Contents
Overview
What Prosemetry does
Most writing tools focus on correction. Prosemetry focuses on understanding. Instead of one quality score, it measures patterns in structure, language, information delivery, guidance, rhythm and style, then shows you exactly which paragraphs produce them.
Every part of the report helps answer three questions:
- Experience. What kind of reading experience does this document create?
- Location. Where do the important patterns occur?
- Action. What, if anything, deserves a closer look or a revision?
Prosemetry is calibrated for academic, professional, technical, administrative, policy and other expository prose. For fiction, poetry, creative nonfiction and experimental work, treat the results as descriptions rather than revision instructions.
The workflow at a glance
After you analyze a document, the dark bar at the top of every page walks you through five steps. Work through them in order: jumping straight to revision can hide the larger patterns that make individual passages meaningful.
| Step | Question | What you’ll find there |
|---|---|---|
| 1. Overview | What kind of document is this, and what stands out? | Writing Galaxy, Revision Priorities, Signal Attribution Maps |
| 2. Readers | Where might reading get hard, and why? | Reader Experience |
| 3. Weigh | Which passages are worth changing? | Guided Analytical Review |
| 4. Revise | How do I change them, in my own words? | The Workspace |
| 5. Compare | Did the revision do what I intended? | Compare |
Analyze a document
Start on the home page. Upload a Word document (.docx). Prosemetry uses Word’s styles to tell headings, notes and bibliographies apart from your body text, so other formats aren’t accepted yet. Analysis usually takes well under a minute; the progress page shows each stage and opens your Overview when it finishes.
- Prosemetry reads body paragraphs. Titles, headings, notes, endnotes and bibliographies are set aside so they don’t distort the measurements.
- Have a PDF or Google Doc? Open it in Word or Google Docs and save or download it as a .docx file.
- Longer is better for the maps. Paragraph-level evidence is most reliable with 20 or more body paragraphs.
Accounts and My analyses
During the private beta, create an account with your invite code to analyze a document. Your analyses belong to your account, so you can open them from any browser, and no one else can see them.
My analyses lists everything you’ve run, newest first, with links to each report, its Workspace draft and, for revisions, the comparison with the original. Only the account that created an analysis can open it.
Overview: see what stands out
Your report opens on the Overview. Don’t revise yet. First establish what kind of document Prosemetry found, what experience it tends to create, and which patterns stand out.
Open your OverviewWriting Identity
Writing Identity names the closest writing population: the group of reference documents the submitted document most resembles, with the variant it leans toward. The most distinctive pattern names the measures where it differs most from that population.
It describes this document, not the writer. It says nothing about intelligence, personality, ability or permanent style, and different tasks often produce different identities.
A highly formal, institutional profile may suit an accreditation report that needs consistency and credibility. The same profile in a public blog post may signal more formality than the audience needs.
Ask: is this profile right for what the document is trying to do?
Writing Galaxy
The Writing Galaxy is a map, not a ranking. Prosemetry compared 5,926 reference documents on 13 measures and found eight recurring populations of writing, each with two variants. The map shows those populations and marks the one your document most resembles, with its correspondence percentage. The flow diagram shows which measurements and behaviours place it there.
- Select a population or variant on the map, or in the list beside it, to read what it describes.
- Hover a measurement or behaviour in the flow diagram to trace where it leads.
- The map is an orientation aid. Correspondence comes from all 13 measures, not from distance on the map. Documents under 750 words of body text can’t be placed.
Revision Priorities
Revision Priorities are Prosemetry’s suggested order of investigation. Each one names a pattern, explains why it may be worth examining, and links to the map that holds the evidence. Priorities are not errors or required changes; the ranking shows where a closer look is likely to be most useful.
This priority doesn’t mean transitions are missing. It means transitional support may be unevenly spread. Open the linked map, read the highlighted paragraphs and their neighbours, and decide whether readers can follow the document’s movement.
Don’t just add transition words. Check that the relationships between ideas are clear.
Signal Attribution Maps
Signal Attribution Maps (SAMs) connect each document-level result to the paragraphs producing it. They are grouped by theme: Organization & Guidance, Information Structure, Style & Positioning, Vocabulary & Expression, Conceptual Development, and the three Reader Experience contributors.
