STEM Undergraduates’ Perceptions of AI Chatbots: A Cross-Sectional Descriptive Survey

We surveyed 297 STEM undergraduates at a single English-medium Sino–UK joint institution to document perceptions of AI chatbots for learning. Students reported high willingness to adopt AI chatbots (78%; 95% CI: 73.1–82.4) alongside concerns about over-reliance (67%; 95% CI: 61.4–72.1), content quality (52%; 95% CI: 46.2–57.5), and reduced human interaction (42%; 95% CI: 36.5–47.8). Over half (52%; 95% CI: 46.3–57.7) requested language/terminology support features, whereas only 16.8% reported language-related barriers. We attempted exploratory factor analysis and k-means clustering, but neither met the inclusion criteria; therefore, we report item-level frequencies only. The findings are descriptive and not generalisable (53% first-year, 80% male convenience sample). These patterns generate testable hypotheses about verification scaffolds, language support utility, and human–AI balance that warrant investigation through controlled studies.