evidoria

← Back to browse

Good practice Imported

Copyleaks — AI content and plagiarism detection for academic integrity

Israel · Tel Aviv · See the Israel profile · See the Tel Aviv profile

Evidence: Observational / pre–post Top 72% 40/100 · Ask Evidence Copilot about this practice

Tel Aviv-based Copyleaks offers an AI-generated-text and plagiarism detector used by many schools and universities. Independent studies give conflicting accuracy pictures — near-perfect in one 2024 comparison, inconsistent elsewhere — a genuinely mixed evidence case.

0 %
False-positive rate in JALT 2024 comparative study (30 detectors tested) (April 2024 study)
~50 to 99+ %
Third-party comparative accuracy range (varies by sample/methodology)
50+ universities
Universities that disabled similar AI-detection features campus-wide (as of reporting)

Details

Maturity
Established
Promoter
Copyleaks Ltd.
Period
2012–present
Keywords
academic integrity, AI content detection, plagiarism detection, edtech

Context

Copyleaks Ltd, founded in 2012 in Tel Aviv by Alon Yamin and Yehonatan Bitton, provides an AI-generated-content detector and plagiarism-checking service that integrates with learning-management systems such as Canvas, Moodle and Blackboard, marketed for academic-integrity screening in K-12 and higher education worldwide.

Results

Independent academic testing of Copyleaks gives a genuinely mixed picture. An April 2024 comparative study in the Journal of Applied Learning and Teaching tested 30 AI-text detectors against university English L1 and L2 student essays and found Copyleaks among only two tools (with Undetectable AI) to record a 0% false-positive rate across the essay sets tested. However, a 2025 peer-reviewed systematic review in the journal Information (MDPI) concluded that AI-detection tools generally, Copyleaks included, produce inconsistent results across studies and content types, and separate third-party comparative tests have reported accuracy figures ranging from roughly 50% to over 99% depending on the sample and methodology used. More than 50 universities, including Vanderbilt and Johns Hopkins, have disabled similar AI-detection features campus-wide over false-positive and bias concerns.

Conclusions

Because detector accuracy is highly sample-dependent and vendor-published accuracy claims exceed what independent testing consistently reproduces, institutions are advised to treat Copyleaks results as a discussion starting point rather than proof of misconduct. This practice is included as an honestly-evidenced, mixed case rather than an unqualified endorsement.

Implementation

Indicative cost
Low (< €50k) — Commercial SaaS/LMS-integration licensing; no published per-institution pricing in sources reviewed.
Time to results
Short (< 1 year) — Founded 2012; ongoing commercial operation and wide-scale LMS integration since.
Staffing & skills
Copyleaks Ltd. (founded 2012 by Alon Yamin and Yehonatan Bitton, Tel Aviv) develops and operates the service; adopting institutions integrate it via existing LMS platforms (Canvas, Moodle, Blackboard)

Conditions for success

  • Integrates directly into existing LMS platforms, lowering adoption friction
  • Best used as a discussion starting point rather than sole proof of misconduct, per the accuracy evidence

Common failure modes

  • Independent third-party results range from near-perfect to roughly 50% accuracy depending on sample and methodology
  • More than 50 universities (including Vanderbilt and Johns Hopkins) have disabled similar AI-detection features campus-wide over false-positive and bias concerns
  • Vendor-published accuracy claims exceed what independent testing consistently reproduces

Where it fits

Governance type
commercial vendor, adopted by individual schools/universities
Scale
global, institution-by-institution adoption
Income level
mixed (deployed across income levels)

Commonly funded by

Own resources / municipal budget

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful