Could a competitor publish this page unchanged?
Paste one URL. A language model reads the visible copy on it and scores 40 checks: whether the writing names a real pain, earns its authority, reads as honest, and could only have come from you.
This is not a technical SEO check. Nothing here looks at schema, headers, speed or links. It reads the words a buyer reads, and it tells you which of them any competitor could have written.
NO ACCOUNT · NO EMAIL · 3 RUNS PER HOUR PER ADDRESS · OUTPUT IS MODEL-GENERATED, NOT A MEASUREMENT
The rubric: 40 checks, four dimensions, 80 points.
Every check is scored pass, partial or fail, worth 2, 1 or 0. Ten checks per dimension, 20 points per dimension, 80 in total. The bands are fixed in advance: 70 to 80 reads as strong copy, 55 to 69 as a solid foundation, 40 to 54 as formulaic, and under 40 as template content.
The rubric is published rather than proprietary, and it is the whole of the instrument. There is no weighting, no adjustment and no hidden composite behind the number. If you disagree with a check, you can see exactly which one moved the score and by how much.
A model reading a page is a judgement, not a measurement.
Everything this tool returns was written by a language model that read your copy. That is an opinion held to a published rubric, which is worth having, and it is a different kind of thing from a measurement. Caul, the instrument behind our paid work, runs a fixed prompt set against engines and counts what came back. This does none of that. Five limits follow, and they hold on every run.
It cannot see whether any engine cited the page.
This reads copy. It does not run a prompt, does not query an engine and does not observe an answer. A score here is not evidence that ChatGPT, Gemini or AI Overviews named you, and it is not evidence that they did not.
It cannot see what your buyer actually is.
The rubric asks whether the copy is calibrated to one clear buyer level. The model infers that level from the page itself. If the page is aimed at the wrong person, a confident page aimed confidently at the wrong person can still score well.
It cannot see anything JavaScript renders after load.
The page is fetched once and read as served. Copy injected by a client-side framework, hidden behind a tab or loaded on scroll is absent from the sample, and absent copy is scored as missing copy.
It cannot be rerun for the same number twice.
Two runs on an unchanged page can differ by a few points. That is what a model-generated assessment is. Read the verdict band and the failed checks, and treat the digit as approximate.
It cannot tell you what the change is worth.
No check here connects to traffic, pipeline or revenue. It says the copy is formulaic; it cannot say what fixing that returns, and anyone quoting a return from a rubric score is inventing one.
The disclosure, on the page rather than in a footnote.
- instrument
- A general-purpose language model, prompted with a fixed 40-check rubric. Not Caul, and not a measurement instrument.
- what was measured
- Nothing is measured. The visible copy of one page is read and scored against the rubric, and the result is the model's judgement of that copy.
- how
- The page is fetched once as served, its visible text, headings and any FAQ blocks are extracted, and that sample plus the rubric goes to the model in a single call. The returned scores are validated against a fixed schema before anything renders.
- over what window
- The moment you press run. There is no history, nothing is stored, and a second run tomorrow is a separate reading rather than a trend.
- what this cannot tell you
- It cannot see whether an engine cited the page, cannot see copy that JavaScript renders after load, cannot know who your buyer actually is, and cannot return the same number twice on an unchanged page. It says nothing about traffic, pipeline or revenue.
For the measurement, the audit is free.
This tool reads your copy. The audit runs a fixed prompt set against the engines and counts how often you are named, against competitors you name, delivered on a call so the findings get explained. That is the measurement, and this is not.
Free. The $1,500 is what this sweep is priced at when it is sold, not a survey of anyone else’s rates.