The scenario
You are an editor who has inherited a site with a hundred and twenty pages and an editorial standard nobody has time to apply by hand.
Northmoor has three published guidelines — voice and tone, structure and grammar, inclusive language. They are held as context items, which means the same three documents that tell a writer what to do also tell the review what to check. That is the design: the guidelines and the evaluation criteria are one artifact seen from two directions.
Try this
Run content review on each of the three pages below. Each was written to fail one guideline while passing the other two.
1. Voice and tone
Run the review on the news item "Fall open day draws the largest attendance on record."
Voice here is scored on the four Nielsen Norman Group dimensions, with a target per surface. News should sit at 2 on matter-of-fact to enthusiastic, with a hard ceiling of 3. This page is a 5. It carries seven exclamation marks against a guideline that forbids them, and it reads as marketing copy on a site where nothing else does.
What you should see. A finding on the enthusiasm dimension with a score, and the exclamation marks cited specifically. The twenty other news items should pass.
2. Structure and grammar
Run the review on the program page for Environmental Science.
Two failures, both mechanical. Headings are in Title Case throughout, against a sentence case rule applied everywhere else on the site. And the second paragraph under the first heading contains a single sentence of about ninety words that should be three or four.
What you should see. The heading case flagged consistently, and the run-on identified by location rather than described in general terms. The other eleven program pages should pass.
3. Inclusive language
Run the review on the faculty biography for Dr. Miriam Frost.
The opening sentence explains her arrival at the university through her husband's job. This is the mild case on purpose. It is not offensive, it is the kind of thing a well-meaning writer produces without noticing, and it defines a professor by a relationship rather than by her work. The other nineteen biographies do not do this.
What you should see. The opening sentence flagged, with the reason stated as defining a person relationally rather than by their own role.
Then run it on the whole site
Three findings across roughly thirty target pages, rather than findings everywhere. That ratio is deliberate — the corpus was generated clean and then three pages were degraded, so the demonstration reads as "three pages need attention" rather than as a triage exercise.
Why the guidelines are the criteria
A prose style guide is not scoreable. Putting each trait on a spectrum with worked examples at either end makes the rule legible to a model and, critically, checkable. That is why the review can cite a number rather than an opinion.