Open Signal Attribution MapsWhat a map shows
Each cell is one body paragraph. A stronger colour means more of that signal in that paragraph, compared with the rest of the same document.
- 0 None
- 1 Very low
- 2 Low
- 3 Moderate
- 4 High
- 5 Very high
Colours are relative to this document and to this one signal. Level 5 in Reader Guidance can’t be compared with Level 5 in Information Density, and neither means better or worse writing.
Click a cell to pin that paragraph’s evidence beside the map while you read. Click another cell to switch.
Information is concentrated near the beginning. That may suit a technical briefing, or it may mean the opening carries too much at once for the intended reader.
Ask: is this concentration intentional, and does it help readers?
Next: open Readers to see what the document may ask of readers.
Readers: where might reading get hard?
Before looking for individual passages, read the document from the reader’s side. Reader Experience shows where reading may get harder or easier and why.
Reader Experience
Reader Experience estimates how much effort the document asks of readers as it unfolds, from beginning to end. The ribbon’s thickness shows combined effort; the streams inside it show three contributors:
- Sentence structure: how much readers must hold in mind to connect the parts of a sentence.
- Word predictability: how expected the wording is in context.
- Connections between ideas: how smoothly each sentence continues from its neighbours.
Read the shape first and the details second. Higher demand is not automatically a defect: a methods section may need to be harder than a conclusion.
Open ReadersKeep measurement and meaning apart
Every finding in Prosemetry has three layers.
| Layer | What it does | Example |
|---|---|---|
| Observation | Reports what was measured. | Reader guidance is concentrated at section transitions. |
| Interpretation | Suggests a possible effect on readers. | Readers get strong orientation at major turns but less within sections. |
| Revision guidance | Offers an option to consider. | Check whether lower-guidance sections make their purpose clear. |
None of these layers tells you what you must do. Meaning depends on audience, genre, discipline and purpose.
Next: go to Weigh and decide which passages are worth changing.
Weigh: decide what’s worth changing
Once you know where the patterns are and what they may mean for readers, decide what to do about the passages that matter most.
Guided Analytical Review
Guided Analytical Review walks you through the passages where a pattern stands out and the Workspace’s sentence-by-sentence check also finds something to look at, one at a time. For each passage it explains what Prosemetry observed, what readers may encounter, how that interacts with purpose, and where it occurs. Then you decide:
Your decisions are saved with the analysis, so they’re still there on another visit or another device when you’re signed in. They also appear beside the matching paragraphs in the Workspace.
Open WeighNext: take the passages you chose to revise into the Workspace.
Revise: change the document for a reason
Before changing a passage, read its evidence, the paragraph itself, and the paragraphs before and after it. Name the effect you want readers to experience, then make one or two focused changes.
Don’t try to maximise or minimise every metric, remove every intense region, or make every map uniform. Variation is often necessary. The goal is not a balanced chart; it is a document that better serves its readers.
The Workspace
The Workspace is where you revise. Your document is on one side and Prosemetry’s findings for the selected paragraph are on the other, including any Guided Analytical Review decisions.
- Edit paragraph by paragraph. Changed paragraphs are marked, and you can compare each with its original or restore it at any time.
- Save your document. Drafts are saved with the analysis, so you can come back later. Ctrl or Cmd+S saves.
- Export the revised draft as .docx or .txt, or a revision report listing each changed paragraph with its comparison, decision, note and linked findings.
- Analyze as Draft 2 (then Draft 3, and so on) runs a new analysis of your saved edits and takes you to the comparison when it finishes. The draft you edited stays exactly as it was.
Compare: did the revision do what you intended?
Choose Analyze as Draft 2 in the Workspace. When the new analysis finishes, the Compare page shows the two drafts side by side:
- At a glance: how many review points, priorities and placements changed.
- Revision Priorities before and after, each marked as no longer a top priority, new, or still a priority.
- Changed paragraphs with your edits highlighted word by word, plus how each paragraph’s reader demand moved.
- Measures for the whole document, before and after.
The revision gets its own full report and appears in My analyses, linked to the original.
Revising is a loop. Basic keeps each draft for a week, Pro for 30 days.
- Read the Overview and Readers to see what a reader meets.
- Weigh which patterns are worth changing.
- Revise in the Workspace, in your own words.
