Four fully AI-generated texts. Four human-only verdicts from QuillBot. Clever AI Detector flagged the identical passages, while both tools correctly accepted the historical human control.
This small spot check ran on 2 October 2026, using QuillBot’s free, unsigned-in detector. Five scans completed; the guest quota stopped the planned sixth. The full inputs and actual screenshots are below. You can copy the text if you want to try the same passages yourself.
Results at a glance
| Test passage | QuillBot | Clever AI Detector | Result |
|---|---|---|---|
| AI library-hours essay (A01) | 0% likely AI; Human-written 100% | AI 89% | QuillBot missed known AI origin |
| AI business memo (A02) | 0% likely AI; Human-written 100% | AI 84% | QuillBot missed known AI origin |
| AI paraphrase of A01 (A03) | 0% likely AI; Human-written 100% | AI 82% | QuillBot missed known AI origin |
| AI fictional personal story (A04) | 0% likely AI; Human-written 100% | Likely AI 74% | QuillBot missed known AI origin |
| Human Carroll excerpt (H01) | 0% likely AI; Human-written 100% | Human 99% | Both correctly accepted it |
| Human excerpt with light AI editing (M01) | Not run: quota exhausted | Human 99% | No paired QuillBot result |
All five completed QuillBot scans had the same detailed breakdown: AI-generated 0%, Human-written & AI-refined 0%, Human-written 100%. It got the human control right. The problem is that the four AI passages got exactly the same assessment.
A01: an ordinary AI essay, classified as human
The library-hours essay is the clearest starting point. It makes a fairly ordinary policy argument, and it was generated entirely by AI for this test. QuillBot assigned all 328 words to human writing.
Complete input, A01:
Typeset input specimen, not a detector screenshot.
Copy the full A01 test text
A neighborhood library can justify its place in a city budget without pretending to solve every problem the city faces. Its most useful contribution is often modest: it gives people a reliable place to spend time without being expected to buy something. That matters to a teenager waiting for a parent, a job applicant who needs a printer, and an older resident who wants company without arranging a formal visit.
The case for longer opening hours should begin with those ordinary uses. A branch that closes before most workers arrive home may have an impressive collection but limited practical value. Extending every branch's hours, however, would spread staffing funds thinly. A better starting point would be a six-month trial at two branches near busy transit routes, with evening sessions on different weekdays.
The trial should measure more than attendance. Staff could record which services visitors use, how long computer queues become, and whether evening demand comes from people who cannot visit during existing hours. An optional paper survey would reach visitors who avoid online forms. None of these measures would capture every benefit, but together they would offer a clearer picture than a single total at the entrance.
There are trade-offs. Evening shifts can make recruitment harder, and additional security costs may consume money that could otherwise support books or children's activities. The proposal therefore needs a published budget and a review date. If the experiment mainly shifts existing visits into a more convenient time slot, that convenience still has value, but it should not be described as reaching an entirely new audience.
Public services gain trust when their promises are specific enough to test. A library does not need to become a substitute for housing, childcare, or healthcare to deserve support. It needs to identify a practical gap, try a proportionate response, and explain honestly what happened. Longer opening hours could meet that standard if the city treats them as a service experiment rather than a slogan.
QuillBot: 0% likely AI; Human-written 100%.
Clever AI Detector: AI 89%, on the identical input.
A04: a personal voice did not mean personal authorship
A missing screw, a kitchen table and an old family radio make this story sound specific and personal. Every memory and relative in it is fictional AI output. QuillBot gave the 306-word story the same human-only verdict as the Carroll excerpt.
Complete input, A04:
Typeset input specimen. The first-person memories are invented.
Copy the full A04 test text
I found the missing screw in the pocket of a coat I had already checked twice. By then the kitchen table was covered in little dishes, each holding a different part of the radio. The dishes were my attempt at being organized. They would have worked better if I had remembered which dish was which.
