Assessment in the Age of Generative AI: Academic Integrity, Authentic Learning and Assessment Design in Tanzanian Universities — A Systematic Review

Generative artificial intelligence (GenAI) tools such as ChatGPT can now produce essays, problem solutions and examination answers that pass as competent student work, raising the question of whether university grades still show what students know and can do. This systematic review with thematic synthesis examined what the international and African literature reports on GenAI and academic integrity, the effectiveness and fairness of AI detection, and the assessment designs that hold up in an AI-rich environment. Following PRISMA 2020, seven databases and organisational sources were searched for studies published between January 2018 and June 2026. Thirty studies were included, eleven from Africa and four from Tanzania; they were appraised with the Mixed Methods Appraisal Tool, AMSTAR 2 and the AACODS checklist and synthesised thematically with frequency counts. Five themes emerged: widespread and rising student use of GenAI (50% of studies); the vulnerability of conventional take-home and online assessment (60%); the unreliability and unfairness of AI-detection software (27%); the partial protection offered by authentic, process-focused and oral assessment (37%); and policy, capacity and training gaps (60%), most acute in African institutions. The evidence indicates that Tanzanian universities cannot rely on detection. They should move from a policing model to a design model of integrity, supported by institutional guidance, staff development and a graded scale of acceptable AI use. A Tanzanian Assessment Redesign Framework (TARF) is proposed for empirical testing, and priorities for intervention research are identified.