- Analyze your saved edits. They become your next draft (Draft 2, Draft 3, …).
- Compare to see what changed.
At the end of every comparison, choose Revise again to keep working on the new draft, or I’m done to download it as your final version with the Complete Analysis Report. The dark bar at the top of each page lists your drafts, so you can open any of them. From Draft 3 on, Compare can show the change since the draft just before or since your first draft.
Look at the passages you meant to change before the wider numbers:
- Did the targeted paragraph become clearer?
- Did the intended reader experience improve?
- Did the change create a new problem nearby?
- Did the Writing Identity shift unexpectedly?
- Does the revised version simply read better?
OriginalThe evidence demonstrates several important trends. The next section discusses additional findings.
RevisedThe evidence demonstrates several important trends. To understand how those trends influence institutional outcomes, the next section examines additional findings.
The revision makes the relationship between sections clear. Whether a metric moves a lot or a little, the real question is whether readers can follow the connection.
Next: judge the passages you changed on purpose. Treat wider score movement as context, not as the definition of success.
Evaluation
What success looks like
A good Prosemetry session usually looks like this:
- You identify one or two meaningful patterns.
- You investigate the evidence behind them.
- You make a small number of focused revisions.
- The revised document serves its audience and purpose better.
- You understand more clearly how the document works.
Success doesn’t require improving every metric, eliminating every intense passage, or moving toward a particular Writing Identity or population.
Boundaries
What Prosemetry can’t tell you
Prosemetry measures patterns. It cannot determine:
- whether a document is objectively good or bad, or must be revised;
- who or what wrote a document;
- whether one Writing Identity or population is better than another;
- whether a metric should always go up or down;
- whether a revision improves factual accuracy or the validity of an argument;
- whether every reader will experience a passage the same way.
Prosemetry provides evidence, context and possible interpretations. You decide what, if anything, to do.
Plain-language dictionary
Terms and concepts
You don’t need specialist knowledge to use the report. Each term below explains what is measured, what higher and lower levels may suggest, and the most useful question to ask. No measure is automatically good or bad; its meaning depends on audience, purpose, genre and where it occurs.
Navigation and reader support
Reader GuidanceHelp for following the document
Measures: language that tells readers where the document is going, how ideas relate, and why a passage appears where it does.
Higher may mean more signposting and statements of purpose. Lower may mean a more implicit structure that relies on readers to infer direction.
Ask: can readers tell what this passage is for and how it connects to what came before? Too much guidance can feel repetitive.
TransitionsConnections between ideas
Measures: words, phrases and structures that signal continuation, contrast, cause, sequence, example or a change of direction.
Higher may mean explicitly signalled movement. Lower may mean implicit connections or abrupt shifts.
Ask: is the logical relationship between these ideas clear? How often transitions appear and how well they work are different things.
Transition RelianceDependence on explicit connectors
Measures: how heavily the document leans on recognisable transition language to organise its movement.
Higher may mean strong explicit organisation or repeated stock connectors. Lower may mean more varied ways of connecting, or under-signalling.
Ask: do the transitions clarify real relationships, or carry relationships the prose hasn’t explained?
Structural RegularityPredictability of form
Measures: how consistently the writing repeats similar sentence and paragraph structures.
Higher may mean stable, procedural organisation. Lower may mean more variety or some unevenness.
Ask: is the regularity helping readers, or making the writing feel mechanical?
Meaning and information
Concentrated MeaningTechnical term: lexical density
Measures: the balance between content words (nouns, verbs, adjectives, many adverbs) and the grammatical words that connect them.
Higher may mean more information packed into fewer words, more precision or abstraction. Lower may mean more conversational or gradually paced language.
Ask: does this passage give readers enough time and support to absorb what it carries? Try an example, splitting competing ideas, or explaining an abstract term, or keep the concentration where precision needs it.
Concept-Heavy LanguageTechnical term: nominal density
Measures: how heavily a passage relies on nouns and noun-based concepts, especially abstract ones such as implementation, evaluation or responsibility.
Higher may mean more abstract, academic, legal or technical writing. Lower may mean more direct attention to people and actions.
Ask: can readers see the people, actions and consequences behind these concepts? “The implementation of the assessment produced an improvement in compliance” can become “The agency implemented the assessment, and more organisations complied.”
