Skip to workbench

Humanize AI text. Sound like yourself again.

your text55 / 250
try a sample
humanized
↓ your rewrite appears here
5 free today · no signup

Paste AI text. Get a human rewrite. See the before/after score. Free. No signup.

how it works

Three steps. About ten seconds.

step 01

Paste your text.

Up to 250 words on Free, unlimited on Pro. We don't store it — processed in memory, discarded on response.

step 02

Humanize.

v3 prompt actively substitutes AI-tell phrases and rewrites abstract noun chains into verb + concrete noun. ~10 seconds.

step 03

Score on both sides.

Score before and after against GPTZero, Turnitin, Originality, and Copyleaks. See the drop on every detector before you ship.

voices

Writers we’ve heard from.

Lightly edited. Names are real first-name-plus-initial; full names on request.

I write in English as my third language. Detectors kept flagging my essays as AI even when every word was mine. This is the only tool that fixed it without changing my meaning.
Priya R.MSc student · ESL

“Before/after scoring on the same screen is the part that sold me. I can see whether to ship or run it again. Most humanizers just hand you back a paragraph and hope.”

Marcus D.
Content marketer · Toronto

“Used the free tier for two weeks before upgrading. The voice profiles paid for the Pro subscription on the first article. Cancel-anytime sealed it.”

Jen K.
Freelance writer
what else it does

Built for serious humanization.

Free covers the casual rewrite. Pro is the rest of the kit — voice profiles, reference style match, academic mode, the bypass loop, batch processing, and a JSON API.

voice profiles

Train it on your writing.

Paste 100–2000 words you wrote, save a style fingerprint, and every humanize run can match your cadence and habits.

  • ·up to 8 saved voices
  • ·150-word style summary
  • ·re-train anytime
open →
reference

Match anyone's voice.

Paste a reference doc and the humanizer analyzes it on-the-fly, then rewrites your text in that style. One-off, not persisted.

  • ·100–2000 word ref
  • ·inline style analysis
  • ·adds ~3–5s per run
open →
academic mode

Quote-aware rewriting.

Preserve direct quotations and in-text citations like (Smith, 2020) verbatim. Humanize the prose around them without touching the cited material.

  • ·straight & curly quotes
  • ·blockquotes
  • ·in-text citation tokens
open →
bypass loop

Iterate until every detector passes.

Run the humanize → detect cycle automatically. Ramps intensity each attempt and stops when every detector lands below your target — or when output stops improving.

  • ·up to 5 attempts
  • ·lenient / standard / strict
  • ·ceiling detection
open →
batch

Twenty docs in one drop.

Drag in up to 20 .txt or .md files, humanize them in parallel with concurrency-limit 5, then download the whole set as a ZIP with a manifest.

  • ·txt + md (pdf/docx soon)
  • ·5 MB per file
  • ·manifest.txt included
open →
public api

Build it into your stack.

Bearer-auth JSON API for humanize, detect, voices, and the bypass loop. Single-call composition: humanize + detect + style match in one round-trip.

  • ·humanize · detect · bypass
  • ·voice profiles via API
  • ·5 keys per user
open →
pricing

Simple pricing. Cancel anytime.

Free covers the casual case. Pro is for people shipping copy at volume. Team rounds out the kit when more than three people share a workflow.

Free

$0
forever · no card
  • ·20 humanizations / day (5 / day anonymous)
  • ·Up to 250 words per request
  • ·Score against GPTZero, Turnitin, Originality, Copyleaks
  • ·Diff view + flag highlights
  • ·Magic-link or Google sign-in
Start free →
best for serious volume

Pro

$24
per month
  • ·200 humanizations / day
  • ·Unlimited words per request
  • ·Intensity 1–5 dial
  • ·5-detector consensus check
  • ·Voice profiles (up to 8) + reference style match
  • ·Academic mode (quote-aware)
  • ·Bypass loop: iterate to a target detector score
  • ·Batch processing (up to 20 docs)
  • ·Public JSON API + 5 keys
  • ·30-day refund · cancel anytime

