An overall style score plus a breakdown of the fifteen metrics that feed into it. The fastest way to see, at a glance, how the prose scores against a general-purpose readability and style baseline.
Endpoint: /score — up to 50,000 words per run.
What it measures
The /score endpoint computes fifteen individual style metrics for your manuscript. Each metric has a current value, an ideal range, and a status — either "good" (in range), "fair" (drifting), or "needs improvement" (out of range). The composite style score is a weighted aggregate of the fifteen, expressed as a 0–100 number with the same three-tone verdict attached.
The fifteen metrics are:
- Style score — the overall composite
- Sentence length — average and distribution
- Readability — grade-level estimate
- Sentence variety — how much length varies from sentence to sentence
- Glue index — see Glue index
- Passive voice — see Passive voice
- Business jargon — corporate-speak density
- Complex paragraphs — % of paragraphs above a difficulty threshold
- Conjunction starts — sentences beginning with "And", "But", "So"
- Slow pacing — density of static (non-action) sentences
- Very long sentences — count above a hard length ceiling
- Emotion tells — "he felt sad" / "she was angry" tell-not-show constructions
- -ing starts — sentences beginning with an -ing verb (a common draft tic)
- Weak adverbs — "very", "really", "just", "quite"
- Dialogue tags — count and diversity of said-tags
Each metric's ideal range is calibrated against a large corpus of published prose in the general fiction and mixed-nonfiction band. Ranges are absolute (not per-genre): "sentence length ideal 14–20 words" is the same target whether you're writing memoir or middle grade. That's a deliberate design choice — the report is a description of where your prose lands against the median, and you interpret that against your genre.
The /score endpoint caps at 50,000 words per run. Longer manuscripts are truncated with a warning; run per-chapter for longer books.
Why it's useful
The overall score is a fast sanity-check — a single glance tells you whether the prose is in a healthy range or whether it needs work in a specific area. If the ring is green and the "to fix" count is low, you can move on. If the ring is red and the "to fix" count is high, you have a targeted list of exactly which of the fifteen dimensions is dragging the score down. The breakdown then tells you where to look — you don't have to hunt for problems, the report names them.
The value is in the granularity. A writer who reads a "your prose scored a 62" verdict has no idea what to do with that. A writer who reads "your prose scored 62 because weak adverbs are 42 (ideal ≤20) and -ing starts are 31 (ideal ≤25)" has a concrete revision list: do a pass for weak adverbs and a pass for -ing starts, and the composite score will move on its own.
The metric family also lets you notice trade-offs. Cutting weak adverbs sometimes lengthens sentences (you replace "very tired" with a stronger verb phrase); cutting passive voice sometimes shifts your dialogue-tag distribution. The radar view is where those trade-offs become visible — you see which metrics move together when you revise.
Use this alongside Passive voice (which drills into the specific passive constructions with exact line locations) and Analyze (which gives you issue-by-issue rewrite suggestions on a 1,500-word window). Style score tells you what's wrong; the other two tell you where and how to fix it.
How to read it
The style-score data feeds two distinct areas of the Rust panel. First: the hero header at the top of the panel shows a colored ScoreRing (green when the status is "good", amber when "fair", red when "needs improvement"), a plain-English verdict word ("strong" / "fair" / "needs work"), the numeric style score in bold ink, and a red "N to fix" count for the number of metrics in the "needs improvement" bucket. This is the at-a-glance sanity check.
Second: the Style scores radar chart lower down in the panel plots all fifteen sub-metrics as a polygon on a 0–100 axis. Each metric axis is labelled around the perimeter (in short-form: "Style", "Sentence length", "Readability", "Sentence variety", "Glue index", "Passive voice", "Business jargon", "Complex paragraphs", "Conjunction starts", "Slow pacing", "Very long sentences", "Emotion tells", "-ing starts", "Weak adverbs", "Dialogue tags"). The polygon shape tells you the profile of your prose — a large well-formed polygon means every dimension is healthy, a polygon with sharp inward dents means specific metrics are struggling.
Hover any axis to see a tooltip with the metric's long name, the current value (mono, tabular), the ideal range, and the status word ("good" / "fair" / "needs improvement"). Below the radar is a legend row with three tiles — "good N", "fair N", "needs N" — coloured green/amber/red, that tally how many metrics fell into each bucket.
A metric with a green tone is fine; leave it alone. Amber means it's outside the ideal range but not dramatically. Red means it's meaningfully off — worth a targeted revision pass on that specific dimension.
Use the refresh icon (top-right of the panel header, RefreshCw icon) to force a fresh compute after major edits. The whole Rust panel re-runs all eight endpoints in parallel — you don't refresh per-endpoint.
When to ignore it
The ideal ranges are calibrated for general-purpose modern prose (mixed fiction + nonfiction, published in the last thirty years). Literary fiction, poetry-adjacent prose, and experimental writing will score amber or red on several metrics deliberately — long sentences, high passive voice, more -ing constructions, richer subordinate-clause structure are all features of that register, not defects. Use the numbers as descriptive, not prescriptive.
Genre skew is also load-bearing. Thrillers should skew short-sentence and low-adverb (green on those metrics); literary fiction should skew long-sentence and higher-glue (amber on those, and that's fine); children's books should skew low-readability-grade (green there, red on vocabulary richness). If your target genre is one that structurally violates a metric's default range, don't try to bring it back into range.
Dialogue-heavy scenes distort several metrics — dialogue is naturally shorter, more fragmented, higher in conjunction starts, and lower in glue than narration. A chapter that's 70% dialogue will produce a very different score from a chapter that's 90% narration, even in the same book. If you're scoring individual chapters, expect the numbers to swing.
Finally, remember: the score is a summary of what the software can measure. It cannot measure whether your prose is beautiful, whether your dialogue crackles, whether your reader will keep turning pages. A book with a green ring can still be bad. A book with a red ring can still be Ulysses.
How to run it
- Click Write in the left sidebar and open the document you want to analyze.
- Open the right rail: click the Reports button (bar-chart icon) in the editor's top toolbar.
- In the rail header, click the Rust tab.
- All 8 Rust endpoints run in parallel on tab open — typically completes in 2–5 seconds. There is no card to click; everything renders in one composite panel.
- The style score appears in two places: the coloured ring in the top-left of the header (with the plain-English verdict — "Good" / "Fair" / "Needs improvement" — plus the "N to fix" count next to it), and the full Style scores radar chart lower down, which plots all fifteen sub-metrics with their good/fair/needs pills.
- Use the refresh icon in the top-right of the header to force a fresh compute after major edits.
What to read next
- Summary report — a wider, deeper snapshot with the same numbers plus more
- Reports — back to the panel overview