Detection-centered academic integrity regimes are unreliable and potentially biased, making sole reliance on AI detectors unsuitable for high-stakes decisions¶

Claim · All claims
Evidence · 1 study (2 entries) · review r2 · q2
Currency · newest study 2025 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person · recheck by 2028: tool and model accuracy changes quickly

Subclaims¶

q2 i? OpenAI's classifier showed meaningful false positives (9% on its challenge set) and was discontinued due to low accuracy. → Sangwa 2025 q2 i? GPT detectors can be biased against non-native English writers, raising equity and due-process concerns for multilingual students. → Sangwa 2025 (2)

Evidence¶

Sangwa 2025¶

Sangwa, S., Ndahayo, C., Dusengumuremyi, F., & Mutabazi, P. (2025). Faculty Readiness for AI-Supported Teaching and Scalable Online Program Delivery in Higher Education: The EPIQ-AI Framework for Epistemic Integrity. https://doi.org/10.35542/osf.io/kj4vt_v1

q2 · i? · review · r2

The synthesis reports, citing OpenAI's 2023 classifier documentation, that "meaningful false positives (9% on its challenge set)" were observed and the tool "was discontinued due to low accuracy". Documented tool-evaluation evidence; no effect size printed.

"OpenAI's own classifier documentation reported meaningful false positives (9% on its challenge set) and noted that the tool was discontinued due to low accuracy"

Sangwa 2025 (2)¶

Sangwa, S., Ndahayo, C., Dusengumuremyi, F., & Mutabazi, P. (2025). Faculty Readiness for AI-Supported Teaching and Scalable Online Program Delivery in Higher Education: The EPIQ-AI Framework for Epistemic Integrity. https://doi.org/10.35542/osf.io/kj4vt_v1

q2 · i? · review · r2

The synthesis reports, citing Liang et al. (2023), that "GPT detectors can be biased against non-native English writers", an open-access analysis the article uses to ground equity and due-process concerns. Second-hand attribution; no effect size printed.

"GPT detectors can be biased against non-native English writers, raising equity and due-process concerns for multilingual and international students"

Discussion¶

Related Claims¶

• Universities disabling AI detection cite false-positive risk at scale, and detector influence persists in individual grading, journal screening, and institutions without policy revision — related
• AI-detection tools produce inequitable outcomes, disadvantaging non-native English speakers through false positives — related
• AI text detectors can return high AI-detection scores for human-written academic text, raising fairness concerns about authenticity evaluation — related
• AI detectors are reported as linguistically biased, unreliable, and psychologically damaging in student discourse — related
• Independent evaluation found commercial AI-text detectors neither sufficiently accurate nor sufficiently reliable for high-stakes use — possibly the same claim (merge candidate)
• AI-content detection tools achieve only marginally higher success than human reviewers, around 80% versus 78.4% — related
• AI-text detectors misclassify authentic TOEFL essays by non-native English speakers at a mean false-positive rate of 61.3% across seven detectors, far more often than native-speaker essays — a narrower finding that bears on this claim
• GPT detectors frequently misclassify non-native English writing as AI-generated, raising fairness concerns for AI policy — related
• Detection-led responses to AI in assessment face validity, fairness and trust limits — a broader claim this one bears on
• AI-generated-text detectors are reported as ineffective against current sophisticated language models, and students can alter text to evade detection — a narrower finding that bears on this claim
• A majority of universities with faculty guidelines discourage reliance on AI detection tools due to unreliability and bias — a narrower finding that bears on this claim
• OpenAI's GPT-2 Output Detector distinguished real student essays from ChatGPT-generated essays on the same subjects in a December 2022 trial — related
• AI detectors using perplexity disproportionately misclassify non-native English writers, and the mechanism plausibly extends to other low-perplexity writers — related
• The stylometric detector misclassified 4 of 18 independently authored documents as GPT-assisted (FPR 22.2%) and 2 of 18 GPT-assisted documents (FNR 11.1%), with wide confidence intervals — a narrower finding that bears on this claim