How scoring works

Every product in your catalog receives an AI Readability Score from 0 to 100. This score represents how well an AI shopping agent can understand, categorize, and recommend the product.

Note that while a high score maximizes your products’ chances of being discovered by AI agents, allowing customers to purchase directly within the AI channel requires your store to meet Shopify’s Agentic Storefronts requirements.

The five dimensions

Description quality (30% of score)

The description dimension assesses whether a product description gives an AI agent enough information to accurately represent the product. Factors include:

• Word count — descriptions under 80 words score lower; 120–200 words is the target range
• Factual density — how many specific, verifiable facts are present (material, dimensions, use case, compatibility, etc.)
• Semantic richness — variety of relevant terms that help with categorization
• LLM comprehension test — a sample prompt is run against the description to verify an AI agent can answer basic questions about the product correctly

Attribute completeness (25%)

AI agents use structured attributes to match products to queries. This dimension checks:

• Presence of key attributes for the product’s category (color, size, material, fit, style)
• Whether attribute values use standard, interpretable terms — “navy blue” vs. “Midnight Forest,” “slim fit” vs. “fits TTS”
• Presence in both product metafields and schema markup

Ambiguous values score lower because AI agents cannot reliably map them to search concepts.

Schema markup (20%)

Schema.org structured data tells search engines and AI agents how to interpret your page content. This dimension checks for presence and completeness of:

• Product — base product type declaration
• Offer — price, availability, currency
• ShippingDetails — shipping cost and time
• MerchantReturnPolicy — return window and conditions
• AggregateRating — review score and count (if applicable)
• additionalProperty — structured key-value pairs for product attributes

Variant structure (15%)

Multi-variant products (e.g., a hoodie available in 5 colors and 4 sizes) need correct ProductGroup schema to prevent AI agents from treating each variant as an unrelated product. This dimension checks:

• Presence of ProductGroup schema linking all variants
• Consistent option naming across variants (e.g., all variants using “Color” not a mix of “Color” and “Colour”)
• Correct Shopify Catalog Mapping variant grouping configuration

Product identifiers (10%)

GTINs, MPNs, EANs, and barcodes allow AI agents to match your products to external databases and verify product identity. Missing identifiers reduce trust signals. This dimension checks:

• Barcode/GTIN population on variants
• MPN presence (especially for non-commodity products)

Score bands

Band	Range	Interpretation
AI-Ready	90–100	AI agents can accurately handle this product
Good	70–89	Minor gaps; generally handled correctly
Needs Work	40–69	Significant gaps; may be misrepresented or skipped
Critical	0–39	Severe gaps; unlikely to be recommended

Catalog-level metrics

The diagnosis also provides:

• Median and mean score across your catalog
• Distribution showing how many products fall in each band
• Category-level breakdown if your products span multiple categories
• Top 5 systemic issues — issues appearing on the most products, ranked by combined impact