05 · Bench
Free Brand Voice Analyzer
Analyze a public page or pasted sample to document repeatable writing patterns. The analysis runs in your browser and does not reduce voice to a score.
Analyze observable writing patterns
How the result is produced
The browser-side parser calculates transparent text statistics and labels observable patterns; no LLM is involved. In URL mode Contextter fetches the public page once, extracts the main text, and caches it for 30 days - the analysis still runs on your device.
Start freeBrand voice rarely fails at the writing. It fails at the handover.
Almost every team has a sense of how it wants to sound. Almost none can write it down so a new writer or a language model can work from it. “Professional but approachable” is not a brief - it is a hope that every participant resolves differently.
Observable properties close that gap. How long are the sentences, and how much does the length vary? Is the reader addressed directly, formally, or not at all? Which turns of phrase recur often enough to have become a signature? These are not quality judgements but descriptions - and that is exactly why they can be handed over.
The side effect is useful: a sample profile can make repeated two- and three-word phrases and a tightly grouped sentence-length distribution visible. Those observations are not errors by themselves. Read the evidence before deciding which measured traits belong in a brief.
What Brand Voice Analyzer checks
A writing voice cannot be scored, but a concrete sample can be described. The analyzer measures sentence and paragraph length, punctuation, configured perspective and style markers, recurring literal phrases, and sample-length-dependent lexical variety.
What comes out is a profile, not a verdict. None of the axes has a good end and a bad end - short sentences are not better than long ones, and addressing the reader directly is not better than keeping a factual distance. What counts is recognisability: two texts from the same brand should show a similar profile, and a new text should fit into it.
The practical use is handover. A described profile can be handed to a new writer, an agency, or a language model. That is considerably more workable than the usual “professional but approachable”, which everyone reads differently.
The terms, briefly
- Sentence rhythm
- The distribution and quartiles of sentence lengths, not only the average. The analyzer reports how tightly or widely the observed lengths are grouped.
- Perspective
- Literal first-, second-, and third-person pronoun matches plus configured sentence-opening address indicators. This is heuristic evidence, not a grammatical parse.
- Recurring phrasings
- Contiguous two- and three-word spans that appear repeatedly. The analyzer reports their counts and examples without treating repetition as automatically good or bad.
- Configured style markers
- Exact locale-specific matches for hedges, assertions, transitions, formality, informality and supported contractions, always shown with counts and examples.
Reading the result
Short and varied
The sentence-length quartiles are far apart and several distribution bands contain sentences.
Record the measured range if this sample is representative; the analyzer does not grade it.
Direct second person
Configured second-person markers recur, with exact counts and examples shown in the profile.
Decide editorially whether comparable texts should keep that address; the count itself is not a quality signal.
Uniform rhythm
The sentence-length quartiles are close together and most sentences fall into the same distribution bands.
Change nothing automatically. Consider variation only if the measured regularity conflicts with the intended brief.
Not enough text
The sample is too short for reliable statements. Below roughly 300 words the properties fluctuate too much.
Combine several representative texts and run it again.
Three mistakes we see constantly
Reading the profile as a score
No axis has a correct side. A profile describes how a text sounds - whether that suits the brand is an editorial decision, not a measurement.
Using an untypical sample
A press release, a product page and a blog post sit in different registers. Analysing one atypical sample describes a text type and calls it a brand voice.
Pasting the result into a prompt unread
A profile also describes traits you would rather not perpetuate - a filler phrase that crept in, for instance. Read it once as an editor before handing it to a language model.
Questions about Brand Voice Analyzer
How much text do I need for a reliable profile?
About 300 words is the floor for stable values; several hundred to a thousand words drawn from two or three typical texts is considerably better. Representativeness matters more than volume: use texts that stand in for what you intend to produce.
Do I get a score for my writing quality?
No, by design. Quality is contextual - a text that works in technical documentation would be wrong on a landing page. The analyzer describes observable properties and leaves the judgement to you.
Can I use this to detect AI-generated text?
No. There is no reliable detection, and tools claiming otherwise produce substantial error rates in both directions. The analyzer can report a tightly grouped sentence-length distribution, but that proves neither origin nor quality.
Does my text leave the browser?
Pasted text never leaves your device: analysis, brief and JSON export run locally, and nothing is sent to an LLM or analysis API. In URL mode Contextter fetches the public page you name and hands the extracted text back to your browser - the analysis still runs locally.
What happens when I check a URL?
A short proof-of-work verification protects the check, then Contextter fetches the public page once, strips navigation and scripts, and returns the main text. That page text is cached for 30 days, so repeating the check reuses the stored snapshot instead of fetching again; the profile itself is computed in your browser.
How do I actually use the profile?
As part of a brief. A described profile with concrete traits - sentence rhythm, perspective, recurring phrasings - is more usable for new writers and for language models than adjectives like “professional” or “approachable”, which everyone interprets differently.
Primary sources
The claims on this page come from the providers' own documentation, not from secondhand summaries. Every source is linked so you can read what it actually says.
Continue with Contextter
A profile describes. A system holds to it.
The difference between a described style and a maintained style is repetition - on every text, by every person.
- Stores the style profile in the workspace so every generated draft starts from it.
- Checks new text against the profile instead of relying on recall.
- Holds the evidence, figures and sources from your own knowledge base, so tone does not come at the cost of substance.
No payment details required. The free check above stays usable without an account.
Related free tools
Where this continues in Contextter
Copy the local brief and opt in by pasting it into a content brief or writing workflow.
Continue in ContextterReady when you are.
Start the check in the form above. Resource-intensive runs require a short browser verification and use strict limits.
Inputs are used only for the selected check. Browser-only tools never upload pasted text or files.
No payment details required. The free check above stays usable without an account.