Sound like you. Not like a bot.
Humanizer strips the AI fingerprints from your drafts and rewrites them in a voice profile built from your own writing — reviewed and approved by you.
Readers can spot AI writing in seconds
When researchers measured vocabulary across biomedical abstracts, at least 13.5% of 2024 PubMed abstracts showed signs of LLM assistance — approaching 40% in some subgroups — and once-rare words like "delve" spiked far past their historical baseline (Kobak et al., 2024). The tells cluster into four signs:
1. The vocabulary
A small set of words AI leans on far more than people do: delve, underscore, tapestry, realm, testament, landscape, intricate, meticulous, vibrant. One or two are normal. A cluster is a fingerprint.
2. The structures
The "it’s not just X, it’s Y" reversal. Everything in threes. Bullet lists where every item is a bold phrase, a colon, then an explanation. Em dashes stitching it all together.
3. The confident emptiness
Profound-sounding prose that commits to nothing. Human writers hedge when they’re genuinely unsure — "I suspect", "as far as I can tell", "I could be wrong". AI writing almost never does.
4. The missing judgments
People say better, worse, best, worst — and pick a side. AI writing balances every claim with a counterpoint until no actual opinion survives.
How it works
Upload samples
Give us writing that is unmistakably yours — emails, posts, essays, docs. We distill it into a voice profile.
Approve your voice
Read the profile, edit anything that doesn't ring true, and approve it. Nothing runs against a profile you haven't signed off on.
Humanize documents
Paste or upload a draft. We strip the AI tells and rewrite it toward your approved voice — usually in under a minute.
Try the tell detector
No account, no upload — detection runs entirely in your browser.
Runs entirely in your browser — your text never leaves this page.
Paste some text to see which AI tells it triggers.
Plans
| Freeaccount required | Subscriberany active product subscription | |
|---|---|---|
| Voice profiles stored | 1 | 5 |
| Humanize jobs | 2 per rolling 7 days | 5 per rolling 24 hours |
| Profile generations | 2 per rolling 7 days | 3 per rolling 24 hours |
| Export / import / delete profiles | Yes | Yes |
Rolling windows, not calendar days. Failed jobs refund their quota automatically.
Frequently asked questions
Is my text used to train AI models?
No. Your samples and documents are processed only to build your voice profile and rewrite your documents. We never use your content for training, and our AI providers do not train on it either.
What file types can I upload?
Plain text, Markdown, Word (.docx), PDF, HTML, RTF, and OpenDocument (.odt), up to 5 MB per file. You can also just paste text.
What exactly is a voice profile?
A structured Markdown description of how you actually write — sentence rhythm, vocabulary, punctuation habits, what you emphasize — distilled from writing samples you upload. You review and edit it, and nothing is used for rewriting until you approve it.
Can I use my profile outside the website?
Yes. Export any profile as a plain .md file compatible with our local Claude Code humanizer skill, and import profiles the same way. Your voice is yours to take with you.
What happens when I hit my quota?
Quotas use rolling windows, not calendar days, so capacity frees up exactly one window after each job. The studio shows when your next slot opens. Failed jobs never count against your quota.
Generic advice says "write like you talk." We go further: we rewrite toward your approved voice — built from your writing, edited by you, applied to every document you run through it.
Open the Humanizer Studio