How to make your fashion PDP readable by AI shopping agents — the agentic commerce checklist (2026)

Why this matters now

Shopping-related searches on generative AI platforms grew 4,700% between July 2024 and July 2025 (BoF × McKinsey, State of Fashion 2026, p.38). ChatGPT already drives 16% of Zara's inbound referral traffic and 8% of H&M's (State of Fashion 2026, p.38, citing SimilarWeb). OpenAI has struck deals with both Shopify and Etsy to enable direct purchasing through ChatGPT (State of Fashion 2026, p.40, p.44). Amazon has launched "Buy for Me". Google and Perplexity both have purchase agents in market.

The McKinsey and Business of Fashion State of Fashion 2026 report makes the commercial priority explicit: brands should "optimise for AI discoverability by overhauling product content and digital infrastructure so AI agents can readily access and read it" (p.37, p.43). It adds that "semantically rich data and API-accessible content will be critical to success" (p.37). And it quotes Marc Bain on the stakes: brands not appearing in an AI agent's consideration set "may as well be invisible" (p.44).

GEO — Generative Engine Optimisation — is what McKinsey calls "the new SEO" and a "critical counterpart to SEO" (State of Fashion 2026, p.7, p.39). For fashion brands, GEO is not a future-state project. It is a current competitive decision.

The checklist below covers the seven areas that determine whether an AI agent can read, evaluate, and recommend your fashion products. For each, we note what the gap typically looks like on a Shopify store, what to do, and — where relevant — how RenderedFits' agentic-commerce layer provides a turnkey implementation for the fit-and-sizing elements.

1. Structured product data — write for machines, not just shoppers

AI agents parse raw text. Vague copy — "a beautiful floaty dress in our signature silk" — cannot be matched against a shopper's requirements. Concrete attributes can.

For every product page, ensure:

• Product title includes garment type, brand, and at least one key attribute (material or fit) — not just a style name.
• Description states: material composition (percentage breakdowns where relevant), care instructions, fit type (oversized/relaxed/slim/structured), body-length reference (mini/midi/maxi, or cm from waist), and country of origin if premium-relevant.
• Size options are labelled in full (Small, not S; UK 10, not 10) and include a prose size guide on the PDP itself — not a separate linked PDF.
• Dimensions table (chest, waist, hip, length) is present per size, in centimetres, as on-page text (not only an image).

2. Product JSON-LD — Schema.org markup on every PDP

JSON-LD is the format AI crawlers and shopping agents read most reliably. Shopify themes often include a partial Product schema block by default; most are missing fields that matter to agents.

Minimum required fields for a Schema.org/Product block:

• name, brand, description
• image (array — include at least 3 angles)
• sku (per variant)
• offers: price, priceCurrency, availability, url

Recommended additions that meaningfully improve agent comprehension:

• material, color, size
• aggregateRating (with ratingCount)
• review (array, with reviewBody and author)
• sizeSpecification (Schema.org/SizeSpecification)

Audit your current markup using Google's Rich Results Test. Missing required fields generate errors; missing recommended fields generate warnings. Errors suppress structured-result eligibility. Warnings reduce agent comprehension.

3. Page metadata — title, description, and Open Graph

Perplexity, ChatGPT Browse, Google AI Mode, and similar systems read page metadata as the authoritative one-line summary of your page. They surface this directly in citations and shopping results.

• Page title should lead with the product name and brand. Avoid leading with your site name.
• Meta description should state the key purchase-decision attributes in one sentence: material, fit, use-case, and price point if distinctive. This is what an AI answer engine surfaces in a citation.
• og:title and og:description should match or closely mirror the title and meta description. These are read by social crawlers and AI systems parsing Open Graph data.
• og:image should be a clean, high-resolution product shot — preferably on a person, not a hanger. AI systems use images as a quality signal for the underlying product.

Shopify's default meta templates use the product title as the page title and a truncated description as the meta description. Both are usually adequate, but most themes do not populate og:image correctly (they use the first product image at full-size rather than a cropped 1200×630 social asset). This is a minor fix with a meaningful impact on how the page appears when cited.

4. Social proof and reviews — structured and on-page

Approximately 80% of sources AI systems draw on are third-party content — reviews, editorial, affiliate — rather than brand-owned pages (State of Fashion 2026, p.41). This has two implications: first, your on-site reviews need structured markup so they are readable as authoritative signals; second, off-site reviews on platforms AI already trusts amplify your visibility beyond your domain.

For on-site reviews:

• Add Schema.org/Review and Schema.org/AggregateRating to your Product JSON-LD block.
• Prioritise displaying reviews that use natural purchase-language: "true to size", "runs small in the shoulders", "surprisingly warm for the weight" — the language an AI agent will match against a shopper's stated requirements.
• Include reviewer attributes where consented (height, usual size, body type): structured agent-readable fit testimonials are more valuable than general praise.

