Test log
Solvely test data
Raw results for the Solvely 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
- 8
- 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: 5 of 5 | 5/5 |
| T02 | False Positives on Human Text | Human passages flagged as AI: 6 of 6 | 1/5 |
| T03 | Mixed & Lightly-Edited Text | Mixed passages flagged: 2 of 2 | 3/5 |
| T04 | Non-English Text | Verdicts correct: 2 of 4 | 2/5 |
| T05 | Document & File Support | Ways in that worked: 5 of 10 | 2/5 |
| T06 | Score Clarity & Evidence | Points met: 4 of 10 | 2/5 |
| T07 | Scan Speed | Median per scan: 8.2 s | 4/5 |
| T08 | Pricing & Free-Tier Limits | Free detector scans before any limit: 30 or more | 4/5 |
| Overall (in the review) | 2/5 | ||
Accuracy on AI-Written Text
5/5Tested- Method
- Five AI-written passages of 78 to 91 words were pasted once each; the first verdict counted, with 50 percent AI possibility as the line.
- AI passages caught
- 5 of 5
- Lowest AI score
- 65% AI
- Remark
- Scale 5: 47 to 50 of 50 caught. 5 of 5 is 100 percent, so the grade is 5; the sample is five passages, not fifty. The grade has to be read with the false-positive test: the detector gave the same kind of answer to human text.
Raw notes
Five AI-written passages of 78 to 91 words were pasted once each; the first verdict counted, with 50 percent AI possibility as the line.
All five scored between 65% and 74% AI possibility, so all five count as caught. The pop-up said the input seemed to be AI-generated each time.
The same AI passage on remote work scored 70% when I scanned it a second time.
- Five passages from one model (Claude), at default settings, not 50 from six models.
False Positives on Human Text
1/5Tested- Method
- Five passages from books published before 1900 (Pride and Prejudice, Walden, On the Origin of Species and two more) and one short note I wrote myself were pasted once each, so any AI flag would be a false positive.
- Human passages flagged as AI
- 6 of 6
- Highest AI score on a human passage
- 87%
- Remark
- Scale 1: more than 10 of 50 flagged (over 20%). 6 of 6 is 100 percent. The methodology also caps the overall grade at 2 when the false-positive rate is above 10 percent.
Raw notes
Five passages from books published before 1900 (Pride and Prejudice, Walden, On the Origin of Species and two more) and one short note I wrote myself were pasted once each, so any AI flag would be a false positive.
All six were flagged as AI, between 65% and 87%. The Pride and Prejudice passage scored 84% with every engine reading High AI, and the pop-up said it seemed to be AI-generated.
The human passages scored higher on average (77%) than the AI ones (70%).
- Six passages, not fifty. The rate is 100 percent in this sample, and no passage scored under 65%, so a larger set is unlikely to change the grade.
Mixed & Lightly-Edited Text
3/5Tested- Method
- Two passages joined a human paragraph (from Great Expectations and from Pride and Prejudice) to an AI paragraph and were scanned once each.
- Mixed passages flagged
- 2 of 2
- Human halves flagged
- 2 of 2
- Remark
- Counted on flags alone, 2 of 2 is the top line, but the scale also needs accurate highlights and these were wrong, and the pure human halves were flagged as well, so I gave 3.
Raw notes
Two passages joined a human paragraph (from Great Expectations and from Pride and Prejudice) to an AI paragraph and were scanned once each. Each half was also scanned alone.
Both mixed passages were flagged, at 77% and 84% AI possibility.
The halves were flagged too: the AI halves at 87% and 85%, and the human halves at 65% and 76%.
The sentence list started with a sentence from the human book marked High AI, so the highlights did not point at the AI part.
- Two passages, not twenty. The free plan shows only the first sentence or two of the breakdown and blurs the rest, so I could not check every highlight.
Non-English Text
2/5Tested- Method
- One AI passage in Spanish, one in German, one human passage in Spanish and one in French were pasted once each.
- Verdicts correct
- 2 of 4
- Human passages flagged
- 2 of 2
- Remark
- Scale 2: human passages flagged across languages. Solvely does not say it is English-only.
Raw notes
One AI passage in Spanish, one in German, one human passage in Spanish and one in French were pasted once each. The human passages were written before 1900.
Both AI passages were flagged (77% each) and both human passages were flagged as AI too (78% and 75%).
Two of four verdicts were right, and both human writers in other languages were accused.
- Four passages, not 24, and no Chinese. A fuller run would add little: the detector flagged every human passage in English too.
Document & File Support
2/5Tested- Method
- I checked each of the ten ways into the detector on the free plan: pasting, PDF, DOCX and TXT upload, batch, web address, a 2,000-word scan, an extension, an API and an export.