Team

—
talk to us
  • ·Everything in Pro
  • ·Shared voice library
  • ·Custom rate limits + SLA
  • ·SSO + audit log (on request)
  • ·Priority email support
Contact us →

30-day refund on Pro. We don’t store your text — see privacy.

detailsSee the full feature-by-feature comparison+
featurefreeproteam
Daily humanizations20 / day (5 anon)200 / dayCustom
Words per requestup to 250UnlimitedUnlimited
Detector scoringIn-house + 4 vendorsIn-house + 4 vendors + bypass loopIn-house + 4 vendors + bypass loop
Intensity dial (1–5)—✓✓
Voice profiles—Up to 8Shared library
Reference style match—✓✓
Academic mode (quote-aware)—✓✓
Bypass loop—Up to 5 attemptsUp to 5 attempts
Batch processing—20 docs / drop20 docs / drop
Public JSON API—5 keys5 keys
Priority supportCommunityEmailPriority email
Refund windown/a30 days30 days
questions

Things people ask.

Is this a cheating tool?+
How you use it is between you and your institution. What this is good for: rescuing prose you wrote that detectors mistake for AI (a known problem for non-native English speakers — Stanford 2023 measured a 61% false-positive rate). What it isn’t for: passing AI text off as original research in a context where original research is what’s being graded. Read our ethics page before you ship anything important.
Is my text stored?+
No. Inputs and outputs are processed in memory and discarded immediately. The only thing we keep is a structured log line with hashed IP, word count, latency, model name, and prompt version — never your prose. The Pro voice-profile feature is the single opt-in exception, and the sample text you upload there is scoped to your account and deletable in one click. Full posture on the privacy page.
Will it actually bypass detectors?+
Most of the time, on text that started obviously AI, yes. We score every rewrite against GPTZero, Turnitin, Originality, and Copyleaks so you see the drop on each one before you ship. We don’t promise 100% on every detector — Pangram is hard, vendor models retrain constantly — but the actual before/after numbers are right there for you to decide.
What’s the difference between Free and Pro?+
Free: 20 humanizations a day at up to 250 words each, magic-link sign-in, the in-house AI score on input and output, and the diff view. Anonymous (no sign-in): 5 / day. Pro ($24 / month): 200 humanizations / day, unlimited words, intensity dial (1–5), multi-vendor consensus check, voice profiles, reference style match, academic mode (quote-aware), the bypass loop (iterate to a target score), batch processing (20 docs at a time), and a JSON API with up to 5 keys.
Can I cancel anytime? Is there a refund?+
Yes to both. Cancel from your account dashboard in one click — your Pro features stay live through the end of the billing period, then you revert to Free. First Pro charge is covered by a 30-day refund: email us and we’ll process it, no questions. Billed monthly via Stripe; no annual lock-in.
Does it work for non-native English (ESL) writers?+
Yes — this is one of the use cases we actively design for. AI detectors flag non-native English at roughly 61% (Stanford 2023) even when the prose is completely human-written, because the cadence and vocabulary patterns trip the classifier. The v3 prompt rewrites the surface features the detectors lean on without changing your meaning. Free tier is enough to test it on your own writing before signing up.
Are the GPTZero / Turnitin / Originality scores real-time API calls?+
No, and we’re upfront about it — each non-in-house cell wears a SIM badge. Live per-request calls to vendor APIs would shred margin (GPTZero and Originality charge per scan, Turnitin doesn’t even sell a consumer API) and we’d have to pass the cost on. Instead each vendor score is a deterministic surrogate: calibrated offline against the real GPTZero API and against published comparison data for the others. The in-house cell is a live computation on every request. Full methodology on the benchmarks page.
Do you have an API?+
Yes, for Pro accounts. Bearer-auth JSON for /v1/humanize, /v1/detect, and /v1/bypass. Up to 5 keys per account. See the docs.