Test log
WriteHuman test data
Raw results for the WriteHuman review. The verdict is in the review.
- Tested
- 6 November 2026
- Tester
- Miles Foster
- Account
- Free demo account, signed up by RankDude, nothing paid to the brand
- Tests
- 7
- Screenshots
- 7
- Method
- Published methodology
- Environment
- Google Chrome on desktop
Results
| ID | Test | Key measurement | Score |
|---|---|---|---|
| T01 | Accuracy on AI-Written Text | AI passages caught: 3 of 3 | 5/5 |
| T02 | False Positives on Human Text | Human passages flagged as AI: 0 of 3 | 5/5 |
| T03 | Mixed & Lightly-Edited Text | Mixed passages the headline flagged: 0 of 1 | 3/5 |
| T04 | Document & File Support | Ways in that worked: 1 of 10 | 1/5 |
| T05 | Score Clarity & Evidence | Points met: 7 of 10 | 3/5 |
| T06 | Scan Speed | Median per scan: 1.0 s | 5/5 |
| T07 | Pricing & Free-Tier Limits | Free detector checks a month: 10 | 3/5 |
| Overall (in the review) | 4/5 | ||
Accuracy on AI-Written Text
5/5Tested- Method
- Three AI-written passages of 78 to 82 words (remote work, the Industrial Revolution, sleep) were pasted once each; the first verdict counted, with 50 percent AI probability as the line.
- AI passages caught
- 3 of 3
- Lowest AI score
- 100% AI
- Remark
- Scale 5: 47 to 50 of 50 caught. 3 of 3 is 100 percent, so the grade is 5; the sample is three passages, not fifty.
Raw notes
Three AI-written passages of 78 to 82 words (remote work, the Industrial Revolution, sleep) were pasted once each; the first verdict counted, with 50 percent AI probability as the line.
All three scored 100% AI probability with the label Likely AI-Generated and every sentence highlighted red.
A fourth scan, of the remote work passage again, also scored 100%.
- Three passages from one model (Claude), at default settings, not 50 from six models.
False Positives on Human Text
5/5Tested- Method
- Three passages from books published before 1900 (Pride and Prejudice, Walden and On the Origin of Species) were pasted once each, so any AI flag would be a false positive.
- Human passages flagged as AI
- 0 of 3
- Highest AI score on a human passage
- 0%
- Remark
- Scale 5: at most 1 of 50 flagged. 0 of 3 flagged is 0 percent.
Raw notes
Three passages from books published before 1900 (Pride and Prejudice, Walden and On the Origin of Species) were pasted once each, so any AI flag would be a false positive.
All three scored 0% AI probability, labelled Likely Human-Written, with green highlights and no sentence marked as AI.
- Old literary text is far from the student and web writing that detectors wrongly accuse; a clean result here is a minimum. Three passages, not fifty.
Mixed & Lightly-Edited Text
3/5Tested- Method
- One passage joined a human paragraph (from Great Expectations) to a two-sentence AI paragraph about project planning, and was scanned once.
- Mixed passages the headline flagged
- 0 of 1
- Headline score
- 0% AI
- Remark
- A passage counts when the detector flags it as containing AI rather than passing it clean. The headline passed it clean, so it is a miss on the verdict; the colours were right, which is why the grade is 3 and not lower.
Raw notes
One passage joined a human paragraph (from Great Expectations) to a two-sentence AI paragraph about project planning, and was scanned once.
The report said 0% AI probability and Likely Human-Written, with 100% human in the bar.
The highlights did land on the right part: the human sentences were green and the two AI sentences orange.
- One passage, not twenty. The AI-draft-with-human-edits kind was not run.
Document & File Support
1/5Tested- Method
- I looked at every way into the detector on the free plan: the input box, the History page and the pages for the extension and API.
- Ways in that worked
- 1 of 10
- Remark
- Scale 1: 3 or fewer points, or paste-only with a tight cap.
Raw notes
I looked at every way into the detector on the free plan: the input box, the History page and the pages for the extension and API.
Worked: pasting text (and typing it). That is 1 of 10 points.
The input box offers Paste and Try a Sample and no upload, no batch and no web address. The longest passage I scanned was 116 words, and the Usage page lists a 250-word limit per request on the free plan.
