HumanPen and WriteHuman: document workflows and monthly free usage
Compare how often free usage renews, how much text each request accepts and whether the output retains document formatting.
Competitor details verified 2026-08-01. Their own site is the authority.
The short answer
WriteHuman has one of the more genuine free tiers in the category - 3 requests a month at 250 words, recurring rather than a one-off trial - so it is easy to try. But it caps words per request, so a long paper still has to be split. HumanPen has no per-request limit and works on the DOCX directly, at the cost of having no genuinely usable free tier.
Side by side
| HumanPen | WriteHuman | |
|---|---|---|
| Input | Upload the document itself | Paste text |
| Output | The same file back, still editable | Plain text |
| Per-request word cap | None | 600 / 1,200 / 3,000 words by tier |
| Free tier | 100 credits (1,000 words), once | 3 requests x 250 words, every month |
| Word-count control | Set a target word range; output converges on it, slight drift possible | None |
| Billing | Metered credits, never expire | Monthly subscription ($12 / $18 / $36 annual) |
| Detection report import | Turnitin / iThenticate | None |
| Built-in detector | None | Yes |
| Agent / API integration | Skill, MCP, and REST API — callable from AI agents | API and MCP |
| Rewrite languages | English only | Multiple |
When to pick WriteHuman
- You want to use it free for a while before deciding. Their free tier recurs monthly; ours is a one-time 100 credits.
- You want an instant score after rewriting. They bundle a detector. We do not - our approach is importing the report your institution issued, which helps once you have one but is less convenient for a quick self-check.
- Your text is short anyway. If your pieces run a few hundred words, the per-request cap never binds and our advantage does not exist for you.
- You need a language other than English.
On HumanizerBench, and where we stop
WriteHuman operates the HumanizerBench leaderboard and discloses this on every page of that site. In the August 2026 cycle, WriteHuman ranked first overall, at 76.69.
We mention it without implying anything. Every input sample, humanized output, raw verdict from all five detectors, and the scoring script are open on GitHub and locally reproducible, and a commit-reveal scheme stops participants pre-tuning against upcoming prompts. Mechanically it is the best public head-to-head data in this category.
We are not on that leaderboard, and no rule we can find on their site excludes us. Their methodology page says they buy each tool’s entry-level monthly plan and test it through its public interface; we sell metered credits against a file upload, and our own reading is that this makes us an awkward fit rather than an ineligible one. Either way there is no head-to-head data between us and WriteHuman, and we are not going to invent one here.
Our free tier is genuinely worse than theirs. If your standard is "try it free until satisfied, then pay", that is a disadvantage and there is no arguing it away. Our shape wins in one situation: a complete, formatted, reasonably long document.
FAQ
Why are you not on HumanizerBench?
No rule on their site rules us out. What their methodology page describes is buying each tool’s entry-level monthly plan and testing it through its public interface, and we are a file upload billed by metered credits — our own reading is that we are an awkward fit for that setup, not an excluded one. They also take submissions from anyone, with no promise of inclusion.
So how effective are you?
We have no third-party head-to-head data, so we will not give you a number. The honest advice is to try it on your own writing - the 100 credits at signup cover a full section of real body text.
Why is your free tier so much smaller?
Our per-job cost is higher: we process a whole document, protect formatting and citations, and hold context across slices. Giving away a full paper is not viable. 100 credits (1,000 words) is enough to judge rewrite quality.
Can I use both?
Yes. Short text there, long documents here, is a reasonable split.
Compare the results on your own document
Use your signup credits to try a small selection of paragraphs. Upload a copy, then compare the returned file with the original.
Try it on your document