Back to home

Scoring Methodology

The Shelfmark AI Shopping Readiness Score (0–100) measures how reliably AI shopping assistants (ChatGPT Search, Microsoft Copilot, Google AI Mode) can parse, comprehend, and recommend your Shopify store’s products.

1. Title clarity

Weight: 25%

AI language models use product titles as primary semantic anchors. When a shopper asks for a “navy blue merino wool cardigan,” titles with clear nouns and attributes match directly.

Fail: Title has fewer than 3 words, or contains only a brand/creative name with no product noun (e.g. “The Luna”).

Warn: Product noun is present, but lacks descriptive attributes (material, color, size, or audience).

Pass: Clearly names what the item is and includes at least one descriptive attribute.

2. Description depth

Weight: 25%

AI shopping agents answer detailed customer inquiries using product descriptions. Vague marketing fluff offers zero facts for an AI to cite.

Fail: Empty, missing, or under 40 plain-text words.

Warn: 40 to 120 words, or heavy promotional adjectives (“stunning, revolutionary, must-have”) with few concrete facts.

Pass: Over 120 words containing concrete facts, use cases, and specifications.

3. Key details present

Weight: 20%

Inspects the title and description for concrete specification categories: materials, dimensions/size, weight, ingredients, compatibility, and care instructions.

Fail: Zero concrete specification types detected.

Warn: Only one specification type found.

Pass: Two or more specification categories explicitly documented.

4. Product type & tags

Weight: 10%

Taxonomy classification helps search agents organize catalogs into standard commerce ontologies.

Fail: Both product_type and tags are completely empty.

Warn: Either type or tags is missing, or uses generic tags like “sale” or “featured”.

Pass: Both a standard product type and meaningful descriptive tags are present.

5. Variant & option naming

Weight: 10%

Shoppers ask AI models to verify size and color availability. Generic labels break multi-attribute queries.

Fail: Generic option names like “Option 1” with multiple variants, or titles consisting solely of cryptic SKU codes.

Pass: Legible option headers (e.g. Size, Color, Volume) and human-readable values.

6. Image coverage

Weight: 10%

Visual AI models cross-reference text with product imagery and alternative text descriptions.

Fail: No product images in the public feed.

Warn: Only a single image, or images missing alternative text.

Pass: Two or more images with descriptive alt text.

What we do not measure

Shelfmark reads only public product feed data. We cannot and do not measure:

  • Private store sales figures or revenue data
  • Conversion rates or advertising performance
  • Customer reviews, ratings, or third-party app widgets
  • Unpublished draft products or password-protected pages

If a field is not exposed in your public store feed, our audit marks it not visible rather than penalizing your score.