Not counted: the Chrome extension and the REST API exist but are built around the humaniser and were not run end to end; no export of a scan was found, only a Copy button on a saved scan.
- A 2,000-word scan was not tested and the extension and API were not run, so a paid plan could add points.
Score Clarity & Evidence
3/5Tested- Method
- Each of the ten points was checked on the scans already run, with one passage scanned twice for repeatability.
- Points met
- 7 of 10
- Remark
- Scale 3: 6 or 7 of 10 points.
Raw notes
Each of the ten points was checked on the scans already run, with one passage scanned twice for repeatability.
Met: a plain label (Likely AI-Generated or Likely Human-Written); a percentage; sentence highlighting; a scan history on the account; repeatability (the same passage scored 100% AI twice); a stated limit (the page says accuracy falls on short, paraphrased or mixed text); no upsell blocking the result. That is 7 of 10.
Missed: a margin or confidence band, a readable explanation of why a sentence was marked, and a download or share report that I could find on a free scan.
- Repeatability used two scans, not three, to stay inside the allowance. The saved scan's note says humanising the text will bring the score down, which points to the paid humaniser but does not block the result.
Scan Speed
5/5Tested- Method
- All 8 scans were timed from pressing Detect AI to the report on screen.
- Median per scan
- 1.0 s
- Slowest scan
- 1.1 s
- Remark
- Scale 5: a median under 5 seconds. The recording is accurate to about 0.2 seconds, which does not change the grade.
Raw notes
All 8 scans were timed from pressing Detect AI to the report on screen. The clock came from my screen recording at ten frames a second, because my first timing script read the page too early and its figures were thrown away.
Median 1.0 second, fastest 0.7, slowest 1.1. No scan failed or queued.
- Passages of 78 to 116 words; a long document may take longer. Free-plan speed only.
Pricing & Free-Tier Limits
3/5Tested- Method
- The free plan was used until I reached 80 percent of the allowance, then the Usage page and the public pricing page were read on the test date.
- Free detector checks a month
- 10
- Cost of a check on Basic, billed yearly
- about 8 cents
- Remark
- Scale 3: a tight free tier. The cheapest paid plan is in the $10 to $19.99 band billed yearly but $20 billed monthly, so I keep the grade at 3.
Raw notes
The free plan was used until I reached 80 percent of the allowance, then the Usage page and the public pricing page were read on the test date.
Free: 10 AI detector checks a month, not a daily allowance, plus 3 humaniser requests and a 250-word limit per request. It needed a Google sign-in and no card.
The cheapest paid plan, Basic, is $12 a month billed yearly or $20 billed monthly, for 160 detector checks a month and up to 600 words a request. That is about 8 cents a check billed yearly.
Pro is $19 a month billed yearly ($29 monthly) with 400 checks and 1,200 words. Ultra is $39 billed yearly ($59 monthly) with unlimited checks and 3,000 words.
- No plan was bought; checkout was not opened. Every plan includes the humaniser, so the price covers more than the detector.
Coverage
Tested
- 8 detector scans: 3 AI passages, 3 human passages from books published before 1900, 1 mixed passage and one repeat
- Time from pressing Detect AI to the report, measured on the screen recording
- The History page, a saved scan, and the Usage page
- The input box, to see which ways in it offers
- The pricing page read on the test date
- The Trustpilot page read in the same browser
Not tested or limited
- The free plan's 10 checks paid for 8 passages of 78 to 116 words, not 50 + 50 + 20 + 20 + 24 full passages
- Resistance to humanisers, AI model coverage and non-English text were not graded: no third-party humaniser was bought, every AI passage was written by one model, and the checks ran out
- Every human passage is from an old book, which is easier to tell from AI than a modern essay
- Only one mixed passage was scanned
- The Chrome extension and the API were read about but not installed or called, and the share card was not found on a free scan
- No plan was bought: the paid plans were read from the pricing page, not tried
Still to test
- The other 40 or so AI and human passages on a paid plan
- Resistance to humanisers and paraphrasers
- AI model coverage across six models
- Non-English passages
- A full-length essay and the Chrome extension and API