Conceptual Trait DensityConcentration of conceptual development
Measures: how densely a passage develops abstract concepts and the relationships between them.
Higher may mean sustained conceptual reasoning. Lower may mean more concrete, narrative or procedural prose.
Ask: are concepts introduced at a pace readers can follow?
Information DensityMaterial competing for attention
Measures: how much substantive material a passage presents relative to its length and structure.
Higher may mean several claims, qualifications or examples in a small space. Lower may mean more explanation, pacing and room for emphasis.
Ask: how many new things must readers understand at once? This is related to Concentrated Meaning but not the same measure.
Vocabulary RichnessVariety in word use
Measures: the range and distribution of vocabulary across the document.
Higher may mean varied word choice. Lower may mean stable terminology or a narrower range of expression.
Ask: is repeated wording helping readers track key concepts, or making the prose feel limited?
Lexical SophisticationSpecialised or uncommon vocabulary
Measures: how much a passage uses less common, specialised or advanced words.
Higher may mean disciplinary precision for expert readers. Lower may mean familiar, broadly accessible vocabulary.
Ask: does the intended audience know these words, and are they the most precise ones?
Sentences and reading demand
Average Sentence LengthTypical sentence size
Measures: the average number of words per sentence.
Higher may mean more developed or layered sentences. Lower may mean faster pacing, or choppiness when sustained.
Ask: does each sentence give readers a manageable amount? The average hides variation.
Sentence ComplexityRelationships inside sentences
Measures: embedded clauses, qualifications and layered relationships within sentences.
Higher may mean nuanced reasoning and more processing demand. Lower may mean more direct construction.
Ask: can readers find the main claim before they process its qualifications?
Sentence-Length VariationChange in sentence rhythm
Measures: how much sentence length varies across a document or passage.
Higher may mean rhythmic variety. Lower may mean steady pacing, or monotony.
Ask: does the rhythm support emphasis and keep readers oriented?
Structural Integration DemandReader Experience: sentence structure
Measures: how much readers must hold in mind to connect the parts of a sentence, using dependency distance, dependency depth and embedded structure.
Ask: where structure is demanding, is the content important enough to justify it?
Lexical UnpredictabilityReader Experience: word predictability
Measures: how expected the wording is in context, estimated with a language model (surprisal). Unexpected words take more effort to process.
Ask: are unexpected words doing useful work, such as introducing a precise term, or are they making the passage harder than it needs to be?
Cohesion DisruptionReader Experience: connections between ideas
Measures: breaks in continuity between neighbouring sentences, using semantic continuity, referential overlap, boundary disruption and transition confidence.
Ask: can readers see how each sentence follows from the one before?
Readability DemandFlesch-Kincaid Grade and Gunning Fog Index
Measures: estimated reading difficulty from sentence length and word length or syllables. Flesch-Kincaid gives an approximate U.S. grade level; Gunning Fog weighs words of three or more syllables.
These are comparative indicators, not literal statements about who can read the document. Necessary technical terms can raise them even in well-designed writing, and they say nothing about accuracy or argument quality.
Ask: is the reading demand appropriate for the intended audience?
Reader BurdenCombined reader effort
Measures: how several features work together to raise processing demand: long or complex sentences, concentrated meaning, dense information, abstract language, limited guidance and specialised vocabulary.
Ask: is the effort this passage asks for justified by its importance? Reader Burden is not a measure of bad writing.
Voice, stance and formality
Formal RegisterProfessional or institutional formality
Measures: language associated with formal academic, professional, technical or institutional communication.
Higher may mean impersonal phrasing and distance from conversation. Lower may mean more personal or direct language.
Ask: does the formality match the audience, relationship and occasion?
Passive Construction DensityFrequency of passive structures
Measures: how often the recipient of an action becomes the focus, as in “The report was approved.”
Higher may mean emphasis on procedures or outcomes rather than actors. Lower may mean more direct naming of who did what.
Ask: is responsibility clear where it matters? Passive voice is not inherently weak.
QualificationLimits placed on claims
Measures: language marking probability, conditions, uncertainty, scope or exceptions, such as may, generally or the evidence suggests.
Higher may mean careful attention to uncertainty, or hesitation. Lower may mean direct claims, or overstatement.
