evidoria

← Back to browse

Good practice Imported

South African Universities Switch Off Turnitin's AI Detector Over Bias and False-Positive Risks

South Africa · Cape Town · See the South Africa profile

Facing evidence that AI-text detectors misclassify non-native English writing, South Africa's UCT, Stellenbosch, UFS and Wits disabled Turnitin's AI detector in 2025-26, replacing it with redesigned, process-based assessment instead of automated policing.

South African Universities Switch Off Turnitin's AI Detector Over Bias and False-Positive Risks

Details

Promoter
University of Cape Town
Period
2023-2026 (Stanford study 2023; university decisions Oct 2025 - Jul 2026)
Keywords
higher education, academic integrity, AI-text detection, assessment policy

Description

The University of Cape Town (UCT) disabled Turnitin's AI-writing detection feature in October 2025, followed by Stellenbosch University at the end of 2025 and the University of the Free State (UFS) from 1 July 2026; the University of the Witwatersrand (Wits) had never enabled the tool. UCT's Sukaina Walji, director of the Centre for Innovation in Learning and Teaching, framed the move as shifting focus "from surveillance" toward learning; Stellenbosch's Dr Hanelie Adendorff called automated detection an "enforcement illusion"; UFS's Prof Francois Strydom said the university would not "outsource academic judgment to inaccurate, automated systems."

The decisions cite a peer-reviewed 2023 Stanford study finding that seven leading AI-text detectors flagged 61% of authentic essays written by non-native English speakers as AI-generated, with 97% of such essays flagged by at least one tool. Rhodes University academics documented further concrete failures: one student had 30% of a human-written submission flagged as AI-generated, Turnitin and GPTZero disagreed on which portions of the same essay were AI-written, and both tools flagged a professor's own assignment questions as AI-generated.

In place of automated detection, the universities are moving toward oral presentations and defences, reflective journals, iterative drafting with in-class writing components, and explicit teaching about appropriate AI use, reasoning that these process-based approaches catch misuse without punishing students for their writing style.

This is a corrective, harm-avoidance practice rather than a proven learning intervention: no institution reports outcome data showing improved integrity or learning under the new regime, and North-West University's retained detector (claimed 98% accuracy, under 1% false positives) shows South African institutions have not reached consensus on the underlying evidence.

Read the full analysis: https://www.itweb.co.za/article/sa-universities-move-beyond-ai-detection-tools/Pero3MZ3oV1qQb6m

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

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.

Attachments

Similar practices you may find useful