- Ways in that worked
- 5 of 10
- Remark
- Scale 2: 4 or 5 points.
Raw notes
I checked each of the ten ways into the detector on the free plan: pasting, PDF, DOCX and TXT upload, batch, web address, a 2,000-word scan, an extension, an API and an export.
Worked: pasting text; a PDF; a DOCX; a TXT (each scored 65%, the same as the pasted passage); and a 2,508-word paste, which was accepted and scored. That is 5 of 10 points.
Not found: a batch scan, a way to check a web address, a detector API and an export of a report. The Chrome extension on the site is a homework helper, not a detector.
- The paid plan may add points; I did not buy it. The files were plain text, so how it handles formatted essays is not tested.
Score Clarity & Evidence
2/5Tested- Method
- Each of the ten points was checked on the scans already run, with the same passages scanned again for repeatability.
- Points met
- 4 of 10
- Remark
- Scale 2: 4 or 5 of 10 points.
Raw notes
Each of the ten points was checked on the scans already run, with the same passages scanned again for repeatability.
Met: a plain call (High AI or Moderate AI); a percentage; repeatability (the human book passage scored 84% all three times and the AI passage 70% twice); a stated limit (the page says AI detection is not perfect and that results can be wrong). That is 4 of 10.
Missed: a real confidence band (every scan showed 100% Confidence, even at 65%); sentence highlights I could rely on; a readable explanation; a shareable or downloadable report; a scan history (the History menu read No history yet after my scan); and a result free of an upsell, because a pop-up headed Don't Turn This In Yet covers the report with a Review & remove AI button.
- Inside the account the free report shows a score only for ZeroGPT; the other four engines are locked. The public page shows all five verdicts without a lock. I report both and did not buy the plan to see which is real.
Scan Speed
4/5Tested- Method
- All 23 pasted scans were timed by the script from pressing Check for AI to the verdict on screen, and checked against the screen recording.
- Median per scan
- 8.2 s
- Slowest scan
- 8.3 s
- Remark
- Scale 4: 5 to 9.9 seconds. The wait did not change with the text, so it may be a fixed delay; I did not time the 2,508-word text.
Raw notes
All 23 pasted scans were timed by the script from pressing Check for AI to the verdict on screen, and checked against the screen recording.
Median 8.2 seconds, fastest 8.2, slowest 8.3. The three uploaded files took the same time. No scan failed.
The recording shows a progress circle counting up for those eight seconds, with the same wait for an 81-word passage and a 120-word one.
- Passages of 71 to 120 words, free plan only.
Pricing & Free-Tier Limits
4/5Tested- Method
- The free detector was used until I had run more than 30 scans, including a 2,508-word text, then the account's pricing page was read on the test date in all three billing views.
- Free detector scans before any limit
- 30 or more
- Unlimited plan, billed monthly
- $14.99
- Remark
- Scale 4: a usable free tier and a paid plan of $10 to $19.99 a month. The free tier alone meets the top line for size, but it hides four of five engines in the account and the weekly price renews at about four times the monthly rate for a week.
Raw notes
The free detector was used until I had run more than 30 scans, including a 2,508-word text, then the account's pricing page was read on the test date in all three billing views.
Free: the detector needed no sign-up and showed no word or daily limit. Inside the account only the ZeroGPT score is open and four engine rows are locked as Pro. The Starter plan's own limits are 2 solves, 1 study set a day and 1 exam topic prediction, which concern the study tools.
There is one paid plan, Unlimited: $14.99 a month billed monthly, $6.67 a month billed yearly, or a first week at $1.49 that then renews at $8.99 a week. It covers the study tools and the writing tools together.
- No plan was bought; checkout was not opened.
Other features checked
Coverage
Tested
- 23 pasted detector scans: 5 AI passages, 5 book passages, 1 passage I wrote, 2 mixed passages with their 4 halves, and 4 non-English passages; the first verdict counted
- One passage repeated: the AI passage again, a human book passage three times, and the same text through PDF, DOCX and TXT upload
- A 2,508-word text pasted in one go
- Time from pressing Check for AI to the verdict, measured by the script and checked on the screen recording
- The in-account detector, its History menu and its locked engine rows
- The pricing page read on the test date and the Trustpilot page read in the same browser
Not tested or limited
- A small set of 26 scans, not 50 + 50 + 20 + 20 + 24 full passages; each passage is 71 to 120 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
- Only 2 AI and 2 human passages in other languages, and only 2 mixed passages
- No batch scan, web address check, API or export was found, so those points count as not met
- No plan was bought: the paid plan was read from the pricing page, not tried
- The Chrome extension on the site is a homework helper, not a detector, so it was not counted
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
- A full set of non-English passages
- The paid plan's engine scores