AI Shopping Score Explained for Ecommerce

Shopping is changing. When someone asks ChatGPT "What's the best waterproof jacket under £200?" or asks Perplexity to compare running shoes, these AI assistants need to evaluate and recommend products. The question for store owners: will your products make the cut?

AI Shopping Score is a metric developed by PageDiag that measures how well your product pages are structured, described, and optimised for AI-powered shopping assistants. It's a score from 0 to 100 that tells you whether an LLM can understand, evaluate, and confidently recommend your products.

This isn't theoretical. AI-assisted shopping is already happening at scale. Microsoft Copilot integrates shopping recommendations. Google's AI Overviews show product suggestions. ChatGPT's browsing mode pulls product data from the web. If your pages aren't AI-readable, you're invisible to a growing channel.

How LLMs Evaluate Product Data

Large language models don't "see" your beautiful product photography or experience your carefully designed UX. They read your page's content, structured data, and metadata. Here's what they extract and how they use it:

1. Product Identity

An LLM needs to understand what your product is unambiguously. This means:

Clear product name (not "SKU-4892" or "The Explorer")
Category context (is "The Explorer" a backpack, a watch, or a car?)
Brand attribution
Variant information (sizes, colours, materials)

Pages that rely on images to convey what the product is - with minimal descriptive text - score poorly. An LLM can't look at a photo and determine that your product is a merino wool base layer in midnight blue.

2. Product Specifications

AI assistants answering comparison queries need hard data:

Dimensions, weight, materials
Technical specifications
Compatibility information
Care instructions

Stores that bury specs in image-based size charts or PDFs make this data invisible to LLMs.

3. Pricing and Availability

Is the product in stock? What does it cost? Are there shipping restrictions? This information needs to be in structured data, not just rendered on the page.

4. Social Proof

Reviews, ratings, and user-generated content help LLMs assess product quality. A product with 2,000 reviews averaging 4.6 stars gives an AI assistant confidence to recommend it over a product with no reviews.

5. Unique Value Proposition

What makes this product different? LLMs evaluate descriptive content for differentiating claims. "Our jacket uses 800-fill responsibly sourced goose down with fully taped seams" is useful. "The best jacket you'll ever own" is not.

The Role of Structured Data

Structured data (Schema.org markup) is the single most important factor in your AI Shopping Score. It's the difference between an LLM having to guess what your page contains and knowing with certainty.

Minimum Product Schema

Every product page should have comprehensive Product schema:

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Alpine Pro Waterproof Jacket",
  "brand": {
    "@type": "Brand",
    "name": "MountainCore"
  },
  "description": "800-fill goose down waterproof jacket with fully taped seams. Rated to -20°C.",
  "sku": "MC-APJ-001",
  "gtin13": "5901234123457",
  "image": [
    "https://example.com/images/alpine-pro-front.webp",
    "https://example.com/images/alpine-pro-back.webp"
  ],
  "offers": {
    "@type": "Offer",
    "price": "189.99",
    "priceCurrency": "GBP",
    "availability": "https://schema.org/InStock",
    "seller": {
      "@type": "Organization",
      "name": "MountainCore"
    },
    "shippingDetails": {
      "@type": "OfferShippingDetails",
      "shippingRate": {
        "@type": "MonetaryAmount",
        "value": "0",
        "currency": "GBP"
      },
      "deliveryTime": {
        "@type": "ShippingDeliveryTime",
        "handlingTime": {
          "@type": "QuantitativeValue",
          "minValue": 0,
          "maxValue": 1,
          "unitCode": "DAY"
        },
        "transitTime": {
          "@type": "QuantitativeValue",
          "minValue": 1,
          "maxValue": 3,
          "unitCode": "DAY"
        }
      }
    }
  },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.6",
    "reviewCount": "2047"
  },
  "material": "800-fill goose down, Gore-Tex shell",
  "weight": {
    "@type": "QuantitativeValue",
    "value": "680",
    "unitCode": "GRM"
  }
}

What Most Stores Get Wrong

Incomplete schema. Having Product schema with just a name and price is like submitting a CV with just your name. Include every relevant property - material, weight, colour, size options, ratings, shipping, returns policy.

Missing aggregateRating. LLMs heavily weight social proof. If you have reviews but don't include them in your schema, the AI can't factor them in.

No brand property. Brand is a critical signal for product comparison queries. "Best Nike running shoes" can't match your page if the brand isn't in structured data.

Generic descriptions. Your schema description should be information-rich. Don't copy your marketing tagline - write a factual summary that an AI could use to compare your product against alternatives.

How PageDiag Calculates AI Shopping Score

PageDiag's AI Shopping Score evaluates your product pages across several dimensions:

Structured data completeness - How many relevant Schema.org properties are present and correctly formatted
Content quality - Is the product description detailed, specific, and factual?
Technical specifications - Are measurable specs (dimensions, weight, materials) available in text?
Review data - Are reviews present and included in structured data?
Price and availability clarity - Can an AI determine cost and stock status unambiguously?
Page accessibility - Can the content be extracted cleanly, or is it locked in JavaScript-rendered components that LLMs can't access?

A score of 80+ means your product pages are well-optimised for AI shopping. Below 50 means AI assistants are likely skipping your products in favour of competitors with better-structured data.

How to Improve Your AI Shopping Score

Step 1: Audit your current score

Scan your store at pagediag.com. PageDiag analyses your product pages and gives you a specific AI Shopping Score with a breakdown of what's missing.

Step 2: Fix structured data gaps

Use the audit results to add missing Schema.org properties. Most ecommerce platforms have apps or plugins for this:

Shopify: JSON-LD for SEO, Smart SEO
WooCommerce: Yoast SEO, Rank Math (both generate Product schema)
Custom: Implement JSON-LD manually in your template

Step 3: Enrich product descriptions

Rewrite thin descriptions. Each product should have at least 150 words of factual, specification-rich content. Cover:

What the product is and who it's for
Key materials and construction details
Dimensions and weight
What's included in the box
How it differs from similar products

Step 4: Ensure content accessibility

Some stores render product information client-side with JavaScript frameworks. If the content isn't in the initial HTML response, many AI crawlers won't see it. Test by viewing your page source (Ctrl+U) - if the product details aren't there, they're invisible to most bots.

Step 5: Add review markup

If you collect reviews, make sure they're in your structured data. The aggregateRating property is essential. Individual Review schema items are even better.

Why This Matters Now

AI-assisted shopping isn't a future trend - it's current behaviour. Gartner predicts that by 2026, 20% of online purchases will involve AI recommendation. Early movers who optimise for AI Shopping Score now will capture this traffic while competitors are still focused solely on traditional SEO.

The stores that rank well in AI recommendations share three traits: complete structured data, detailed factual content, and strong review profiles. None of this is mysterious, and none of it conflicts with traditional SEO best practices. It's simply a higher bar for data quality.

Check your AI Shopping Score today at pagediag.com and see exactly where your product pages stand.