Is JustDone's AI detector legit or accurate? What is published
No published study in the index has tested JustDone's AI detector, so every accuracy figure on the record is JustDone's own: 80% overall detection accuracy, 98% on academic and scientific texts and a 10.3% error rate, per the detector page read 8 November 2026 (justdone.com/ai-detector). The vendor is GM Appdev Limited of Limassol, Cyprus, the privacy policy took effect on 25 June 2025 (justdone.com/legal/privacy-policy, read 8 November 2026), and the detector scans up to 15,000 words in one pass.
Whether that record makes JustDone legit depends on what the reader needs the word to mean. A real company publishes real terms and a real limit; no outside paper has put the tool on a shared sample. The table sets JustDone beside Scribbr and ZeroGPT, both free and both aimed at student writing, on the same eight fields from the index. This page is one of the questions answered from the index; the siblings ask whether a legit AI detector exists at all and whether any free AI detector has no limit.
| Field, read 8 November 2026 | JustDone | Scribbr | ZeroGPT |
|---|---|---|---|
| Input limit per check | Up to 15,000 words; at least 350 characters | 15,000 characters per detection on FREE; 100,000 to 350,000 on paid plans | |
| Free tier, vendor's words | No setup and no account needed for your first scan; quantity per period not published | Unlimited free AI checks, no sign-up required | $0 forever: 15,000 characters per detection, words allowance 1,250,000 /month |
| Vendor accuracy statement | 80% overall; 98% on academic and scientific texts; 10.3% error rate | "we can't guarantee 100% accuracy"; highest found 84% premium, 68% free | 98.4% detection accuracy |
| Vendor false positive statement | Under 1% | ||
| Study records in the index | 0 | 1 | 5 |
| Lowest published false positive rate, with base | 2% of 42 human papers (Walters 2023) | 0 of 18 human documents (Weber-Wulff 2023) | |
| Overall accuracy in a study, with base | 88% of 126 documents (Walters 2023) | 87% of 126 documents (Walters 2023); 32 of 54 (Weber-Wulff 2023) | |
| Retention statement | Personal data stored as long as reasonably necessary; text never sold or published | Submissions not stored or shared | Not used to train models; retained only as needed to operate |
What does JustDone publish about its AI detector?
JustDone publishes four kinds of fact on the pages the index read on 8 November 2026. The product: a text detector that accepts pasted text or a PDF, Word or TXT upload, checks up to 15,000 words in one scan, asks for at least 350 characters, and names ChatGPT, GPT-4, GPT-5, Claude, Gemini, Llama and DeepSeek among the generators covered (justdone.com/ai-detector). The company: GM Appdev Limited, Office 202, 2nd Floor, 15 Nafpliou Str., 3025 Limassol, Cyprus. The policy: section 11 of the privacy policy, effective 25 June 2025, keeps personal data "for as long as it is reasonably necessary", and the detector page adds that text is never sold or published (justdone.com/legal/privacy-policy). The claim: 80% overall, 98% academic, 10.3% error, under the label "Independently benchmarked".
The blog, updated 19 March 2026, gives the same evaluation in more detail: 80% on pure AI text, a 10.3% total error rate, a 67.5% full correct classification rate, a table with ArXiv at 100%, News at 100% and PubMed at 83.3%, and a statement that the detector flags 60% of AI-edited human drafts (justdone.com/ai-detector, read 8 November 2026, which links the post). All of those figures come from JustDone's own evaluation framework, as the blog describes. Prices are the one field the JustDone card leaves blank, because the pricing page renders its figures client-side (justdone.com/pricing, read 8 November 2026).
Where do JustDone's own pages disagree?
Three places. First, the academic figure: the detector page states 98% on academic and scientific texts, while the blog states "almost 100%" and lists PubMed at 83.3% in the same table that puts ArXiv at 100% (justdone.com/ai-detector, read 8 November 2026). The card uses the detector page figure because the detector page is the product claim, and quotes the blog beside it.
Second, the label: the detector page marks the 80% and 10.3% figures "Independently benchmarked", and the blog that explains them describes JustDone's own evaluation framework. The nine records in the studies index, each with a DOI or publisher link, are what the index counts as a test from outside the company, and none of them names JustDone.
Third, the scope: the detector page FAQ mentions analysing "images, text, keywords", and no image detection product exists on the site, so the card records text as the single modality. The page also shows a Turnitin estimate with the vendor's own disclaimer: "Estimation based on our own model analysis. JustDone is not affiliated with Turnitin, and institutional results may differ." The Turnitin card carries Turnitin's own published rate, under 1% false positives on over 700,000 pre-ChatGPT papers (Turnitin FAQ, read 8 November 2026).
What can a reader conclude, and what can they not?
A reader can conclude that JustDone is a registered company with a working detector, a stated 15,000-word limit, a published privacy policy and an openly stated error rate of 10.3%, which is more than several vendors in the index publish, and that JustDone says in its own words "Every AI detection result is a probability, not a verdict" (justdone.com/ai-detector, read 8 November 2026).
