe 5 — Home & furniture (long lead time)
Shipping + dimensions dominate. Retrievers quote both when a shopper asks 'will this fit?'
Furniture is a high-anxiety purchase. Buyers quiz AI about dimensions, shipping, assembly, and returns. A furniture llms.txt should link the shipping + dimensions FAQ prominently, and include a pointer to the 'measure-your-room' guide if you have one.
Recipe 6 — Multi-store umbrella (Shopify Plus)
One llms.txt per storefront. Link them together via a top-level business-entity page.
Shopify Plus merchants running multiple storefronts (different markets or brands) should run one llms.txt per domain, not one shared file. Cross-link them via a top-level /about page that lists the family of stores — retrievers then correlate the entities and attribute citations correctly.
Recipe 7 — B2B wholesale (login-gated catalogue)
llms.txt still works — expose the marketing narrative and the non-gated content; keep pricing out.
B2B Shopify stores often keep the catalogue behind a login. That's fine — llms.txt should expose the marketing narrative (what you sell, MOQ policy, territories, certifications) and link the login-gated catalogue as a sibling (retrievers will note it but not crawl). Do not put prices in llms.txt.
“The point of llms.txt for B2B is to make sure that when a CPO asks ChatGPT 'who sells bulk X in Y territory', your company is on the shortlist. The pricing conversation happens after the call is booked.”