THE ENGINE
Any renderable URL becomes a structured verdict.
The page is rendered in full, put through 311 checks across 27 categories, scored on 7 dimensions, and returned as one structured verdict — the same engine behind the API and the white-label reports.
WHAT THE ENGINE DOES
URL in. Structured verdict out.
One request. The page is rendered, then 311 checks fire across 27 categories. Structured JSON returns.
Any live page the engine can render — homepage, pricing, product, landing. No tag, no SDK, no access to your code.
7 dimension scores · ranked findings · verified strengths.
Universal checks run on every page. Vertical checks fire only when the site type calls for them — so every finding is relevant to what is actually being scanned.
THE SEVEN DIMENSIONS
Seven dimensions. One score.
Every check maps to one of seven scored dimensions. Each is returned as a 0–100 sub-score under its exact contract key, so a verdict is never a single opaque number.
How the page structures the path from arrival to action — hero, hierarchy, and where the CTA sits in the flow.
Whether proof is present and credible — outcome-specific testimonials, named sources, contextualized logos, not decoration.
Whether a cold visitor grasps what this is and who it is for in seconds — headline, subhead, and body pulling one direction.
Whether the page is prepared to convert the traffic pointed at it — intent match and continuity from the click to the page.
The mechanics that gate conversion before copy matters — render, speed, and mobile behavior that quietly make or break trust.
Whether the page anticipates and neutralizes buyer doubt — risk reversal near the CTA, FAQs, and clear ICP fit.
Whether the offer reads as a complete unit — what it does, who it is for, what it costs, and the differentiator, stated not implied.
Keys are stable across every scan and every plan — build against them once.
TWO-PASS RECONCILIATION
One pass is not trusted. So the engine runs two.
A single pass of a language model against a rubric wobbles. The engine runs the full rubric twice and reconciles the passes. Per-pass variance is about ±3 points, so the coverage score is published as a tight band rather than a false-precision number.
BENCHMARKING
Scored against an industry corpus.
A bare score is hard to act on. Each dimension is positioned against an industry corpus, so a reading comes back as above or below the norm for that vertical — not just a number floating in space.
The framing is deliberately honest: benchmarks sharpen as corpus volume grows. The engine returns where a page sits relative to its vertical, not a manufactured percentile.
Positioned against an industry corpus. Benchmarks sharpen as corpus volume grows.
WHY THE FULL RUBRIC
The full rubric is what makes the score reproducible.
It is tempting to run a handful of checks and call it a score. But a subset drifts — the same page lands in a different band from one run to the next, and the findings point in inconsistent directions. Running the complete rubric is the whole point: it is what makes a coverage score reproducible, and reproducibility is what you can actually build on.
Drifts. Directionally inconsistent.
A subset of checks gives a different band on re-run and disagrees with itself on what matters. You cannot track a page against a number that moves on its own.
Reproducible. The same URL lands in the same band.
Every check runs, every scan. Reconciled across two passes and published as a band, the score holds still — so a change in the score means a change in the page, not the method.
TWO OUTPUTS, ONE ENGINE
The same engine ships two ways.
Same 311 checks. Same 7 dimensions. Same reconciled score. What changes is the surface it comes out of.
OUTPUT · API
Structured JSON.
POST a URL, get the verdict back as one structured object — score, 7 dimension sub-scores, ranked findings with on-page evidence, and drop-in copy rewrites. The shape never changes, so you build against it once.
OUTPUT · DASHBOARD
A white-label scored report.
The same verdict, rendered as a client-ready report — the score, what is working, the ranked fixes, and how the page benchmarks. Branded as yours, no code required. This is the door agencies use.