Scoring Model
How Colter scores agent readiness across 5 dimensions. Weights, signals, score ranges, and how to improve.
TL;DR: Colter scores stores from 0 to 100 across transaction, discovery, ecosystem, content quality, and security. The fastest way to improve the score is usually UCP, ACP, JSON-LD,
llms.txt, and security headers.
Formula
composite = round(
transaction * 0.30 +
discovery * 0.25 +
ecosystem * 0.20 +
content_quality * 0.15 +
security * 0.10
)
Dimensions
| Dimension | Weight | High-value signals |
|---|---|---|
| Transaction | 30% | UCP, ACP, checkout capabilities, fast protocol endpoints, compatible PSP, declared agent-payment rails |
| Discovery | 25% | JSON-LD, sitemap, llms.txt, ai-plugin.json, robots posture, Open Graph |
| Ecosystem | 20% | Google coverage, OpenAI coverage, MCP, WebMCP |
| Content Quality | 15% | Complete product schema, price, availability, images, SSR |
| Security | 10% | HTTPS, HSTS, CSP, XFO, X-Content-Type-Options, rate limiting |
Score Bands
| Score | Meaning |
|---|---|
75-100 | Strong readiness |
60-74 | Basic readiness |
40-59 | Needs work |
0-39 | Not ready |
Agentic Commerce Level
Colter also emits an agentic_level and label:
| Level | Label | Meaning |
|---|---|---|
| 1 | Web Forms | Basic site presence |
| 2 | Descriptive Search | Agents can find and read the catalog |
| 3 | Persistence | Stronger agent identity and handoff support |
| 4 | Delegation | Agents can browse, cart, and checkout |
| 4.9 max | Progress toward Level 5 | Level 5 behavior is not directly measurable today |
Fix Tiers
Every fix Colter surfaces is classified by tier. The tier controls how the fix appears in the report and how its absence affects your score.
| Tier | Absence | Label in report | When used |
|---|---|---|---|
| Required | Costs you score and is flagged as a "Gap" | Gap: prefix | Feature of a main protocol (UCP, ACP) or blocks a main agent (Claude, ChatGPT, Gemini) from shopping the site |
| Recommended | Costs you score and is flagged as an "Opportunity" | Opportunity: prefix | Real operational value, but your current stack may already cover it via another vendor |
| Bonus | Never costs score; never shown as a fix | Surfaced in the "Leading edge" commentary | Emerging protocols or leading-edge rails — credit if present, never penalized if absent |
The score accumulation respects tier: Required items must be present to hit full dimension credit; Recommended items contribute when present but never subtract; Bonus items only add.
Highest-Impact Fixes
| Fix | Typical effect | Tier |
|---|---|---|
| Add UCP | Big transaction and ecosystem lift | Required |
| Add ACP | Big transaction and ecosystem lift | Required |
| Add complete Product JSON-LD | Discovery and content quality lift | Required |
Add llms.txt | Discovery and ecosystem lift | Recommended |
| Add security headers | Security lift | Recommended |
| Enable Google AI Mode / Gemini checkout (UCP catalog search + structured data with gtin/mpn/sku) | Transaction lift, Gemini compare-and-buy readiness | Recommended |
Declare agent-payment readiness (x402, AP2, MPP, TAP, Google Pay, card networks, or compatible PSP handlers) | Small transaction lift and payment-trust commentary | Recommended |
Shopify Storefront Catalog (UCP migration)
Shopify opened UCP with Shopify Catalog as the default Shopify-native readiness path. Colter treats Shopify Catalog / Agentic Storefronts as platform-provided coverage when a Shopify store does not expose merchant-specific public UCP evidence, while still distinguishing public endpoint evidence from Shopify-native baseline coverage.
Shopify migrated the Storefront Catalog MCP to UCP at /api/ucp/mcp with new tool names (search_catalog, lookup_catalog, get_product). Legacy Storefront MCP tool names are being deprecated.
- Before 2026-06-15: stores on legacy-only tool names get partial credit and a Recommended-tier fix prompting migration.
- On or after 2026-06-15: legacy-only is scored as a Required-tier Gap. Migrate to the new endpoint before that date.
Current UCP Extensions
Colter detects current UCP extension declarations including fulfillment, buyer consent, AP2 mandate, and loyalty. These are surfaced as report evidence so merchants can see whether the manifest is tracking the latest UCP surface area. They do not change score weight until the public scoring model is recalibrated against live adoption data.
Payment And Trust Rails
Payment and trust rails are scored conservatively. Colter gives recommended/bonus credit when a UCP manifest or related commerce surface declares recognized payment handlers such as Stripe, Visa, Mastercard, Amex, Adyen, Braintree, Google Pay, PayPal, x402, AP2, MPP, or TAP.
These signals do not replace UCP/ACP checkout coverage. They help distinguish stores that expose live product, stock, cart, checkout, payment-policy, and settlement primitives from stores that only publish catalog metadata.
How To Use This Page
- Use Check to see current scores.
- Use Fix to get ranked improvements.
- Use Verify when you need conformance evidence.