§ 01 Why sizing matters
Why sizing is one of the three fields Shopify's AI doc names explicitly
Shopify's AI optimization doc lists the recommended fields for AI visibility once for the whole platform. Most categories receive a generic recommendation ('detailed specifications and technical details'). Apparel gets three category-specific fields named: sizing guides, material information, and care instructions. The reason is that fit is the question AI shopping engines are asked most often in apparel, and the engines route that question to whichever fields the merchant populates. Brands without structured sizing answers get skipped on fit queries regardless of how good their PDP copy is.
The Shopify AI optimization doc1 publishes the field list verbatim and calls out sizing, material, and care as the apparel-relevant surfaces. The Catalog optimization doc2 separately lists Variants (with Option name) as an AI-readable field — Size as a variant axis interacts with sizing schema as a content axis, and both have to be populated consistently for the engines to surface the answer cleanly. A garment with "Size" as an Option name and no sizing schema sends the engine inconsistent signal.
§ 02 The three layers
Three implementation layers — schema, metafield, FAQ
Layer one: schema additionalProperty on the Product schema. PropertyValue blocks carrying 'Chest (Size M)', '40 inches', and equivalent measurements per size. Layer two: structured metafields with a measurement-set type (chest, waist, hip, inseam, length) per variant, rendered on the PDP as a sizing table. Layer three: Knowledge Base FAQs answering the prose questions buyers ask the engine ('how does this fit', 'should I size up if I'm between sizes', 'how do these run compared to standard sizing').
The schema layer is the AI-readable surface. schema.org's size property3 accepts SizeSpecification or QuantitativeValue, and the install adds additionalProperty blocks with the measurement detail engines parse most cleanly — chest, waist, hip, inseam, length, sleeve, shoulder. The metafield layer is the merchant-editable surface that renders the same data as the PDP sizing table for human readers. The Knowledge Base FAQ layer4 is the conversational layer — the answer text AI agents surface when the buyer's question is "will this fit me" rather than "what's the chest measurement."
§ 03 The fit FAQ
The Knowledge Base fit-FAQ pattern
Five FAQ entries cover most apparel fit queries. 'How does this fit?' (true-to-size / relaxed / fitted, with reference body type). 'Should I size up if I'm between sizes?' (yes/no recommendation, with the silhouette logic). 'How do these run compared to other brands?' (cross-brand reference for buyers familiar with another label). 'What's the difference between Size M and Size L in chest measurement?' (numeric anchor for the engine to extract). 'What if it doesn't fit?' (return and exchange mechanics, linked to the return policy).
The Knowledge Base managing-FAQs doc4 specifies 1-2 sentence answers per FAQ, stored as metaobjects under Content > Metaobjects. The five-FAQ set above is the apparel default; brands with technical fit specifications (compression wear, performance apparel, formal tailoring) add category-specific FAQs on top. The AI agents Shopify routes through Knowledge Base have access to these answers when buyers ask the engine fit questions, which is the moment the brand earns or loses the citation.
§ 04 Multi-region
Multi-region sizing — US, UK, EU, AU
Apparel brands selling internationally face the size-conversion problem at the AI extraction layer. A US Size 8 dress is UK 12, EU 38, AU 12 — and AI shopping engines parse the buyer's geo signal to surface the size in the correct system. The install adds region-specific size labels to the variant graph (or, more commonly, an additionalProperty mapping table on Product schema) and a Knowledge Base FAQ covering size conversion explicitly.
The structured solution is an additionalProperty mapping table on the Product schema with a PropertyValue per region (US, UK, EU, AU, JP) for each Size variant. The Knowledge Base FAQ "What size am I in your sizing if I wear UK 10?" carries the prose answer for engines routing conversational queries. Brands operating on Shopify Markets can attach region-specific Catalog feeds with localized size labels, which lets the engine surface the right label without parsing a conversion table — the cleaner path when international volume justifies it.
§ 05 Anti-patterns
What not to do — three patterns that suppress AI extraction
Three patterns suppress AI extraction of sizing. First: sizing as a static image only (the sizing chart as a PNG with no underlying structured data). Second: sizing in the PDP body as freeform prose without measurement anchors. Third: sizing as a separate Page URL linked from the PDP but not embedded as schema on the PDP itself.
The static-image pattern is the most common — a sizing chart designed for human readers, uploaded as a PNG, with no alt text beyond "sizing chart" and no underlying structured data5. AI engines cannot extract measurements from the image. The freeform-prose pattern ("our medium fits a typical 38-40 inch chest") is better but still suppresses extraction because the engine has to parse the prose for numbers and units. The separate-Page pattern (sizing chart at /pages/size-guide) keeps the data off the PDP, which means the engine connecting a product to its sizing has to chain two requests instead of one. The install fixes all three with the schema + metafield + FAQ stack on the PDP itself.
