Test log
Decopy AI Detector test data
Raw results for the Decopy AI Detector 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
- 8
- Method
- Published methodology
- Environment
- Google Chrome on desktop
Results
| ID | Test | Key measurement | Score |
|---|---|---|---|
| T01 | Accuracy on AI-Written Text | AI passages caught: 1 of 1 | 5/5 |
| T02 | False Positives on Human Text | Human passages flagged as AI: 0 of 1 | 5/5 |
| T03 | Mixed & Lightly-Edited Text | Mixed passages flagged: 1 of 1 | 3/5 |
| T04 | Document & File Support | Ways in and out that worked: 6 of 10 | 3/5 |
| T05 | Score Clarity & Evidence | Points met: 8 of 10 | 4/5 |
| T06 | Scan Speed | Median per scan: 16.4 s | 3/5 |
| T07 | Pricing & Free-Tier Limits | Free credits a month: 10 | 3/5 |
| Overall (in the review) | 3.5/5 | ||
Accuracy on AI-Written Text
5/5Tested- Method
- One AI-written passage of 78 words, about remote work, was pasted and scanned once; the first verdict counted, with 50 percent AI Probability as the line.
- AI passages caught
- 1 of 1
- AI Probability
- 100%
- Remark
- Scale 5: 47 to 50 of 50 caught. 1 of 1 is 100 percent, so the grade is 5; the sample is one passage, not fifty, and the credits ran out before I could add more.
Raw notes
One AI-written passage of 78 words, about remote work, was pasted and scanned once; the first verdict counted, with 50 percent AI Probability as the line.
It scored 100% AI Probability with the Final Verdict AI Text, so it counts as caught. Every sentence was highlighted red and the Why this sentence was flagged note named Predictable Structure and Generic Claim.
- One passage from one model (Claude) at default settings, not 50 from six models. The Spanish AI passage was missed, which is reported under non-English.
False Positives on Human Text
5/5Tested- Method
- One passage from Pride and Prejudice, published in 1813, was pasted and scanned once, so any AI flag would be a false positive.
- Human passages flagged as AI
- 0 of 1
- AI Probability on the human passage
- 0%
- Remark
- Scale 5: at most 1 of 50 flagged. 0 of 1 is 0 percent, so the grade is 5. The sample is one passage, so the rate is a signal and not a measurement, and the grade has to be read with the mixed-text result below.
Raw notes
One passage from Pride and Prejudice, published in 1813, was pasted and scanned once, so any AI flag would be a false positive.
It scored 0% AI Probability, the Final Verdict was Human Text and the whole passage was highlighted green. The sentence list said No AI or Mixed sentences found.
- One passage, not fifty, and none in another language. The mixed-text scan also called a human paragraph AI, so I cannot say the detector never accuses a real writer.
Mixed & Lightly-Edited Text
3/5Tested- Method
- One passage joined a human paragraph from Great Expectations (74 words) to an AI paragraph (42 words) and was scanned once.
- Mixed passages flagged
- 1 of 1
- Human words highlighted as AI
- 74 of 74
- Remark
- Counted on the flag alone, 1 of 1 is the top line, but the scale also needs accurate highlights and these marked the human half as AI, so I gave 3.
Raw notes
One passage joined a human paragraph from Great Expectations (74 words) to an AI paragraph (42 words) and was scanned once.
It was flagged: 100% AI Probability and the Final Verdict AI Text, with 0% Mixed Text and 0% Human Text.
The highlight did not point at the AI part: the whole passage, including the human paragraph from Dickens, was marked red as one AI sentence. The AI Vocab tab then underlined nine phrases, all from the AI paragraph.
- One passage, not twenty. The report treated the text as a single sentence, so I could not check sentence-by-sentence marks.
Document & File Support
3/5Tested- Method
- I checked each of the ten ways into and out of the detector on the free plan: pasting, PDF, DOCX and TXT upload, batch, web address, a 2,000-word text, an extension, an API and an export.
- Ways in and out that worked
- 6 of 10
- Text limit stated
- 200,000 characters
- Remark
- Scale 3: 6 or 7 points. The files were loaded and read but not scanned, because of the credit limit; the pasted passages show how a scan reads.
Raw notes
I checked each of the ten ways into and out of the detector on the free plan: pasting, PDF, DOCX and TXT upload, batch, web address, a 2,000-word text, an extension, an API and an export.
Worked: pasting; a PDF, a DOCX and a TXT (each filled the box with the 545 characters of the passage); a 2,508-word text (17,094 characters against a stated limit of 200,000, accepted but not scanned); and an export, because the Download Report menu gave a PDF Report that downloaded and an Online Report with its own web address. That is 6 of 10 points.
