HOW AI TEXT REFINEMENT TOOLS HAVE NULLIFIED TURNITIN'S DETECTION ARCHITECTURE AND RESHAPED ACADEMIC INTEGRITY IN SOUTH AFRICAN DISTANCE LEARNING

This qualitative study investigates how South African distance education students use AI humaniser software to evade Turnitin's AI detection protocols within Canvas. Drawing on interviews, documentary evidence, and focus groups with 30 participants across three universities, the research reveals that while raw AI submissions trigger automatic rejection (Canvas error code 989), humanised submissions reliably receive passing grades when bibliographic references are authentic. The study documents markers' increasing reliance on algorithmic scoring systems and argues that similarity-based integrity frameworks are obsolete. Recommendations include mandatory referencing software certification, prohibition of undisclosed text optimisation, and development of detection methodologies targeting citation authenticity and stylistic consistency.