Grade your copy the way the machine reads it.
Paste a draft. A homepage, a new page, a paragraph you're about to ship. See where a synthesis engine resolves you soft, before you publish.
What this checks
An AI engine does not read your page the way a person does. It pulls sentences out and tries to use them somewhere else, on their own, with no page around them. Five things decide whether a sentence survives that. The checker scores each one and shows you the lines that cost you.
Does the sentence stand on its own
Lift one line out of your page and drop it into an answer with no context. If it still makes sense and still says who it is about, a model can use it. If it needs the sentence before it, it cannot.
How much of it is hedging
Words like leading, innovative, world class and industry leading carry no fact a model can check, so a sentence built out of them reads as no answer at all. The checker counts them and shows you the density.
How hard the reader has to work
Soft connectives, clauses that double back, and bridges that carry no new information all slow a reader down and give a model nothing to lift. This is the part most drafts fail on.
Whether the answer comes before the explanation
Put the claim in the first words and the explanation after it. A paragraph that builds to its point gets cut before the point arrives.
Whether the claim is backed by something checkable
A superlative with a number, a date and a source behind it is a claim. The same superlative on its own is a decoration, and a model treats it as one.
These five are the Authorship pillar of the Stage30 methodology, and the same five the paid Resolution Audit scores across every page on a site rather than one paragraph. The vocabulary is defined in the lexicon.