Understand reports
The five editorial scores, what each one measures, how to read the bands, and why context still matters more than the number.
Working card by card tells you about one sentence at a time. Reports steps back and scores the document as a whole, the same way the Word add-in does. It is computed from the same analysis that produced your suggestions — there is no separate report-generation step, and it does not spend credits.
Reports needs enough text to say anything meaningful: below roughly 500 characters, the panel shows “Process more suggestions to calculate your scores” rather than a number it cannot stand behind.
The five scores
Each score runs 0–100, computed from a different slice of the writing.
Reader Engagement
Synthesizes every other dimension into one overall measure. Higher scores balance clarity, concision, flow, tone, and word choice to keep a reader oriented and interested.
Concise and Readable
Sentence length, word length, and sentence variation. Higher scores trim unnecessary padding and vary rhythm without oversimplifying.
Flowing and Cohesive
How smoothly ideas connect across sentences and paragraphs. Higher scores reflect effective transitions, logical sequencing, and clean modifier placement.
Crisp and Punchy
Word choice. Higher scores favor concrete verbs and direct phrasing over abstract, inflated, or cumbersome language.
Clear and Direct
How plainly the document communicates. Higher scores reflect a modern tone and straightforward phrasing, restrained on legalese, without sacrificing precision or authority.
Reading the number
Scores run 0–100, and the product’s own guidance is direct about what each band means:
| Score | Interpretation |
|---|---|
| 85–100 | Excellent — exemplifies best practices in this category |
| 75–84 | Strong — above average, with room for polish |
| 65–74 | Average — typical of competent legal writing |
| 50–64 | Below average — worth a targeted revision |
| 0–49 | Needs attention — significant room for improvement |
The average document lands around 70. A 65 is not a failing grade; it is the middle of the pack.
Context the score cannot see
The product’s own caveat is worth repeating in full, because it is easy to over-read a single number:
No algorithm can fully capture what makes writing effective for a particular audience and purpose.
Specifically:
- Document type — a scholarly article may legitimately score lower on Crisp and Punchy than a motion in limine. Different genres reward different things.
- Subject matter — complex regulatory topics may need longer sentences and more hedging than a straightforward factual dispute.
- Audience — writing for specialists may justify technical terms that would hurt the score but serve the reader.
- Strategic choices — passive voice or a qualifier sometimes serves a legitimate purpose. The score does not know your strategy.
Treat these scores as one input among several — a prompt to look again at a passage, not a verdict to defer to.
Readability
Below the five scores, three concrete metrics with their own pass/fail-style flag:
- Average characters per word
- Median words per sentence
- Passive voice index — how many sentences use the passive voice
Each metric shows its own score and a short message; a metric flagged high is worth a look even if the overall Reader Engagement score is fine, since a strong average can still hide one dense paragraph dragging the passive-voice count up.
Strengths and opportunities
The final section is plain language rather than a number: what the document is already doing well, and where it has room to improve, generated from the same analysis pass. Read this section first if you want the takeaway before the metrics that support it.
Sharing a report
Copy summary puts a short plain-text digest on your clipboard — suggestions considered, accepted, ignored, and how many citation suggestions are outstanding — ready to paste into an email or a note. An email draft option builds the same summary into a message so you do not have to.
Refresh the report after editing the document if real-time updates are off; otherwise the numbers describe the text as it stood at the last analysis, not the text on screen now.
RealityCheck has its own, separate export — a PDF with a verdict per citation. See Use RealityCheck.