Takes your writing from “a model wrote this” to “I wrote this.”
Paste two paragraphs you've actually written. EchoWrite measures how you sound, then rewrites anything in it.
your last 40 sentences
One of these is better writing. It just isn't yours.
Save it and point it at anything.
We'll carry the writing you just pasted straight into your account — name it once and reuse it on every draft from here.
no card · 15 rewrites included
Read them again. The left one is better writing — and it isn't yours.
It was never the vocabulary.
A model picks the most probable next word, then does it again. Same rule every step — so the rhythm comes out flat.
Yours doesn't. You write nine words. Then thirty-four, because you got attached to a subclause halfway through. Then three. That variance is the signal.
↑ the same shape as the ridgeline you just scrolled past
Three steps. About ten seconds.
No prompt engineering. No wrestling with tone every time. Set your voice once, reuse it forever.
Paste your samples
Two or three things you've genuinely written. EchoWrite builds a private fingerprint of how you actually sound.
Add your draft
A rough idea, a bullet list, or a stiff AI first draft — whatever's fastest to get out of your head.
Get it in your voice
Back in seconds: your cadence, your phrasing, the model-speak stripped out. Copy and ship.
Every voice gets a colour.
One for your brand, one per client, one for the newsletter. Each keeps its own fingerprint — and its own colour, everywhere it shows up in the app.
Not a humanizer. A voice engine.
Most tools flatten everything into the same "casual" mush. EchoWrite learns your specific voice and holds it across everything you publish.
Style fingerprint
Deeper than tone. It maps sentence rhythm, punctuation habits, and the function words you never chose on purpose — then tunes the same voice per surface, so LinkedIn and email both still sound like you.
AI-ism scrubber
Drops the tells that scream template — automatically, on every rewrite. Twenty ready-made library tones sit alongside your saved voices when you want a starting point.
Rewrite history
Every rewrite is saved and re-openable, colour-tagged by the voice that produced it. Go back, re-run it in a different voice, or lift the version you liked three weeks ago.
Fact-safe rewrites
It rewrites meaning and voice, not substance. Numbers, names, dates and claims pass through untouched — so you can point it at a launch email without re-checking every figure.
You already have a writer. You don't have your voice.
A general model writes well in a voice. A humanizer flattens everything into the same loose register. Neither one is yours, and neither one remembers.
| ChatGPT | Humanizers | EchoWrite | |
|---|---|---|---|
| Learns your specific voice | ✕ | ✕ | ✓ |
| Reusable across every draft | ~ | ✕ | ✓ |
| Keeps your facts and numbers | ~ | ✕ | ✓ |
| Separate voice per client or brand | ✕ | ✕ | ✓ |
| No prompt writing, ever | ✕ | ✓ | ✓ |
✓ yes · ~ possible, but you rebuild it every time · ✕ no
Style is a measurable thing.
Detectors score generated text on two axes: perplexity, how predictable each word is, and burstiness, how much that predictability swings from sentence to sentence — measured, literally, as the standard deviation of the per-sentence scores.1 AI text scores low on both.
Detectors were built on this, and they're bad at it. They've flagged the U.S. Constitution as machine-written.2 So we're not in the business of beating a checker — a score that fires on the Constitution isn't a standard worth optimising against, and the biggest ones have since moved to other methods anyway.3
But the measurement underneath is real. Stanford researchers showed that model-written text sits in a measurably different region of a language model's probability landscape than human writing does — the signal exists whether or not any product detects it reliably.4
And it's old. In 1964 two statisticians ended a 175-year argument over the disputed Federalist Papers by counting words like upon and whilst — excluding topic words entirely, using only articles, pronouns, prepositions and conjunctions.5 Function words. The ones nobody picks on purpose. Sixty years on, the finding holds: the most identifying thing about your writing is the part you aren't thinking about.
You are a distribution.
Which is why pasting a paragraph works and a tone dropdown never will.
Sources & further reading
- Chen, Kang, Zhai, Li, Guo & Liu (2023) — On the Generalization of Training-based ChatGPT Detection Methods. Sets out how perplexity and burstiness are actually computed, burstiness being the standard deviation of per-sentence perplexity. arxiv.org/abs/2310.01307
- Benj Edwards, Ars Technica (2023) — Why AI detectors think the US Constitution was written by AI. The false-positive problem, explained with the creator of GPTZero. arstechnica.com
- GPTZero (2023) — What is perplexity & burstiness for AI detection? Their own note that they moved off these metrics in autumn 2023 in favour of a deep-learning architecture. gptzero.me
- Mitchell, Lee, Khazatsky, Manning & Finn (2023) — DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature. Model-written text occupies negative-curvature regions of the log-probability function; the underlying statistical difference is real. arxiv.org/abs/2301.11305
- Mosteller & Wallace (1964) — Inference and Disputed Authorship: The Federalist. The founding work of stylometry: contextual words excluded, authorship settled on function words alone. press.uchicago.edu
That fingerprint is the thing you're buying. Not a rewrite button — a saved measurement of how you write, that you can point at anything.
Simple, flexible pricing.
Start free. Move up when you're writing every day — or pay once and keep it for life.
- 1,000-character rewrites
- 15 rewrites to start
- 1 saved voice
- Core voice-matching
- 3 starter styles
- 3,000-character rewrites
- ~200 rewrites / mo
- 3 saved voices
- Everything in Free
- Rewrite history + AI-ism scrubber
- 8,000-character rewrites
- ~600 rewrites / mo
- 10 saved voices
- Everything in Starter
- Full library, length control, priority
- Unlimited length
- ~1,500 rewrites / mo
- 25 saved voices
- Everything in Pro
- Fastest speed
Cancel anytime · no card needed to start · keep access to the end of your paid period
Pay once. Yours forever.
- Unlimited length
- Everything in Studio
- Unlimited saved voices
- Highest rewrite limits
- Every future feature
- No renewal, ever
Good to know.
How many samples does it need?
Two or three paragraphs you actually wrote is plenty. More real "you" gets a sharper match — polished, edited pieces work better than raw notes.
Is my writing kept private?
Your voice profiles are yours. They're stored to your account so you can reuse them, and aren't shared with other users.
What counts as one rewrite?
One rewrite of one draft counts as one. Re-running the same draft in a different voice counts as another. Your monthly allowance resets at the start of each billing month.
Can I keep separate voices?
Yes — that's the point. Keep one for your personal brand, one per client, one for the company newsletter, and switch between them.
How is this different from ChatGPT?
ChatGPT writes well, but in its voice — and you rebuild the instructions every session. EchoWrite measures how you write, saves that as a reusable profile, and applies it to any draft in one click. See the comparison above.
Can I get a refund?
Yes — 14 days, no questions asked, on your first purchase of any paid plan. Email us and we'll refund it in full. Full detail on the refunds page.
What does Lifetime include?
Everything in Studio, unlimited saved voices, the highest rewrite limits, and every feature we ship from here — paid once, no renewal.
Can I cancel anytime?
Anytime, from your account — no email required. You keep full access until the end of the period you've already paid for, and you're not billed again.