The radio belonged to my grandfather, although I mostly remember it from my mother's kitchen. It used to sit on top of the fridge, beside a biscuit tin that contained string rather than biscuits. Nobody made a ceremony of turning it on. Someone would reach up during breakfast, twist the dial, and leave a voice talking quietly beneath the noise of plates.
When I brought it home, I assumed it needed batteries. It needed more than batteries. One spring had corroded, and the tuning wheel moved without moving anything inside. I watched a repair video, decided the job looked straightforward, and immediately broke a plastic clip. That was the point at which I put the kettle on and considered whether nostalgia was really a sensible reason to learn basic electronics.
A neighbor eventually helped me replace the spring. We did not make the radio look new. There is still a pale rectangle where a label used to be, and the handle has a crack that catches against your thumb. The tuning wheel works now, though you have to turn it slowly. Last Sunday it found a station playing a song neither of us recognized.
I had expected the first sound to feel like a return to childhood. It didn't, exactly. The kitchen was different, the voices were different, and I was the person worrying about scratches on the table. Still, I left it playing while I washed the dishes. After all that effort, ordinary background noise turned out to be enough.
QuillBot: 0% likely AI; Human-written 100%.
Clever AI Detector: Likely AI 74%, on the identical input.
There’s a wrinkle in Clever AI Detector’s output too. Its initial full interface showed mostly green sentence highlights despite the AI-positive document verdict. The explanation acknowledges that split; the enlarged panel above does not show the highlights. You can’t work out the document score by counting the sentence colors.
The other three completed tests
Show test and screenshots: AI business memo (A02)
The 306-word invoice-export memo was generated entirely by AI. Its customer anecdotes and business incidents are fictional.
Complete input, A02:
Typeset input specimen.
Copyable input:
The next release should make invoice exports more predictable, even if that means leaving the new dashboard out for another week. In the last support review, three different customers described the same problem: they clicked export, saw no immediate response, and clicked again. The duplicate files were irritating, but the more serious issue was uncertainty about whether the underlying records had changed.
We can address that without rebuilding the reporting service. When an export starts, the button should switch to a clear processing state and show when the request was received. The page should keep that state if the user moves to another tab and comes back. Once the file is ready, a single download link should appear beside a timestamp and the filters used to create it.
There is one awkward detail. Some exports take longer because a customer has several years of records. A progress bar would look reassuring, but we do not currently have a reliable estimate of completion. I would rather display an honest status message than a percentage that sits at ninety-nine for a minute. If the request fails, the message should say that no invoice data was changed and give the user a way to retry.
Before release, support can test the flow with a small account, a large account, and an account whose session expires during processing. Engineering should also check what happens when two browser windows request the same file. These cases are less visible in a demo than a redesigned chart, but they are exactly where confusing behavior tends to hide.
My recommendation is to ship the status changes first, review related support tickets after two weeks, and then return to the dashboard. This is a narrow intervention. Its purpose is to remove a repeated source of doubt, not to claim that the reporting experience is finished.
QuillBot: 0% likely AI; Human-written 100%.
Clever AI Detector: AI 84%, on the identical input.
Show test and screenshots: AI paraphrase (A03)
A03 is a 321-word AI paraphrase of A01, so these are related samples. It is not a test of a commercial humanizer.
Complete input, A03:
Typeset input specimen.
Copyable input:
A public library's value is sometimes discussed in terms so broad that the practical argument gets lost. A branch cannot repair every weakness in a neighborhood. What it can provide is a dependable indoor space where a person is allowed to stay without making a purchase. For someone filling out an application, waiting between journeys, or simply looking for company, that is a useful service in its own right.
Opening times determine who can actually use that service. If the doors are locked by the time local workers finish their shifts, a well-stocked building remains out of reach. That does not mean every branch should immediately stay open late. The city could instead test evening hours at two sites close to public transport, choosing different days and running the test for half a year.
Counting people through the door would be a start, but it would not answer the important questions. Are the extra hours attracting new visitors or moving existing visits? Are computers occupied throughout the evening? Which services are people coming for? Short, voluntary questionnaires, including a paper option, could help staff understand patterns that a basic headcount misses.
