What the tool detects
Hook: openings with a number perform 35% better (n = 1.18M posts); those starting with a question, 34% worse. If your first 140 characters contain no number, no named entity and no tension, the tool flags it: that is all a mobile reader sees before "See more".
AI phrasing: "Here's how…" and "It's not X, it's Y" are associated with measured visibility drops (−9.7% and −9.2% on 78,281 posts). These are risk indicators, not bans — the tool never says LinkedIn "will penalize" your post.
Structure: more than two hashtags triggers an alert (the data is non-convergent — documented); an external link in the body costs about 11% of measured impressions.
Readability: a style indicator, computed with empirical per-language thresholds (75th percentile of professional English writing: 29 words per sentence). It informs you; it predicts nothing.
What the tool does not do
It does not predict performance. No text alone can: author, audience and timing weigh more than form. The displayed score is a heuristic index aggregating detected rules — not a grade on your future.
Privacy
The analysis runs entirely in your browser. No text is sent, stored or shared with anyone.