The same words.
A fairer comparison.

Every ranked humanizer gets the same 93-word passage. We publish its first complete rewrite, explain the scores, and keep the original available for your own comparison.

What the score means

Our score is an editorial judgment of one tested workflow. It combines the quality of that rewrite with the tools available to use it. It is not a detector pass rate, a scientific benchmark across all writing styles, or a promise about your next result.

Each dimension receives a score from 0 to 10. The overall score is the sum of each dimension score multiplied by its weight, rounded to one decimal place. Tied overall scores share the same editorial standing; alphabetical ordering only makes the list stable.

The original test passage

This AI-generated passage describes a fictional library experiment. Its numbers and qualifications are deliberate. A rewrite must not turn a limited observation into proof of higher productivity.

A local library tested a four-day schedule for six weeks. It kept the same 40 weekly opening hours, closing Wednesdays and extending Monday and Thursday evenings. Visits rose from 1,200 to 1,320 a week, although Wednesday visitors had to change plans. Staff reported fewer interruptions, but productivity was not measured. Longer evenings may help people who work during the day; this does not prove every library should open fewer days. Before making the change permanent, the team will survey residents, compare costs and repeat the trial in winter, when travel habits may differ.

How we run a test

1. Open the current product and save its homepage. Record the date, selected plan, mode, language and any login requirement.
2. Paste the complete passage without edits. Use the free or default mode where it accepts the full sample. If a tool cannot handle the full passage, do not shorten it to produce a comparable score.
3. Run one rewrite and save the first complete output. A technical failure can be retried and recorded; an unattractive result is never discarded in search of a better score.
4. Capture the input and result on the real product page. Copy the complete output so readers can inspect text as well as screenshots.
5. Check the editing controls and pricing page. Keep features observed in the tested plan separate from capabilities advertised for a paid plan.
6. Read the rewrite against the original, assign each dimension score, and explain the specific changes behind the verdict.

Scoring anchors

9–10: Excellent for the tested task, with no material issue in that dimension. 7–8: Useful with minor cleanup or a modest limitation. 5–6: A mixed result that needs substantial attention. 3–4: Serious limitations. 0–2: The capability is unusable or unavailable in the evaluated workflow.

For meaning and facts, we check the six-week trial, unchanged 40-hour week, Wednesday closure, Monday and Thursday evenings, both attendance figures, the lack of a productivity measurement, the limited conclusion and the planned follow-up. Invented evidence or a reversed conclusion is a major error even when the sentence sounds smooth.

Readability covers grammar, punctuation and ease of comprehension. Natural voice covers rhythm, specificity and whether the rewrite sounds forced. Changing more words does not automatically earn more points.

Features, free plans and integrations

A Free plan tag means we completed the benchmark without payment. It does not mean unlimited words. No signup means our test completed without creating an account. Word and sentence highlighting are separate from the ability to click a sentence and rewrite it.

API, MCP and agent-skill tags require an identifiable official integration or documentation. Third-party wrappers are described separately. A documented integration is not labelled end-to-end tested unless we actually used it. No integration tag means it has not met this verification threshold.

For integrations, an unverified offering earns no availability points rather than being treated as proven absent. A documented humanizer API earns 5/10, documented API and MCP earn 8/10, and an additional product-specific agent skill can earn 10/10. Preview waitlists do not count. These scores concern documentation and availability, not live integration reliability.

Value considers the useful free allowance, the billing unit and whether limits are clear. We record monthly and annual billing separately and distinguish a monthly equivalent from an annual charge. Promotions are snapshots of the date checked.

What we do not infer

A vendor's “human score” is a claim from that vendor. It is not an independent authorship test. We do not infer that a rewrite will pass Turnitin, GPTZero or another detector from that number. We also do not claim plagiarism clearance or factual accuracy from a detector result.

One paragraph cannot represent every genre or language. We do not test long documents, paid modes or integrations by implication. The review states the plan and settings that produced the shown result.

Updates stay at the same address

Each tool has one profile under /tools/ and one review article under /reviews/, using its domain with dots replaced by hyphens as the slug. A product update triggers a revision of that review, with the new test date and relevant changes. We retain earlier test evidence internally rather than creating a fresh article for the same product.

Compare tool profiles, read the reviews or use the test kit yourself.