Ask: is the level of certainty appropriate to the evidence?
Conventional LanguageFamiliar genre phrasing
Measures: reliance on standard expressions common in academic, professional or technical writing.
Ask: is the conventional wording making the document easier to recognise and use, or just making it sound generic?
Formulaic PressurePredictable writing formulas
Measures: how much language and structure rely on recurring templates or highly predictable patterns.
Ask: is the pattern providing useful consistency, or flattening the document’s voice?
Repetition and emphasis
RepetitionReuse of words or phrases
Measures: how often language recurs across the document.
Higher may mean emphasis and continuity, or redundancy. Lower may mean varied language, or inconsistent names for key concepts.
Ask: is the repeated language helping readers remember and connect ideas?
Repeated Phrase PressureConcentration of recurring phrases
Measures: whether particular multi-word expressions recur often enough to shape the document’s style or rhythm.
Ask: does this phrase need to stay fixed, or could some repetitions change without losing precision?
Profiles, evidence and review
Writing IdentityWhere this document fits
A document-level description of where the text fits: its closest writing population in the Prosemetry reference archive, with the pattern that most sets it apart.
It is not a grade, a diagnosis, a permanent label for the writer, or a style to imitate.
Writing GalaxyA map of writing styles
A map of eight populations of writing found in 5,926 reference documents, showing the one this document most resembles and which measurements put it there.
The Galaxy is not a ladder. No population is better than another.
Signal Attribution Map (SAM)Paragraph-level evidence
A map showing where a pattern appears and how strongly each paragraph contributes to the document-level result.
A SAM tells you where to look. It doesn’t tell you what to change.
Intensity level0 None to 5 Very high
The relative strength of a signal in one paragraph. Intensity is relative to the document and specific to one measure, so levels can’t be compared across measures.
Contribution and top contributorWhich paragraphs drive a result
Contribution is how much a paragraph helps produce the document-level result. A top contributor is one of the paragraphs most responsible for it.
A top contributor is not automatically the worst paragraph or the one most in need of revision.
Revision PriorityA ranked opportunity to investigate
A suggested place to look first, based on the analysis and its evidence. It is not an error, an automatic correction, a required change or a deduction.
Review decisionKeep it, Revise it or Doesn’t apply here
Your judgement on a passage in Guided Analytical Review. Decisions are saved with the analysis and shown in the Workspace.
Observation, interpretation and revision guidanceThe three layers of a finding
Observation is what was measured. Interpretation is what it may mean for readers. Revision guidance is an optional action to consider after reading the evidence.
Revision and comparisonAnalyzing a revised draft
A revision is a new analysis of your Workspace draft, linked to the original. The Compare page shows what changed between them.
No terms match that search.
Avoid
Common mistakes
- Treating Writing Identity as a permanent label.
- Treating Signal Attribution Maps as error maps.
- Comparing intensity levels across different measures.
- Revising only to move a score.
- Reading highlighted paragraphs without their surrounding context.
- Making broad changes before working through Revision Priorities.
- Assuming every high-intensity passage should be reduced.
Workflow summary
Checklist
- Read the Overview and Writing Identity.
- Look at the Writing Galaxy, then Readers, and ask whether they fit the document’s purpose.
- Read the highest-ranked Revision Priority.
- Open its Signal Attribution Map and pin the strongest paragraph.
- Read the evidence together with the surrounding paragraphs.
- Work through Weigh and record a decision for each passage.
- Make one or two purposeful revisions in the Workspace and save the draft.
- Choose Analyze as Draft 2 (the button names the next draft), then check the passages you changed on the Compare page.
Teaching and coaching
For instructors and writing center tutors
In classrooms, writing centers and coaching sessions, Prosemetry works best as a starting point for conversation about rhetoric, not as an automated grader.
- Before the session: the writer reads the Overview and Readers, then picks one or two passages to discuss.
- During the session: writer and tutor read the evidence together, talk about audience and purpose, and record decisions in Weigh.
- After the session: the writer revises selectively in the Workspace, analyzes the revision, and compares the targeted passages.
Prosemetry can find patterns and provide evidence. The conversation about meaning, audience and purpose still belongs to people.
Write like you, only better.
Understand the document. Weigh the evidence. Revise with purpose.
Start a new analysis