Content Repetition: Vocabulary Variety for AI Trust

What it measures

This metric measures how much new wording your page brings as it goes. It reads the share of distinct words inside every rolling 100-word window and averages the result — a moving-average type-token ratio.

The rolling window matters. A plain "unique words ÷ total words" ratio falls off steeply as a page gets longer: a 60-word stub scores about 85% and a 3,000-word article about 25%, on writing of identical quality. Scored directly, that ratio would pass every thin page and fail every real article. The rolling window removes the length effect, so a long guide and a short one are judged on the same scale.

On top of that, sentences repeated near-verbatim elsewhere on the page subtract from the score, up to 4 points. Pages under 100 words are too short for a rolling window, so they are compared against the unique-word share pages of that length normally reach; under 40 words the metric stays neutral.

Why it matters for AI

AI systems use vocabulary variety as a quality signal to distinguish human-quality content from thin, repetitive pages. Content that keeps adding new language is perceived as more natural and authoritative. Heavily recycled wording often indicates keyword stuffing, boilerplate templates or auto-generated content, which AI deprioritises.

Varied wording	Score	What it signals
< 55%	Low	Heavily recycled wording — reads as templated or auto-generated
55–70%	Average	Typical for directory, category and listing pages
≥ 70%	Good	Varied wording — natural, original writing

How to improve

1. Use synonyms — Avoid repeating the same terms; vary your word choices
2. Write naturally — Focus on communicating ideas rather than optimising for keyword density
3. Vary sentence structure — Mix short punchy sentences with longer explanatory ones