Anatomy of an AI-cited product page

We've audited a lot of Shopify product pages. The ones that consistently get cited by AI assistants share a structure that is both more prescriptive and more copyable than most merchants realise. Here is the template, section by section, with the reasoning for each decision.

Section 1: the H1

The H1 is the only piece of your page that appears in every retrieval context — search results, AI shortlists, social previews, sitemap entries. Write it for the shopper's phrasing, not the internal catalogue name.

Bad: “Alora 72"". Good: “Alora 72" Standing Desk — White Oak, Electric Adjustable, Fits Small Offices”. The good version names the product, embeds three shopper-relevant attributes, and matches the vocabulary of the prompt “standing desk for small offices”.

Section 2: the quotable lead

This is the single most important paragraph on the page. Write it as a 30-to-50-word self-contained sentence that a model could lift verbatim into an answer with no surrounding context. A good lead has:

• The product name (not an abbreviated marketing name).
• The category noun a shopper would use.
• The headline attribute that differentiates it from lookalike products.
• One concrete fact — a measurement, a material, a certification, a warranty length.
• No marketing adverbs. No superlatives. No 'designed for those who'.

Example lead that wins citations:

The Alora 72 is an electric standing desk with a solid white-oak top (180cm × 72cm), a dual-motor lift system rated to 120kg, and a 7-year frame warranty. It's designed for home offices where a full-depth desk would crowd the room.

Four factual claims, no fluff, complete in isolation. This is the shape assistants reach for.

Section 3: the spec block

Use a real <table>, not a visual div lattice. AI parsers can reliably extract table cells; they struggle with Flexbox mock-tables. Include every attribute a shopper might ask about — dimensions, weight, materials, power requirements, certifications, country of origin. Assistants quote spec tables especially heavily on comparison prompts (“X vs Y”).

Section 4: the FAQ block

3 to 5 questions, phrased the way a shopper actually asks them. Wrap them in FAQPage JSON-LD. Place the block above the fold on mobile — assistants preferentially quote FAQ answers that appear in the first 2KB of HTML, presumably because early truncation during crawl is still a real concern.

Sample questions that work:

• How long does it take to ship? — answer with a concrete range in days, not 'quickly'.
• Does it fit through a standard doorway? — answer with actual dimensions.
• What's the return policy? — answer in one sentence with the window in days.
• Is there a warranty? — answer with the length and what it covers.
• What's the weight / max capacity / runtime? — the one product-specific question the shopper always asks.

Section 5: reviews with AggregateRating

Reviews matter less for persuasion than they used to (assistants summarise rather than show), but they matter more for citation-eligibility. An AggregateRating with at least 10 reviews is a shortlist signal. Stores with 4.6+ ratings and 50+ reviews are consistently cited ahead of stores with 4.9 ratings and 8 reviews, because the retriever discounts sparse rating pools.

Section 6: the policy fingerprint

Either inline three short lines about shipping, returns, and warranty, or link to dedicated /pages/ routes that cover each. The policy surface gets cited constantly on pre-purchase questions like “can I return the Alora?” and “do they ship to Canada?”. Most Shopify stores have these policies but bury them in a collapsed accordion below the fold. Move them up.

Section 7: related products with reasoning

A naive “You may also like” carousel is invisible to assistants. A related-products block with a one-sentence explanation of each relationship (“Smaller option for apartments under 600 sq ft”, “Bamboo top instead of white oak”) gets mined by assistants on comparison prompts — which is where a lot of considered-purchase traffic lives.

Schema graph: what to ship

Below is the JSON-LD block we ship by default. Every field is there for a reason — none of it is boilerplate. The graph visualisation below shows how the nodes interlink, because the retriever's extractor walks the graph edges as much as it reads the node properties.

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Product",
      "name": "Alora 72\"\" Standing Desk",
      "image": ["https://…/alora-72-oak.jpg"],
      "description": "…the quotable lead goes here…",
      "sku": "ALO-72-OAK-E",
      "brand": { "@type": "Brand", "name": "Alora" },
      "offers": {
        "@type": "Offer",
        "price": "899.00",
        "priceCurrency": "USD",
        "availability": "https://schema.org/InStock",
        "priceValidUntil": "2026-12-31",
        "itemCondition": "https://schema.org/NewCondition",
        "hasMerchantReturnPolicy": { "@id": "#return-policy" },
        "shippingDetails": { "@id": "#shipping" }
      },
      "aggregateRating": {
        "@type": "AggregateRating",
        "ratingValue": "4.7",
        "reviewCount": "184"
      }
    },
    {
      "@type": "FAQPage",
      "mainEntity": [
        { "@type": "Question", "name": "How long does it take to ship?",
          "acceptedAnswer": { "@type": "Answer", "text": "…" } }
      ]
    }
  ]
}

Putting it together

If you do all of this, your PDP will look visually unremarkable. That's the point. AI-cited PDPs don't win on design — they win on linguistic and structural legibility. The reward is that when a shopper asks an assistant “what's the best standing desk for a small office under $1000”, your quotable lead is the sentence that comes back.