WriteHuman AI Review: Pricing, Terms and Scores (2026)
Alex Halpin
11/12/2026
WriteHuman's homepage lists a "human score" of 98% for Copyleaks, 97% for Turnitin and 95% for Originality. Its terms of service say the company does not guarantee that any text will receive "any particular score, label, or result from any AI detection tool, including ours."
I read the homepage, pricing page, terms, privacy policy and FAQ titles on 10 November 2026, plus the App Store listing and the HumanizerBench leaderboard the homepage points to. I did not run my own scans for this review. What follows is what WriteHuman claims, what it asks you to accept, and where its own pages differ.
My view: WriteHuman has clearer pricing than most humanizers and its terms are unusually frank about accuracy. The same terms say your text trains its models with no way to opt out, and every payment is final.
What the homepage claims
Under the heading "Natural writing quality, verified by leading platforms", the homepage shows a human score per detector after WriteHuman processing: Copyleaks 98%, Turnitin 97%, ZeroGPT 97%, GPTZero 96%, Grammarly 96%, Winston AI 96%, Originality 95% and Pangram 94%. A separate figure gives a "97.5%" average human score.
The page does not say how these scores were measured, how many texts were used or when the scans were performed. It also states that it is "#1 humanizer on HumanizerBench". The footer of the HumanizerBench leaderboard reads "Operated by WriteHuman", showing that this ranking is from WriteHuman itself. The leaderboard displayed WriteHuman as first place with a score of 70.11, followed by StealthGPT with a score of 69.47 and last tested on 30 October 2026. Unfortunately, it was not possible to determine which detectors are being used by WriteHuman.
To its credit, the headline wording is softer than many of the articles in this category. I could find no use of the word "bypass" in the homepage text.
Pricing and limits
The pricing page has a comparison table, and the plan cards on the homepage agree with it on prices. The cards also list extras the table says the plan lacks.
| Plan | Monthly | Annual, per month | Requests a month | Words per request |
|---|---|---|---|---|
| Free | $0 | n/a | 3 | 250 |
| Basic | $20 | $12 | 80 | 600 |
| Pro | $29 | $19 | 200 | 1,200 |
| Ultra | $59 | $39 | Unlimited | 3,000 |
Annual billing works out to $144 for Basic, $228 for Pro and $468 for Ultra, as the cards show. The paid plans also differ in output variations (2, 3 and 5), in detection checks a month (160, 400, unlimited) and in access to its MCP server, which the table gives to Pro and Ultra.
Three details do not line up. The pricing table says the free plan has 3 requests a month, while the homepage says "Start with 5 free humanizations". The homepage cards list MCP access and priority support for Basic, which the table says it does not include. And the iPhone app sells different prices again, including "Ultra Weekly $12.99" and "Pro $26.99". Check which price you are paying before you subscribe on a phone.
The word "Unlimited" is defined in the terms. It "means the plan has no stated monthly allowance for that feature. It does not mean there are no limits."
What the terms say
These terms were last updated on 13 October 2026 and are the most useful document on the site.
On accuracy they are direct. They say "AI outputs and AI detection results are probabilistic and can be wrong", and that detection results "are not proof of who or what wrote a piece of text." The disclaimer section repeats that the company does not warrant any detection result, and that you use the services "AT YOUR OWN RISK."
On money, section 12 states: "All payments are non-refundable except where the law requires a refund." There are no refunds or credits for partial billing periods or unused words. Canceling takes effect at the end of your current billing period. Subscriptions will automatically renew at the then-current price; prices may change at any time.
On your text, section 8 gives WriteHuman a "perpetual, irrevocable, and transferable license" to your content and says "We use Your Content to train, evaluate, and improve our AI models." It adds that there is no setting to exclude your content. The privacy policy repeats this and says samples may be sent to third-party detection services to test its tools, without your name or email. Text sent through the developer API or MCP server is the exception: the policy says it is not saved or used for training.
Two more clauses matter. Acceptable use prohibits "academic dishonesty", which is an odd line for a product sold to students. And disputes go to individual arbitration under Virginia law, with a class-action waiver. The company is WriteHuman LLC, with an address in Midlothian, Virginia.
What users say
The App Store listing showed 4.8 out of 5 from 228 ratings when I read it on 10 November 2026, version 1.3.2 from 8 October. The homepage still shows an older 4.7 from 188 ratings. The reviews I could see were mixed in content. A teacher wrote that they keep catching text that GPTZero scored as "100% human", and another reviewer said text they wrote themselves was shown as AI.
Two reviews are not a measurement and app ratings mostly reflect the pleasantness of the app. These can still be a useful counterweight to the homepage that shows 94% to 98% across eight detectors without a method.
Who it is for
WriteHuman suits someone who wants predictable pricing, is fine with their text being used for training, and treats the detector scores as marketing. It is a poor fit for unpublished work, client documents or anything covered by a confidentiality agreement, unless you use the API route the policy exempts.
The honest way to judge any humanizer is to test it on your own text against the detector that your reader uses. Our guide to comparing AI humanizers explains how to do this and Are AI detectors accurate? explains why a single score from an AI detector doesn't necessarily show the humanizer's accuracy.
The short version
WriteHuman publishes clear plans, from a free tier with 3 requests up to $59 a month, and admits in its terms that detection results can be wrong. Its homepage scores have no method behind them, the leaderboard ranking comes from a site it operates, payments are non-refundable and your text trains its models. If you want to compare it with our tool, the WriteHuman alternative page lays out the differences, and the AI humanizer is where to try ours. For another tool read the same way, see our look at BypassGPT.
If you want the longer treatment:
Keep reading
- BypassGPT Review: Does It Work? Refunds and Terms (2026)
BypassGPT's homepage promises 100% undetectable text and a 99%+ human score. Its terms promise neither. I read the pricing, refund and licence pages and compared them.
- Phrasly AI Review (2026): Pricing, Terms and a 99.7% Claim
Phrasly's pricing page cites a 99.7% human score. Its terms say it cannot guarantee passing any detector. Here is what both pages say, and what an outside test found.
- Clever AI Humanizer Review: Free, Fast, and It Changed My Facts
Clever AI Humanizer rewrites 3,000 words at a time, free and without an account. On the fixed sample it moved an independent detector less than four points, and changed what the text actually said.
Humanize AI Text and Improve Your Writing Right Now
Rewrite for clarity, flow, and readability while keeping your original meaning and writing style.