Academic Voice
Measure how you write. Then see where a draft drifts.
Give Syft a few pieces of your own work. It measures the habits it can count — sentence length, hedging, formality, first person — and reports where a new draft reads differently, dimension by dimension, with your reading and the draft’s side by side.
A comparison, not an identity check · You choose every sample
Similarity to your samplesLow
2 samples · 16 dimensions
- Longer sentences than your typical range
- Yours · Typically 6–23 words · avg 13.3Draft · Avg 18.8 words · 9 sentences
- Differs
- More assertive than your usual writing
- Yours · Certainty markers in 0% of sentencesDraft · Certainty markers in 11% of sentences
- Differs
- Less first person than your samples use
- Yours · First person in 57% of sentencesDraft · First person in 11% of sentences
- Differs
- More formal than your usual register
- Yours · Moderately formalDraft · Formal
- Differs
- Harder to read than your usual writing
- Yours · Grade level 4.8Draft · Grade level 11
- Differs
- Heavier vocabulary than usual
- Yours · 1.27 syllables per wordDraft · 1.68 syllables per word
- Differs
And 10 more measured dimensions. A stated rule over the measured differences — not a model score, and never evidence of who wrote the text.
The profile
Built from work you deliberately choose.
Add essays, chapters or reports you wrote and are happy to be measured against. Syft reads them for recurring tendencies and tells you when it has enough text to say anything useful. Nothing is inferred from documents you did not add.
The profile lives with your account and is used by this lens only. You can see what it measured, add to it, or delete it.
What it measures
- Sentence length
- Average length, and how much it varies from sentence to sentence.
- Hedging
- How often a claim is softened — may, might, appears, tends to.
- Connectives
- Whether sentences open with however, moreover, therefore.
- Formality
- Contractions, plain words against formal ones, register.
- First person
- How often you write as I or we rather than impersonally.
- Punctuation
- Semicolons, dashes, parentheses — the marks you reach for.
Six families, and sixteen measured dimensions underneath them on the draft above. Not a personality, a style score, or a judgement about quality.
The comparison
Both readings, side by side.
Each dimension shows what your samples do and what the draft does, in the engine's own words. A difference is reported as a difference — not as a mistake, and not as a score.
Similarity to your samplesLow
2 samples · 16 dimensions
- Longer sentences than your typical range
- Yours · Typically 6–23 words · avg 13.3Draft · Avg 18.8 words · 9 sentences
- Differs
- More assertive than your usual writing
- Yours · Certainty markers in 0% of sentencesDraft · Certainty markers in 11% of sentences
- Differs
- Less first person than your samples use
- Yours · First person in 57% of sentencesDraft · First person in 11% of sentences
- Differs
- More formal than your usual register
- Yours · Moderately formalDraft · Formal
- Differs
- Harder to read than your usual writing
- Yours · Grade level 4.8Draft · Grade level 11
- Differs
- Heavier vocabulary than usual
- Yours · 1.27 syllables per wordDraft · 1.68 syllables per word
- Differs
- More abstract nouns than you usually use
- Yours · 1.1 nominalisations per 100 wordsDraft · 3.2 nominalisations per 100 words
- Differs
- Shorter paragraphs than you usually write
- Yours · Typically 5–6 sentencesDraft · Avg 3 sentences · 3 paragraphs
- Drifts
And 8 more measured dimensions. A stated rule over the measured differences — not a model score, and never evidence of who wrote the text.
How to read it
- Consistent — the draft sits where your samples sit on this dimension.
- Drifts — a noticeable gap, worth a look if you meant to sound like yourself.
- Differs — the draft is doing something your samples rarely do.
A draft can differ from your profile for good reasons: a different assignment, a co-author, a discipline with its own conventions. The lens reports the difference and leaves the decision with you.
Suggestions
Small, mechanical, and only where one is safe.
Where a measured difference has a safe span-level move behind it — expanding a contraction, dropping a signposted opener, splitting a long sentence — Syft offers it. Where it does not, the difference is reported and left to you.
Most of what makes a voice cannot be edited mechanically without inventing content. Those differences are reported and never rewritten.
What it never changes
- Your argument
- A difference in habit is reported; the claim is never touched.
- Quotations
- Someone else's words are never adjusted toward your profile.
- Numbers and citations
- Figures and references survive every suggestion.
- Discipline conventions
- A field's house style is not a deviation from you.
Method
What this can and cannot tell you.
The honest description of a measurement, so the result is read for what it is.
It compares measurements, it does not learn a voice. Syft counts the same features in your samples and in your draft and reports the gaps. There is no model here that writes in your style, and nothing on this page claims one.
It is not evidence of authorship. A draft that matches your profile was not necessarily written by you, and one that differs was not necessarily written by someone else. Writing changes with the task, the deadline and the subject. Syft never presents this as proof and neither should anyone else.
It needs enough text to say anything. Below a real amount of sample writing the lens says so and reports nothing, rather than producing a reading from too little evidence.
It is separate from AI detection. This lens never reads Detect’s result, and a difference from your profile is not a signal that anything was AI-written. The two answer different questions.
Two different questions
Machine-like is not the same as unlike you.
Humanize asks
Does this passage carry the habits that generated prose tends to have — uniform sentences, announced transitions, stock openers?
Academic Voice asks
Does this passage read like the writing you already do? A sentence can be perfectly natural and still unlike your usual work.
Part of Syft
Four capabilities, one workspace.
They read the same document and the same analysis, so a signal you find in one is the passage you revise in another.
See the whole workspace — documents, metadata, provenance and code as well as writing.
Start with two pieces of your own work.
Add them once, and every draft after that can be measured against them.