A hybrid bibliometric and systematic review of generative AI in engineering education since the advent of ChatGPT

This review examines how engineering educators have empirically studied generative artificial intelligence (GenAI) and large language models (LLMs) in their courses since the November 2022 release of ChatGPT. From a 402-record Scopus corpus on AI in engineering education, a GenAI/LLM filter applied to titles, author keywords, and abstracts identified 184 candidates (142 paywalled, 42 open-access); 50 full-text records (42 open-access plus 8 institutionally acquired) constitute the synthesis base. Each record was coded on engineering sub-discipline, instructional application, and study design, paired with a bibliometric overview of citation anchors and venue distribution. Four findings dominate. First, the literature is overwhelmingly perceptual rather than experimental: only 6 of 50 records (12%) report controlled or quasi-experimental interventions, while 18 (36%) are surveys and 16 (32%) are perception or opinion pieces. Second, evidence is concentrated in software and computer engineering (n = 11, 22%), computing-adjacent settings (n = 9, 18%), and general engineering (n = 15, 30%); chemical (n = 4), mechanical (n = 4), civil (n = 3), and other sub-disciplines are thinly populated, and a sensitivity analysis excluding computing records leaves the principal findings unaffected. Third, application activity centres on assessment and grading (66%), perceptions and acceptance (62%), and ethics and integrity (28%), while safety and accreditation, both central to engineering as a profession, each appear in a single record. Fourth, the specific model version studied is unreported in the majority of records, and reporting practice has not improved across the review window. The six controlled studies are synthesised in detail and a framework for assessment redesign is proposed. It is argued that the most urgent contribution this review can make is to render visible the field’s methodological composition, and implications are drawn for educators, program leaders, accreditation bodies, and researchers. Not applicable.