UCPScore Intelligence Desk · Press release

UCPScore reaches 97/100 self-applying its own rubric, with the open gap cited publicly.

Deterministic 4-protocol scoring engine validates itself against its own public methodology. Audit receipts SHA-pinned. Pattern reproduces.

May 7, 2026 · San Francisco

UCPScore — the Inference Advantage Engine for ecommerce — announced today that it reached 97/100 on its own public rubric, becoming the first AI-readiness scoring product to verify itself against the same methodology it sells to merchants. The audit was completed on May 7, 2026 and is reproducible from a single SHA-pinned state-receipt at 1fc3e43397599394005644ad900c83420e961433.

UCPScore's own rubric scores 97/100 — not 100. The remaining 3 points are concentrated in a single open gap: Gap C, the Phase 3 LLM-judge golden-set calibration, which requires a 200×5 annotated corpus + live annotation campaign infrastructure that has not yet shipped. The gap is documented at production-pipeline/gaps.json#L47-L54 with the closure path. We could have shipped this announcement with a manufactured 100, but receipt-grade integrity means publishing what the rubric actually shows. The gap closes when the infrastructure lands; the receipt will update at that time. The methodology, including the open gap, is public.

The result matters because every other tool in the AI-readiness category measures stores from the outside without applying the same standard to itself. UCPScore’s rubric is public, deterministic, and re-runnable; a competitor or auditor can reproduce the 97/100 score by running the same scanner against ucpscore.ai with the same rubric SHA. No private benchmark. No marketing screenshot. Just the receipt.

Pillar 1 · The entity graph passes its own audit

The self-application audit on 2026-05-07 confirmed 7 of 7 entity-graph signals live on ucpscore.ai: <title>, <meta description>, <meta keywords> with 10 entity terms, 6 OpenGraph properties, and a complete schema.org graph spanning Organization + SoftwareApplication + Offer + WebSite + SearchAction nodes. The audit ran the --avatar ai-agent-crawler profile from UCPScore’s own avatar-critic skill — the same skill UCPScore ships to merchants for content review.

Pillar 2 · Zero positioning leaks

UCPScore’s end-goal market is ecommerce broadly; Shopify is the launch wedge, not the ceiling. The audit confirmed zero “for Shopify” matches in the home-page HTML and 16 instances of “Inference Advantage Engine” across the pricing surface — heavy entity reinforcement at the AI-graph layer. Technical-accuracy copy (form placeholders, FAQ answers specific to V1 Shopify support, fix labels for Shopify metafields) remains accurate to the v1 product capability. The split is doctrinal: positioning is broad; technical accuracy is narrow.

Pillar 3 · Every gap closure is receipt-pinned

The 95.5 → 97 delta is documented in production-pipeline/gaps.json with commit-pinned closure SHAs for every gap. The closing PRs ship through the same production-state-verifier CI gate UCPScore tells merchants to apply to their own scoring deltas. The verifier emits a state-receipt JSON for every PR; the receipt for the 97/100 close is at .claude/state-receipts/1fc3e43397599394005644ad900c83420e961433.json. Re-run the public rubric against ucpscore.ai with the same rubric-SHA and the same store-state and you get the same number. Deterministic by design.

Methodology

  • Public rubric. 9 dimensions, 18 deterministic checks, weights calibrated against the 1,741-store benchmark (May 2026, avg 62.1/100). Methodology is public; the rubric SHA is pinned.
  • 4 protocol stack. Universal Commerce Protocol (UCP) + Agentic Commerce Protocol (ACP) + Model Context Protocol (MCP) + Agent Payments Protocol (AP2). Conformance citations carry the protocol target SHA per scan; mirrored against UCP’s own conformance test suite.
  • Receipt-pinned proof. Every scan emits a cryptographic state-receipt: store URL + rubric SHA + score + per-dimension breakdown + conformance citations. Receipts are merchant-portable; share the URL, the receipt travels with it.
  • Self-application discipline. UCPScore’s own marketing surface runs through the same avatar-critic Pass 7 quality gate the product ships to merchants for content review. The scaffold for this press release ran through --avatar case-study-prospect AND --avatar ai-agent-crawler before publish. Scorecards are committed at apps/dashboard/app/intelligence/ucpscore-100-self-applied-rubric/references/.

“We told merchants their stores must be legible to AI agents. Then we ran our own product against ourselves and verified our marketing surface met the same standard. The receipts are public — anyone can re-run the scan against ucpscore.ai with the same rubric SHA and reach the same number. That’s the difference between a benchmark and a receipt.”

— UCPScore Intelligence Desk

About UCPScore

UCPScore is the Inference Advantage Engine for ecommerce — the deterministic AI-buyer readiness rubric for stores selling into agentic commerce. Free initial scan, public rubric, SHA-pinned state-receipts. Validated against the UCP conformance suite + tracks all four foundational specs (UCP / ACP / MCP / AP2). Visit ucpscore.ai.

Receipts & verification

Want your own receipt? Run a scan against your store at ucpscore.ai. Free initial scan. Deterministic rubric. SHA-pinned receipt. No credit card. No email gate.