Not found: a batch scan, a way to check a web address, a browser extension for the detector and a detector API.
- The paid plans may add points; I did not buy them. The files were plain, so how formatted essays read is not tested.
Score Clarity & Evidence
4/5Tested- Method
- Each of the ten points was checked on the scans already run.
- Points met
- 8 of 10
- Remark
- Scale 4: 8 or 9 of 10 points. A missed point because I did not test it counts as missed.
Raw notes
Each of the ten points was checked on the scans already run.
Met: a plain call (AI Text, Mixed Text or Human Text); a percentage; sentence highlighting; a readable explanation (Why this sentence was flagged and a Suggested Direction, plus an AI Vocab tab); a downloadable and shareable report (a PDF, and an online page kept for 30 days); a scan history (30 days or 100 records); a stated limit (the page says no AI detection tool can be 100% accurate); and no upsell blocking the result. That is 8 of 10.
Missed: a stated confidence band or margin, since the AI, Mixed and Human split is a mix and not a margin; and repeatability, because I did not scan one passage three times.
- The highlights met the point by being there, but the mixed passage shows they can mark human writing as AI.
Scan Speed
3/5Tested- Method
- Each of the four scans was timed from pressing Detect text to the verdict on screen.
- Median per scan
- 16.4 s
- Slowest scan
- 19.2 s
- Remark
- Scale 3: 10 to 19.9 seconds.
Raw notes
Each of the four scans was timed from pressing Detect text to the verdict on screen. The script timed the last two; the first two were read off the screen recording to the nearest half second.
The four scans took about 23 seconds (the first, an AI passage), about 5 seconds (the human passage), 13.6 seconds (the mixed passage) and 19.2 seconds (the Spanish passage). The median is 16.4 seconds. None failed.
The recording shows the box greyed out with a spinning circle for the whole wait.
- Four passages of 78 to 116 words on the free plan; the wait is not the same each time, and I did not time a long text.
Pricing & Free-Tier Limits
3/5Tested- Method
- I used the free plan until 80 percent of its monthly credits were gone, then read the pricing page on the test date in both billing views.
- Free credits a month
- 10
- Lite plan, billed monthly
- $9.99
- Remark
- Scale 3: a tight free tier, or a paid plan of $20 to $29.99. About five scans a month is tight, and the plan that makes detector scans unlimited is $29.99 a month.
Raw notes
I used the free plan until 80 percent of its monthly credits were gone, then read the pricing page on the test date in both billing views.
Free: 10 credits and 10 quotas a month, claimed on login. A scan cost 2 credits for each passage, so the free plan buys about five scans of 80 to 120 words a month. Quotas are for basic features and credits are for premium ones, and the AI content detector uses credits on the Lite and Basic plans.
Paid: Lite is $9.99 a month billed monthly with 800 credits, Basic is $19.99 with 2,000 credits and Premium is $29.99 with unlimited detector scans and 5,000 credits. The yearly view carried a Halloween sale of 50% off: $6.99, $13.99 and $20.99 a month.
- No plan was bought and checkout was not opened, so I do not know how many credits a long text costs.
Other features checked
Coverage
Tested
- Four pasted detector scans, the first verdict counted: one AI, one human, one mixed and one Spanish AI passage
- A PDF, a DOCX and a TXT loaded through Upload Document (loaded, not scanned)
- A 2,508-word text pasted into the box (accepted, not scanned)
- The AI Sentences and AI Vocab tabs, the History panel, the PDF report and the online report
- Time from pressing Detect text to the verdict, checked on the screen recording
- The pricing page read on the test date and the Trustpilot page read in the same browser
Not tested or limited
- Four scans, not 50 + 50 + 20 + 20 + 24 full passages, because the free plan gives 10 credits a month; each passage is 78 to 116 words, not 300
- Resistance to humanisers and AI model coverage were not graded: no third-party humaniser was bought and every AI passage was written by one model
- The non-English test was not graded: only one Spanish passage was scanned and it was an AI one, so there is no human passage in another language to check for a false flag
- The same passage was not scanned three times, so repeatability was not checked; the files and the 2,508-word text were loaded but not scanned
- No batch scan, web address check, browser extension or API was found on the site
- No plan was bought: the paid plans were read from the pricing page, not tried, so how many words a credit buys on a longer text is not known
Still to test
- The other 45 or so AI and 45 human passages on a paid plan
- Resistance to humanisers and paraphrasers
- AI model coverage across six models
- A full set of non-English passages, human and AI
- Repeatability on one passage scanned three times