A reader cannot conclude how often JustDone flags human writing, because JustDone publishes no false positive rate; cannot set JustDone's 80% against Scribbr's 88% of 126 or ZeroGPT's 87% of 126 in Walters 2023 (DOI 10.1515/opis-2022-0158), because the figures were measured on different texts by different people; and cannot check the 10.3% error rate against anything, because the test set is JustDone's own. The methodology page explains why the index never places a vendor figure in the same sort as a study figure.
How does JustDone compare with detectors that have study records?
Scribbr is the nearer match: a free, unlimited detector aimed at students, with one study record. In Walters 2023, Scribbr classified 88% of 126 documents correctly, 100% of 42 GPT-3.5 papers and 69% of 42 GPT-4 papers, and called 2% of 42 human papers AI (DOI 10.1515/opis-2022-0158, read 8 November 2026); Scribbr's own page puts the ceiling for any free tool at 68% (scribbr.com, read 8 November 2026).
ZeroGPT matches JustDone's 15,000 figure, in characters rather than words, and has five study records: 32 of 54 documents with 0 false accusations in 18 human documents in Weber-Wulff 2023 (DOI 10.1007/s40979-023-00146-z), 87% of 126 in Walters 2023, 77 of 104 AI samples undetected in Perkins 2024 (DOI 10.1186/s41239-024-00487-w) and a 1.42% false positive rate at threshold 0.5 in RAID 2024 (DOI 10.18653/v1/2024.acl-long.674). ZeroGPT states a detection accuracy of 98.4% and a false positive rate under 1% (zerogpt.com, read 8 November 2026), and the five records show where the vendor figure holds and where the figure fails. A detector with five records can be read five ways; a detector with none can be read only in the vendor's words. The ranking by independent testing lists the eight detectors with a record, and the home board shows which AI detectors exist and who has tested them.
Questions about JustDone's AI detector
Is JustDone's AI detector accurate?
JustDone states 80% overall detection accuracy, 98% on academic and scientific texts and a 10.3% error rate, on the detector page read 8 November 2026 (justdone.com/ai-detector). The figures come from JustDone's own test set, and no study in the index has tested JustDone, so no outside figure exists to set beside them.
Is JustDone's AI detector free?
The first scan is free with no setup and no account, per the detector page read 8 November 2026 (justdone.com/ai-detector). The pricing page renders its figures client-side, so the JustDone card carries no plan table; Premium is described on the detector page as unlimited AI checks up to 15,000 words per scan.
Does JustDone's Turnitin estimate predict a Turnitin score?
No. JustDone says the estimate is "based on our own model analysis", that JustDone "is not affiliated with Turnitin", and that institutional results may differ (justdone.com/ai-detector, read 8 November 2026). Turnitin's own published figure is a false positive rate under 1% for documents with over 20% AI writing (Turnitin FAQ).
Sources
All read 8 November 2026.
- JustDone AI detector: https://justdone.com/ai-detector
- JustDone privacy policy: https://justdone.com/legal/privacy-policy
- JustDone pricing: https://justdone.com/pricing
- Scribbr AI detector: https://www.scribbr.com/ai-detector/
- ZeroGPT DeepAnalyse page: https://www.zerogpt.com/deep-analyse
- ZeroGPT pricing: https://www.zerogpt.com/pricing
- GPTZero blog, 25 October 2026: https://gptzero.me/news/best-ai-detectors
- Turnitin AI writing detection FAQs: https://guides.turnitin.com/hc/en-us/articles/28477544839821
- Walters 2023: https://doi.org/10.1515/opis-2022-0158
- Weber-Wulff et al. 2023: https://doi.org/10.1007/s40979-023-00146-z
- Perkins et al. 2024: https://doi.org/10.1186/s41239-024-00487-w
- Dugan et al. 2024 (RAID): https://doi.org/10.18653/v1/2024.acl-long.674
Revision log: 2026-11: page created
Related pages: How many AI detectors exist? The index count and every study's count, Which AI detectors share an owner? The ownership map from the vendor pages, AI detectors ranked by independent testing Two records added in the fifth wave on 9 November 2026, Pegoraro 2023 and Fleckenstein 2024, changed nine counts and none: Pegoraro 2023 tested ZeroGPT, GPTZero, Copyleaks, Originality.ai, Writer, Content at Scale (BrandWell), DetectGPT, the GPT-2 Output Detector and GLTR, lifting GPTZero to 12, the GPT-2 Output Detector to 10, Copyleaks to 7 and Originality.ai and ZeroGPT to 6; Fleckenstein 2024 tested 289 teachers and no software, so it counts for no card ([Fleckenstein 2024 record](/studies/fleckenstein-2024-teachers/)). Illuminarty, one of four cards added in the same wave, joined with its one image record from Ha and Passananti 2024, so 18 of 72 cards now qualify. and Questions about AI detectors, answered from the index.