AI-Readiness Scoring Methodology
Version 2.0 — June 2026
Overview
The ShoppingPartnerLab AI-Readiness Scanner assesses European e-commerce websites on their readiness for AI shopping agents. Each website is scanned across 10 categories with 120+ individual checks, and independently assessed by three AI models.
What we scan
For each website, we analyze:
- The homepage
- A sample of up to 5 product pages (discovered via sitemap)
- Standard web files (robots.txt, sitemap.xml, llms.txt, /.well-known/*)
We do not perform full site crawls. Our scanning mirrors how AI shopping agents evaluate websites in practice: they land on the homepage, look for structured data, and assess a handful of product pages.
The 10 categories
| Category | Weight | What it measures |
|---|---|---|
| AI Visibility Score | 15% | Independent assessment by 3 AI models (Gemini, Claude, Perplexity) |
| Agentic Commerce Readiness | 15% | UCP/ACP profiles, MCP discovery, checkout automation, API discoverability |
| Schema.org Compliance | 13% | Product schema completeness, GTIN validity, Merchant Center alignment |
| Product Content Intelligence | 13% | Content depth, freshness, alt text quality, multilingual quality |
| Technical Setup | 12% | JS shell detection, sitemap, robots.txt, AI crawler permissions, security headers |
| Trust & Authority | 12% | Business verification, external trust (Trustpilot, Google Reviews), social presence |
| Transaction Readiness | 10% | Payment methods, price consistency, availability accuracy, cart functionality |
| Operational Maturity | 10% | Shipping detail, return policy, error handling, cookie consent, performance |
| Google Agent-Friendliness | Bonus | Semantic HTML, layout stability, accessible names, WebMCP (based on Google web.dev guidance) |
| Product Data Integrity | Bonus | Price, availability, and review consistency between schema and visible page |
AI model assessment
Three AI models independently evaluate each website:
- Google Gemini: assesses semantic clarity and structured data quality
- Anthropic Claude: evaluates technical SEO and content structure
- Perplexity: measures content richness and citation potential
Each model provides a score (0-100) and a written assessment. The AI Visibility Score is the average of all three assessments. Models do not see each other's scores.
Scoring
Each category produces a score of 0-100 based on individual check results. The overall score is a weighted average using the weights listed above.
Trust levels
| Level | Score | Meaning |
|---|---|---|
| Exceptional | 90-100 | Top-tier AI-readiness, fully optimized |
| High | 80-89 | Trust Registry qualified, strong readiness |
| Moderate | 60-79 | Basic readiness present, improvements needed |
| Low | 40-59 | Significant gaps in AI-readiness |
| Critical | 0-39 | Not suitable for AI agent interaction |
Consistency checks
We verify data integrity by comparing structured data (JSON-LD) with visible page content:
- Price consistency: schema price vs. displayed price (±5% tolerance)
- Availability consistency: schema stock status vs. page indicators
- Review consistency: schema ratings vs. displayed ratings (±0.3 stars, ±1.5× count tolerance)
Limitations
- PageSpeed and Core Web Vitals are deliberately excluded — AI agents don't experience visual load times
- The scanner analyzes a sample of pages, not the entire website
- AI model assessments are automated and may not capture every nuance
- Scores reflect a point-in-time assessment
- The discovery engine may have a selection bias toward more discoverable shops
Data sources and references
Our methodology is informed by:
- Google web.dev: "Build agent-friendly websites" (2026)
- HTTP Archive: Web Almanac structured data analysis
- Schema.org specification (v30, March 2026)
- Universal Commerce Protocol (UCP v2026-04-08)
- Agentic Commerce Protocol (ACP v2026-04-17)
- Model Context Protocol (MCP, modelcontextprotocol.io)
Update frequency
- Trust Registry qualified shops (≥80): rescanned weekly
- Moderate shops (60-79): rescanned biweekly
- Low shops (40-59): rescanned monthly
- Critical shops (<40): rescanned every 60 days
Contact
Questions about our methodology: info@shoppingpartnerlab.com
Full datasets: shoppingpartnerlab.com/en/datasets