For off-site reviews:

• The Shopify App Store (if you are an app) and Google Business reviews are indexed and cited by AI systems.
• Trustpilot, Yotpo published pages, and editorial mentions in fashion media (Vogue, WWD, Drapers) carry disproportionate weight as third-party validation sources.
• AI answer engines use grounding queries to validate brands they surface — searches for "[brand] review", "[brand] returns policy", "is [brand] legit". Having rich, independently indexed content at these query shapes is what survives the validation pass.

5. API-accessible fit and sizing data — the critical gap for fashion

Fit and sizing are the decisive purchase-confidence factors in fashion, and the hardest to expose in a machine-readable way. A size guide written in a table is better than a PDF; a size guide in JSON-LD is better still; a product-level JSON endpoint an agent can call with a product ID is the most machine-useful form.

The McKinsey exec priority is explicit: "semantically rich data and API-accessible content will be critical" (State of Fashion 2026, p.37). In fashion, fit IS the semantically rich data. Brands that expose it in structured, API-accessible form give AI agents exactly what they need to make confident product recommendations.

Three implementation options in increasing order of machine-readability:

For most Shopify fashion brands today, the realistic starting point is Option A (structured text) combined with an immediately deployable Option C via RenderedFits. The try-on widget, fit endpoint, and fit_profile metafield install together — the fit-data infrastructure is a by-product of the purchase-confidence improvement the widget delivers to human shoppers.

6. JSON-LD for fit and try-on capability — Schema.org and AEO extensions

Where fit data exists, it should be expressed in JSON-LD as well as in the API endpoint. This makes it readable by any crawling agent or answer engine that parses structured data — without requiring an active API call.

The relevant Schema.org types are:

• SizeSpecification — for declared size dimensions and size system (e.g. UKSize).
• SizeGuide — for linking to or embedding a structured size guide.
• ItemList combined with review data — for surfacing fit-consensus signals (e.g. "87% of reviewers found this true to size").

For RenderedFits merchants, a /apps/rendered-fits/fit-jsonld/{productId} endpoint provides Schema.org-compatible JSON-LD ready to embed in the PDP <head>. This makes the fit-confidence signal directly readable by any agent or answer engine that parses structured data, with no manual markup authoring required on the merchant side.

7. Agent capability manifest — declare what agents can do with your store

An agent capability manifest is a machine-readable declaration of the endpoints and capabilities available at your store. Think of it as a sitemap.xml for agent interaction rather than content indexing. It tells AI shopping agents what data is available, how to access it, and what attribution to apply when those capabilities drive a sale.

A minimal fashion-store agent manifest might declare:

• That a fit endpoint exists and what product IDs it accepts.
• That a fit-jsonld endpoint provides Schema.org-compatible fit data.
• The attribution format expected when fit data is used in a purchase recommendation.

Agent manifests are not yet a universal protocol — there is no single standard as of mid-2026. But OpenAI's Shopify integration and Google's Shopping Graph are both building in the direction of structured capability discovery, and early publishers of agent manifests will benefit from the same first-mover dynamic that early sitemap publishers captured in the SEO era.

For RenderedFits merchants, an /apps/rendered-fits/agent-manifest endpoint is available at every store, listing the fit capabilities (fit.query, fit.sizing, fit.visual.tryon) with endpoint URLs, schema references, and attribution metadata. This is ready without any configuration on the merchant side.

The full checklist at a glance

Frequently asked questions

Related reading

Sources

• BoF × McKinsey — The State of Fashion 2026: When the Rules Change. Key figures cited in this page: 4,700% growth in AI-platform shopping searches (p.38); 53% of US gen-AI users used it to shop Q2 2025 (p.37–38); $3–5 trillion agentic commerce value by 2030 (p.42); 11–18% of B2C retail by 2030 (p.42); ~80% of AI-cited sources are third-party content (p.41); ChatGPT = 16% of Zara's inbound traffic (p.38, citing SimilarWeb); OpenAI/Shopify and OpenAI/Etsy deals (p.40, p.44); exec priority "optimise for AI discoverability" (p.37, p.43); "semantically rich data and API-accessible content" (p.37); GEO = "the new SEO" (p.7, p.39); "may as well be invisible" quote (p.44). mckinsey.com / businessoffashion.com
• Schema.org — Product, SizeSpecification, SizeGuide, Review, AggregateRating. Markup references throughout. schema.org/Product
• Google — Rich Results Test. For auditing PDP JSON-LD. search.google.com
• RenderedFits — Agentic Commerce Layer documentation. fit_profile metafield, /fit/ endpoint, /fit-jsonld/ endpoint, /agent-manifest endpoint — built 2026-06-18 on branch feature/agentic-commerce.