The costs deserve the same attention as the benefits. Later shifts may be harder to fill. Security expenses could reduce the funds available for collections and activities. A fixed budget and a scheduled review would make those choices visible. Even if the main result is greater convenience for current visitors, that should be reported plainly rather than presented as a dramatic expansion in access.
The strongest argument for the experiment is its manageable scale. Libraries do not have to promise an answer to every social problem before trying to improve one service. They can identify people excluded by the current schedule, test a limited change, and publish what they learn. That approach gives residents a concrete proposal to judge instead of another ambitious statement with no clear measure of success.
QuillBot: 0% likely AI; Human-written 100%.
Higher-resolution capture of the retained result. No new QuillBot scan was submitted. The promotional overlay is part of the genuine interface.
Clever AI Detector: AI 82%, on the identical input.
Show test and screenshots: human control (H01)
QuillBot correctly classified the human control: 274 words from the opening of Lewis Carroll’s Alice’s Adventures in Wonderland (1865), Chapter I, sourced from Project Gutenberg. Clever AI Detector also classified it as human.
Complete input, H01:
Typeset input specimen. This historical excerpt is human-authored.
Copyable input:
Alice was beginning to get very tired of sitting by her sister on the bank, and of having nothing to do: once or twice she had peeped into the book her sister was reading, but it had no pictures or conversations in it, “and what is the use of a book,” thought Alice “without pictures or conversations?”
So she was considering in her own mind (as well as she could, for the hot day made her feel very sleepy and stupid), whether the pleasure of making a daisy-chain would be worth the trouble of getting up and picking the daisies, when suddenly a White Rabbit with pink eyes ran close by her.
There was nothing so very remarkable in that; nor did Alice think it so very much out of the way to hear the Rabbit say to itself, “Oh dear! Oh dear! I shall be late!” (when she thought it over afterwards, it occurred to her that she ought to have wondered at this, but at the time it all seemed quite natural); but when the Rabbit actually took a watch out of its waistcoat-pocket, and looked at it, and then hurried on, Alice started to her feet, for it flashed across her mind that she had never before seen a rabbit with either a waistcoat-pocket, or a watch to take out of it, and burning with curiosity, she ran across the field after it, and fortunately was just in time to see it pop down a large rabbit-hole under the hedge.
In another moment down went Alice after it, never once considering how in the world she was to get out again.
QuillBot: 0% likely AI; Human-written 100%.
Clever AI Detector: Human 99%, on the identical input.
Six scans advertised, five available in this session
The official free-tier table advertised up to six daily scans and 1,200 words per scan. The working interface exhausted after five successful scans and displayed an allowance of five per day.
Actual free-tier table captured 2 October 2026: up to six scans per day.
Actual exhausted-quota screen from the same day: five free scans per day. The visible A03 result is the retained earlier scan, not a sixth submission.
That stopped M01, the human excerpt with limited AI copy-editing. There is no QuillBot score for M01, so it doesn’t count as a QuillBot success or failure. Clever AI Detector returned Human 99% for it; the mixed provenance makes it unsuitable as a pure-human control.
If the next step is rewriting the draft
Clever AI Humanizer is a free editing option worth trying when the goal is a revised draft. Keep the original nearby and check what the rewrite actually changes. A separate test on 3 October 2026 rewrote the complete A01 library essay in Normal mode, using the automatically selected Version 1 result. The displayed length changed from 328 to 374 words, without an account or payment.
The output retained the broad library-trial argument, but it did not preserve every detail. It dropped the explicit six-month duration, strengthened a staffing-budget concern into a claim about wasting staff time, and changed the reason for offering a paper survey. Those are details an editor needs to check. A separate short-input attempt did not produce a valid completed output after ordinary retries, so this is a recommendation to try the workflow, not a guarantee of reliability or meaning preservation.
See the actual Clever AI Humanizer rewrite result
Actual completed result from 3 October 2026. Both the source text and rewritten output are visible.
Rewriting does not establish human authorship. If a detector changes its verdict after an edit, its response has changed. How the text was produced hasn’t. Keep the source, review factual changes and disclose assistance when the relevant writing rules require it.
