ToHuman.io is a rewriting tool with a free tier of 2,500 words a month, a $19 Pro plan, and an API. It is more candid than most of its category about what it does and does not achieve. This page covers its published figures, a clause in its own terms worth reading, and who is better served by a detector.
Paste some text or load the example to estimate AI likelihood.
Read from tohuman.io on 16 August 2026. The site returned HTTP 200 throughout.
The product rewrites pasted AI text at three intensity settings named minimal, medium and heavy. It advertises a response time of three to five seconds and states that it runs a self hosted language model with no OpenAI dependency, and that data is never shared or used for training. An API is documented with examples in cURL, Python, Node.js and Ruby.
| Option | Price | Allowance | Notes |
|---|---|---|---|
| Free | $0 forever | 2,500 words per month | No signup needed to try, all intensities, document history, API access at 60 requests per minute |
| Pro | $19 per month | 100,000 words per month | API at 300 requests per minute, priority support |
| Word pack | $9 one off | 50,000 words | No subscription, and the site states the words never expire |
Two details stand out. The free tier is the most usable in this group: 2,500 words a month, no account needed, and API access included rather than gated behind payment. The $9 word pack lets you buy capacity without a recurring charge, which is unusual here. When the free allowance runs out, the site says humanization pauses until it resets on the first of the next month, or resumes on upgrading. Billing runs through Polar, and no refund policy is published anywhere.
The footer and terms identify tohuman, LLC, with a 2026 copyright. No registered address, state of incorporation or company number is published, and the terms name no governing law. The contact email is obscured behind Cloudflare's email protection script. The terms carry a last updated date of March 23, 2026, the most recently revised legal text of any tool in this group.
We also confirmed the service runs. An unauthenticated POST to tohuman.io/api/v1/humanizations/sync returned HTTP 401, meaning the endpoint exists and enforces authentication rather than erroring, and response headers identify a Ruby on Rails application on Railway. That is a working backend, not a landing page.
Most tools in this category advertise round numbers approaching perfection. ToHuman.io instead publishes a dated, specific and notably imperfect figure: In our April 2026 testing, ToHuman's heavy intensity mode produced output that scored below Turnitin's AI detection threshold in roughly 79% of test runs.
That is the vendor's own testing, reported here as theirs, with no methodology or sample size published alongside it. But note what it concedes: on their own numbers, roughly one run in five did not clear the threshold, using the most aggressive setting available. Anyone treating this class of tool as a guarantee should sit with that figure, and it is to ToHuman.io's credit that it is on the page at all.
The most useful thing we found, if you are considering the tool for coursework.
The homepage headline reads Make AI Text Undetectable in 5 Seconds, and the copy beneath states the tool converts text to writing that bypasses GPTZero, Turnitin and Originality.ai.
Section 4 of the terms takes a different position. Users must not Use the Service to misrepresent AI-generated content in contexts where disclosure is required by law (e.g., academic submissions governed by institutional policies).
Read together, the marketing names Turnitin as a system to be bypassed while the contract you accept prohibits the academic use that naming implies. We are not calling this deceptive or speculating about intent. Both texts were live on the same domain on 16 August 2026, and the terms are the part you are legally bound by. If your intended use is a graded submission, you would be agreeing to terms that exclude it.
Several of these are genuine.
The free tier is the best in this category. 2,500 words a month with no signup, all intensity levels, and working API access. Nothing we offer free competes as a rewriting allowance, because our free tier is a detector.
You can buy capacity without a subscription. The $9 word pack for 50,000 words with no expiry suits occasional use far better than a monthly plan. Most competitors sell nothing except a recurring charge.
A documented API on the free tier. 60 requests a minute at no cost, rising to 300 on Pro, with examples in four languages. A low friction way to start building.
Self hosted inference and a no training commitment. The site states it runs its own model with no OpenAI dependency and does not share or train on your data. For anyone handling sensitive text, that is stated more plainly than most.
Recently maintained legal terms. A March 2026 revision date suggests active upkeep, more than several competitors can show.
TextSight is not a quieter version of this tool. ToHuman.io sells output that clears detectors. We sell the detector. Our rewriting exists to make writing clearer and more specific, not to move a score, and we publish no figures about how our output performs against Turnitin, GPTZero or anything else. That is a difference in purpose, not a gap in our feature list.
So the straightforward advice: if you want another tool that does what ToHuman.io does, keep looking in that category. This page will not talk you into something you did not come for.
People on the other side of the document: editors receiving freelance copy, hiring managers reading written assessments, course leaders working through submissions, content leads auditing what an agency produced last quarter. The question is what a piece of text is, and the deliverable is a judgement plus a record of it.
There is also the diagnostic case. Writers run their own drafts to see which passages read as machine generated, then edit for specificity and rhythm. That tends to produce better prose than any processing pass.
It runs without an account. Anonymous use gives three checks a day, pooled across the AI Detector and the Document Detector, capped at 1,000 words per check. A free account raises that to 60 a day at 5,000 words, and any scan run while signed in is saved to Scan History, free accounts included. Starter gives 200 daily, Pro and above unlimited. Minimum input 25 words.
Three bands are reported: below 30 is likely human-written, 30 to 69 is mixed signals, 70 and above is likely AI-generated.
One thing we will not bury. The detector is an English only ensemble of DeBERTa, RoBERTa and ELECTRA, and it has a known false positive bias against writers who learned English as a second language. If you are assessing ESL writers, a high score means look closer, not case closed. We would rather lose a sale than have someone treat our output as proof.
Annual billing is 25 percent below monthly.
| Plan | Monthly | Annual | Adds |
|---|---|---|---|
| Starter | $9.99 | $7.49 | Sentence level analysis, PDF export, 200 scans daily |
| Pro | $19.99 | $14.99 | Unlimited scans, Chrome extension, REST API |
| Business | $39.99 | $29.99 | Teams, branded reports |
| Enterprise | $79.99 | $59.99 | Full capability at scale |
File upload, URL scanning and bulk scanning come with the paid plans. Note that ToHuman.io's Pro at $19 and our Pro at $19.99 sit at almost the same price while buying entirely different things: theirs buys 100,000 rewritten words, ours buys unlimited scanning plus the API and extension.
ToHuman.io has the most generous free tier in its category, honest pricing mechanics and unusually candid claims. If you want rewriting it is a reasonable choice, subject to its academic use clause. TextSight does the different job of evaluating English text and producing a report you can act on.
tohuman.io homepage: product description, intensity levels, response time, self hosted model and no training statements, free allowance, Pro price, word pack, April 2026 testing claim, company line. HTTP 200. 16 August 2026.tohuman.io/pricing: all tiers, API rate limits of 60 and 300 requests per minute, allowance reset behaviour, Polar as billing provider. HTTP 200. 16 August 2026.tohuman.io/terms: tohuman, LLC, acceptable use section 4, last updated March 23, 2026, no governing law or refund terms stated. 16 August 2026.tohuman.io/api/v1/humanizations/sync: unauthenticated POST returned HTTP 401 with Railway and Rails headers. 16 August 2026.The deeper head-to-head: feature table, where GPTZero wins, where TextSight wins, and migration steps.
Read the comparePerplexity, burstiness, and transformer scoring explained, so you can read any detector's verdict critically.
Read the guideWhy detectors over-flag ESL and formal writing, and how sentence-level evidence helps you defend real work.
Read the guideThe full ranking with detection approach, pricing, and use-case fit side by side.
See the rankingStart with TextSight's free tier. No card, no signup, no commitment. Your first scan in about six seconds, with the evidence shown line by line.
Honest comparisons vs other tools.