Internal
Academic style audit
Reads the rendered prose on every public route and flags three things: phrasing that belongs to a product page rather than a paper, sentences that read as mechanistic where the manuscript is descriptive, and pages that state headline results without carrying their qualifications.
Run the audit
Each route is rendered in an offscreen frame and its main region is stripped of tables, code, figures and chart fallbacks, so only prose is scanned. Mechanistic matches inside a hedged sentence are downgraded to a review item rather than dropped, because hedging can be doing too little work. The rules live in src/lib/style-rules.ts.
Rule set
| Rule | Severity | Editorial reason |
|---|---|---|
| Hype adverb | flag | Intensifiers overstate dispersion the paper reports numerically. |
| Product language | flag | This is a research artifact, not a product page. |
| Conversational filler | flag | Slogan and filler phrasing compresses qualified claims. |
| Second person address | review | Academic prose addresses the reader impersonally. |
| Contraction | review | Contractions read as informal in a scholarly register. |
| Unqualified superlative | review | Superlatives are claims the manuscript does not make. |
| Causal verb | flag | The decomposition is an accounting identity, not an identification strategy. |
| Explanatory claim | flag | The paper observes rankings, not the beliefs and expectations behind them. |
| Effect language | review | Reserve “effect” and “impact” for estimates the paper identifies. |
| Preferences asserted | review | A submitted list is informative about ordering, not a preference parameter. |