Methodology · v2.0
What we measure: the degree to which a website is structured to allow autonomous AI agents to discover, understand, and transact with it.
The NOD Score is a formative composite index measuring 7 capability dimensions across 71 automated checks. Like a credit score for AI agent readiness.
Grade Tiers
Agent Optimized is a qualified tier, not a score threshold. Reaching 80+ is necessary; meeting a multi-criterion qualification gate (valid NOD Protocol manifest, no critical access failures, category minimums) is required. Sites that score 80+ but fail the gate are reported as Ready with a cap at 79.9 and a gating flag.
Weights reflect construct importance — what an agent needs to successfully interact with a site — not empirical discrimination. In a formative composite, weights are about what matters, not about what correlates.
How the site identifies itself to machines — whether agents can recognize what kind of entity they're interacting with, what it offers, and how it relates to the rest of the web.
How agents find site content — whether the site's structure and directories surface what's there instead of hiding it behind interactive navigation.
How agents extract meaning from pages — whether content is readable without executing a browser, and whether structure communicates hierarchy and relationships.
How agents interact and transact with the site. The combined weight splits between programmatic interaction (APIs, docs, endpoints) and transaction readiness (purchase, book, apply) — the split varies by business model.
Whether agents can reach the site at all — not blocked, not throttled, not hidden behind bot-detection that treats every non-browser request as hostile.
Whether the site publishes a NOD Protocol manifest declaring its agent-facing contract. Optional for a baseline score; required for Agent Optimized.
Sites are scored against the checks that are appropriate for their business model. A SaaS product isn’t penalized for lacking a shopping cart. An e-commerce store isn’t penalized for lacking API documentation.
Every site is classified into one of four primary archetypes. Classification determines which checks apply and how the Programmatic Access / Transaction Enablement weight splits.
Every site receives two scores. Both are 0–100.
The headline score. Weighs all seven dimensions including whether the site publishes a NOD Protocol manifest.
The anchor for longitudinal and cross-site comparison. Protocol adoption is a contract the site chooses to sign; the Base Score measures capability independent of that choice. A site that’s excellent everywhere except protocol gets a Base score near its ceiling.
Sites are benchmarked against their industry peer group. A site’s sector percentile shows where it ranks within its cohort — a 60 is stronger in a sector averaging 25 than in one averaging 45. Percentile is reported when the cohort has enough sites to make the ranking meaningful; otherwise the site falls back to its parent-group benchmark.
Across 10,000 sites. Scores computed from two random halves of the applicable check set correlate at 0.80 — the scale is internally consistent.
Study in progress. 500 sites rescanned at 24-hour and 7-day intervals. Published in the next methodology revision.
Every check returns a normalized result. Checks are aggregated equally within a dimension, dimensions are aggregated by weight. Dimensions that don’t apply to a site’s archetype are excluded, and the remaining weights are rescaled so the score remains on a 0–100 scale.
This is v2.0. A formative framing, launched alongside a continuity anchor.
v2.0 reframes NOD as a formative composite, adds business-model archetypes, and introduces dual scoring. For 90 days, every site shows both its v1.1 and v2.0 scores with the conversion documented. Read the migration notes.
75 checks. 7 categories. Your score in 60 seconds.