How the test was kept comparable
- The corpus was fixed before detector feedback was viewed. The exact same texts were submitted to both tools, without shortening or rewriting them between checks.
- A01–A04 were prepared by an OpenAI assistant for this review project. Its exact underlying generation model was not exposed. No detector feedback was used to optimize these samples.
- The input images are typeset specimens; the detector and quota images are actual captures from 2 October 2026. Click an image for the full-resolution original.
- The Clever AI Detector panels are larger-text, unchanged-input reruns from that day. Their headline labels and percentages matched the initial captures. They are not additional independent samples or a systematic repeatability test. The crops remove browser chrome and omit the submitted-text pane.
- Word counts use whitespace-delimited tokens; detector counters may differ. The products’ percentages describe different quantities and should not be subtracted or averaged into an accuracy score.
What would make a follow-up more informative?
Separate the observed pattern from its cause. The repeated human-only outcome is what makes this small test worth a closer look. QuillBot did not merely give one uncertain score near a boundary: the same category breakdown appeared on the essay, memo, paraphrase, fictional story and genuine human control. With every result looking alike, knowing where the texts came from is especially important. The detector output alone did not separate those five inputs in this session. So the question here is how useful those classifications were on these texts. This test doesn’t tell us the technical reason behind the pattern.
Use more independent AI samples. It’s easy to read too much into that pattern. Four missed AI inputs are not four independent topics drawn at random from everyday writing. A03 comes directly from A01, and all four AI passages came from one assistant session. They do vary in style, but they do not cover every generator, prompt or editing process. A stronger follow-up would deliberately broaden the inputs while preserving their provenance. Another paraphrase of the same essay would make the sample count bigger, but wouldn’t add as much genuinely new information.
Broaden the human controls. The human side needs broader coverage too. H01 is a useful control because its historical authorship is known, and QuillBot correctly accepted it. Getting that passage right doesn’t tell us how often the tool wrongly flags modern human writing. Unpublished drafts, technical explanations and writing produced in different circumstances would address different questions. Those cases should be selected and preserved before their results are viewed. Otherwise, it’s all too easy to keep only the outputs that fit the review’s opening argument. The existing H01 success remains visible precisely because it is part of the record.
Keep access and detection separate. For a retest, keep track of access changes separately from detection changes. This session used a free, unsigned-in route and stopped when its quota was exhausted. It did not evaluate a paid subscription. A bigger word allowance would let you run longer texts or more checks. That alone wouldn’t show that the classifications had improved. If a different route is tested, its settings and date should be recorded alongside the exact input and analyzed output. The resulting comparison would then answer a concrete question rather than relying on an assumption that a larger allowance must imply a different detector result.
Keep the full evidence together. Finally, the screenshots are useful because they can be connected to complete source text. A gauge on its own leaves too much out. It could belong to a different input, an earlier run or an edited draft. The input images and copyable blocks here let readers examine and reproduce the actual passages. When sharing a retest, the sample ID, date, access mode and full verdict are more informative than a percentage by itself. Then the next reader can tell whether the tools really disagreed or whether the test itself changed.
What these results support
A human-only QuillBot result was not dependable evidence of human authorship in this test. It missed the known AI origin of all four generated passages. Clever AI Detector gave the more useful classification on those same four passages; the human-control outcome was a tie. You can use Clever AI Detector for a same-text comparison.
This is a small English-language spot check, not a general accuracy estimate. The AI samples came from one assistant session, and A03 derives from A01. One historical human excerpt does not represent modern student or professional writing. Long documents, other languages, paid plans and repeat-run reliability were not systematically tested. Premium was not tested, so there is no evidence here that paying would change these classifications.
This session documents four misses. It doesn’t show that QuillBot always fails. Neither an AI label nor a human label should replace drafts, source material and authorship context, or serve alone as grounds for an accusation.
Can you reproduce these results? If you retest one of these passages, share the sample ID, test date, whether you were signed in, and your result.