Which AI humanizers are students actually using? A teacher’s guide (2026)

The Clever vs. Walter vs. QuillBot thread ended with a question about which other humanizers are turning up in student work. I have pulled together a field guide from what I see in my own classes, what colleagues in my department have mentioned, and, for every price and limit below, the vendor’s own website. I have linked each source so you can check it yourself, because these numbers change every few months. Everything was checked on November 2, 2026, and prices are in US dollars.

Two things up front. First, for hands-on detector results, see the Walter review thread and “It Works. That’s the Problem.” here on the forum. This post sticks to what each tool offers. Second, this is not a how-to for students. If you teach older grades, assume your students already know these names. The point is that we should know them too.

The short version: Walter Writes is the most capable humanizer students are likely to find, followed by QuillBot, Clever AI Humanizer, Grammarly’s humanizer and Phrasly. Undetectable AI and StealthWriter are the names with the most “beat the detector” marketing. A humanizer changes how text sounds, not who did the thinking, and that is where our assessment design has to focus.

What a humanizer actually does

A humanizer takes text, usually from ChatGPT, Claude or Gemini, and rewrites it so it reads less like a chatbot: varied sentence length, less predictable word choice, fewer of the tidy transitions we have all learned to spot. Some tools stop there. Others bolt on an AI detector so the student can check the result before handing it in. That second feature is the one that changes the classroom picture, because it closes the loop: write, rewrite, check, submit.

The tools, roughly in order of capability

1. Walter Writes. This is the most complete tool I have come across, and the one I would expect a motivated senior student to settle on. Its humanizer page offers three rewrite levels (Simple, Standard and Enhanced), three tones (casual, academic and brand-safe), a built-in detector to scan text before and after, file upload, and support for 80+ languages. You can humanize 300 words free without signing up (Walter’s humanizer page). Paid plans start at $12 a month, or $8 a month billed yearly, for 30,000 words (Walter pricing). The academic tone setting is the detail teachers should notice: it is designed for exactly the register we ask for in essays. The language list matters for us too, given how many of our students write in French or in a first language other than English.

Good for (from a student’s point of view): polished essays where tone matters. From ours: the reason a “this doesn’t sound like them” conversation now matters more than a detector score.

2. QuillBot. Most of our students already know QuillBot as a paraphraser, and many boards have had long debates about it (see the QuillBot ban thread). It now has a humanizer with a free Basic mode limited to 125 words at a time, and its page lists Canadian English as one of its English dialects alongside French, Spanish, German and Portuguese (QuillBot’s humanizer page). Its browser extension can humanize text inside ChatGPT and Gemini. To QuillBot’s credit, the page tells users to follow their institution’s AI rules and cite AI use.

3. Clever AI Humanizer. Completely free, with no fixed monthly limit and up to 3,000 words per request, in English, Spanish and Portuguese (Clever AI Humanizer). “Free with no account” is a powerful combination for a grade 11 student at 11 p.m. It does not cover French.

4. Grammarly’s humanizer. This one surprises colleagues. Grammarly offers a free humanizer, and a free account adds an agent with four preset voices plus a custom voice built from a sample of the student’s own writing, in six languages including French. Grammarly’s own FAQ says the tool is not intended to bypass AI detectors (Grammarly’s humanizer page). Worth knowing: its parent company, Superhuman, now owns GPTZero (Superhuman, September 2026), so the same company sells a humanizer and one of the best-known detectors.

5. Phrasly. One free run a day of up to 500 words without an account, and a paid plan with unlimited humanizing (Phrasly). Its page also asks users to follow institutional AI guidelines.

6. Undetectable AI. The name says it. It offers a free basic model, a 250-word trial of its paid model, and paid plans from $9.99 a month (Undetectable pricing). It even offers a refund if output is flagged. Less capable than Walter on tone and languages, but its marketing is aimed squarely at the detector question.

7. StealthWriter. Needs a free account, then allows 10 humanizations a day of up to 1,000 words each (StealthWriter pricing). Interestingly, its own FAQ admits no rewrite guarantees a particular detector score.

Others you may hear about: WriteHuman, which has iOS and Android apps (relevant if your students do everything on their phones), and BypassGPT.

The Canadian wrinkle: French and multilingual writers

This is the part I have not seen discussed much here. Turnitin’s product update page lists AI writing detection in English, Spanish, Japanese and Arabic, with Turkish, German and Dutch added in October 2026 (Turnitin product updates). French is not on that list. Its detection of text run through humanizers launched in August 2025 for English only (Turnitin). For French Immersion and francophone classrooms, that means the detector-based safety net many of us assume exists is thinner than we think.

The other side of that coin is fairness. A Stanford team found that common GPT detectors misclassified writing by non-native English speakers as AI-generated while correctly identifying native writing (Liang et al.). In classrooms with many English language learners, a high score can say more about the writer’s English than about the tool they used. If you work with ELL students, this thread on AI writing tools for ELL students is worth reading alongside this one.

What I am doing differently this year

• I stopped treating a detector score as evidence. It is a reason to talk to a student, never a finding. Vanderbilt turned Turnitin’s AI detector off back in 2023, estimating that even a 1% false positive rate would have wrongly flagged about 750 of its 75,000 papers (Vanderbilt).
• More writing happens in class, in stages: a handwritten plan, an in-class first paragraph, then the take-home draft. Humanizers polish a final product; they cannot fake a process I watched.
• Short conferences. Two minutes of “walk me through your second argument” tells me more than any score.
• I name the tools in my AI policy, and I explain the difference between a tool that helps you edit your own words and one used to pass off someone else’s thinking.
• I keep samples of each student’s in-class writing from September, so I have a baseline for voice.

On the ethics

I do not think every use of these tools is cheating. A student who drafts their own essay and uses Grammarly’s voice feature to tidy it is in a very different place from one who pastes a chatbot answer into a humanizer and submits it. The tools themselves are neutral; the honesty question lives in what the student claims as their own work. The forum has had good discussions on this, and I would point new members to The Ethics of AI Text Humanizers in Education, A Conversation We’re Not Having Seriously and the case for and against banning them.

Bottom line

If you only have time to look at one tool, look at Walter Writes. It is the most capable of the group, with three rewrite levels, an academic tone, a built-in detector and 80+ languages (Walter Writes), which makes it the best illustration of where this technology is now. Spend ten minutes on its free sample with a paragraph of your own writing and you will understand why the conversation has to move from detection to assessment design.

What are you seeing in your schools? I am especially curious about French-language classrooms and about anyone whose board has put humanizers in writing in its AI guidance.