Skip to content
Paz.aiAgentic Store Score

Scoring methodology

The Agentic Store Score evaluates how ready an ecommerce store is for AI shopping agents. Each scan checks the public parts of your store that agents can discover, retrieve, and use to begin a shopping journey. The public scan never submits checkout or payment.

Score model

The score is out of 100 points, divided equally across five essential categories (20 points each). A bonus of up to +5 points can be earned from emerging signals - their absence never lowers the score.

01. Discovery and crawlability

20 pts
  • robots.txt exists and allows AI agent crawlers
  • XML sitemap published with product URLs
  • robots.txt references the sitemap location
  • Homepage content available without JavaScript (SSR)
  • Homepage has Organization/Store JSON-LD structured data

02. Product structured-data consistency

20 pts
  • Product JSON-LD on all sampled PDPs
  • Consistent schema fields across products (name, image, description, sku, brand, offers)
  • PDPs accessible without JavaScript (server-side rendered)
  • Product images referenced in JSON-LD
  • Product descriptions in JSON-LD and visible HTML

03. Price, stock, and variant fidelity

20 pts
  • Price visible on all sampled PDPs (JSON-LD offers.price or HTML)
  • Availability/stock status on all sampled PDPs
  • Product variants (size, color) exposed in structured data

04. Agent-operable controls and cart handoff

20 pts
  • Cart/checkout endpoint accessible to agents
  • Add-to-cart controls present on homepage
  • Add-to-cart form controls on all sampled PDPs

05. Shipping, returns, merchant, and trust clarity

20 pts
  • Shipping policy page accessible
  • Return/refund policy page accessible
  • Merchant/contact information accessible
  • Terms of service accessible

Bonus signals

Emerging agent protocol signals earn bonus credit (capped at +5 points total). These are forward-looking indicators of agent readiness - they never count against you when absent.

llms.txt+1.25 pts
agents.md+1.25 pts
MCP server descriptor+1.25 pts
AI commerce catalog+1.25 pts

How the scan works

  1. Fetch homepage - The scanner fetches the store homepage HTML using a server-side fetch with an agent user-agent string.
  2. Discover PDPs - Product detail pages are discovered from JSON-LD ItemList, product-like anchor links, and the XML sitemap.
  3. Sample 5 products - Five representative products are evenly sampled from the discovered URLs.
  4. Fetch each PDP - Each product page is fetched and analyzed for structured data, pricing, availability, variants, images, and cart controls.
  5. Run store-level checks - Policy pages, cart endpoints, robots.txt, sitemap, and emerging signals are checked.
  6. Aggregate score - All check scores are combined and normalized to 0–100. At least three product pages and 70% of essential check weight must have conclusive evidence; otherwise the report is marked insufficient evidence and no score is assigned. Unavailable checks are excluded rather than counted as merchant failures.
  7. Store result - The completed scan is stored in a durable report store and made available via the report page and API.

Grade scale

A90–100Excellent agent readiness
B80–89Strong agent readiness
C70–79Moderate agent readiness
D60–69Important blockers remain
E40–59Significant gaps detected
F0–39Not ready for AI agents
N/AUnscoredInsufficient product evidence