Abstract
<jats:p>How do students actually experience high-stakes testing—and does the empirical record match the often polarised policy narratives? This registered report protocol describes a systematic review of empirical studies published between 2015 and 2025 that collected primary data from students about the motivational, learning, and well-being consequences of high-stakes or standardised assessments. The review will draw on three complementary theoretical lenses: washback theory, which frames how testing systems shape classroom behaviour; Self-Determination Theory (SDT), which foregrounds students’ needs for autonomy, competence, and relatedness; and Expectancy-Value Theory (EVT), which captures how beliefs about success and task importance drive effort and engagement. Four databases will be searched (Scopus, ERIC, PsycINFO, and Web of Science) using an expanded query that includes both educational assessment and psychological construct terms, and all eligible studies will be included without subsampling. The review will be conducted by a single author using an AI-augmented workflow: LLM-assisted coding will provide the first-pass extraction, the author will verify and refine, and an independent trained coder will validate a stratified subset to establish inter-rater reliability. The AI-assisted component follows a three-phase pilot–validate–refine protocol in which BERTopic-based topic modelling and LLM-assisted coding are calibrated on a pilot sample, then refined and validated by an independent coder on 15% of the corpus, with phased extension if reliability is borderline. Motivational constructs will be coded using a four-level hierarchical scheme that distinguishes construct presence, direction of effect, mixed-effect subtypes, and evidence strength. RQ2 will adopt an exploratory framing — characterising how SDT and EVT constructs manifest across contexts—rather than testing a directional hypothesis. Expected outcomes include a descriptive map of the student-experience literature, an analysis of motivational patterns across regions and education levels, and a transparent evaluation of what a single researcher can achieve with AI-assisted evidence synthesis. This manuscript constitutes the Stage 1 preregistration.</jats:p>