Arenza is now part of the OpenAI Partner Network 🎉
← State of Agentic Commerce
Agentic Commerce Score · scanned 2026-07-23 · rubric v0.2

gymshark.com

56/100
grade C
not yet agent-buyable
platform: shopify

Can an AI shopping agent buy from gymshark.com?

As of 2026-07-23, no: gymshark.com scores 56/100 (grade C) on the Agentic Commerce Score and does not meet the agent-buyable bar. The failing checks are: Homepage readable without JavaScript, Product JSON-LD (Offer: price/currency/availability), Brand + product identifier (sku/gtin/mpn), and 3 more.

67
Discover — can agents fetch + read the store? (30%)
44
Evaluate — can agents parse + trust the products? (45%)
64
Transact — can an agent complete a purchase? (25%)

Discover — can agents fetch + read the store?

✓ AI crawlers allowed (robots.txt)
No AI shopping crawler is root-blocked in robots.txt.
✗ Homepage readable without JavaScript
Homepage carries a noindex robots meta — AI systems are told to skip it.
✓ Sitemap discoverable
Sitemap found.
✓ llms.txt present
/llms.txt found.

Evaluate — can agents parse + trust the products?

✓ Machine-readable product catalog
Open product feed at https://gymshark.com/products.json?limit=100 (100 products sampled).
✓ Required fields (title/image/price/description)
Coverage over 100 feed products: title 100% · image 100% · price 100% · description 100% (avg 100%).
✗ Product JSON-LD (Offer: price/currency/availability)
0/3 sampled product page(s) carry Product JSON-LD with a complete Offer (price + priceCurrency + availability); 0/3 carry any Product schema.
✗ Brand + product identifier (sku/gtin/mpn)
0/3 sampled product page(s) expose both brand and an identifier (sku/gtin/mpn); brand alone on 0/3.
✗ AggregateRating in Product JSON-LD
0/3 sampled product page(s) expose machine-readable review ratings.
✗ Product image alt text
0% of content images on sampled product page(s) carry non-empty alt text.
✓ Description depth (≥120 chars)
100% of feed products have a description agents can quote (≥120 chars).

Transact — can an agent complete a purchase?

✓ Agentic-checkout rail (platform prerequisite)
Platform: shopify. Open /products.json feed + Shop Pay signals — Shopify storefront; platform ships ChatGPT-checkout (ACP) and Google (UCP) catalog rails for eligible stores.
✓ Machine-readable price + availability
Feed exposes price on 100% and availability on 100% of products.
✓ Per-variant stock readable
100% of products expose an availability flag on every variant (size/colour level stock).
✗ Shipping + returns policies (with terms)
Neither a shipping nor a returns policy could be found — agentic checkout surfaces these before completing a purchase.

Highest-impact fixes

  1. Serve real HTML without JavaScript: agent fetchers read the raw response. Pre-render or SSR your storefront.
  2. Add complete Product JSON-LD on product pages: an Offer with price + priceCurrency + availability.
  3. Publish shipping and returns policies as real pages with concrete terms (state the return window in days) — agentic checkout surfaces merchant policies before completing a purchase.
  4. Publish brand plus a product identifier (sku / gtin / mpn) in Product JSON-LD — agents match, dedupe and price-compare products by identifier; without one your listing is an orphan.
  5. Expose reviews as AggregateRating in Product JSON-LD (ratingValue + reviewCount) — agents quote ratings when ranking options, and skip products that have none they can read.
  6. Write descriptive alt text on product images — it is the only thing a text-mode agent can read about your photography.

Is this your store?

This is a point-in-time scan of public signals (2026-07-23) — if you have shipped fixes since, re-run it in ten seconds: npx agentic-commerce-score gymshark.com.

ACS covers whether agents can buy from you. The other half is whether AI assistants recommend you over competitors — run Arenza's free Agentic Visibility Score or write to hello@arenza.ai.

Think a check is wrong? Every result carries the evidence it was based on, and the scanner is open source — so a disputed score is a reproducible question, not an opinion. Mail hello@arenza.ai and we will re-scan, correct the published page if we got it wrong, and note the correction. Fixed the gap instead? Tell us and we will re-scan and update this page.