Nick Huntington-Klein (2026) — Single-author study.
Research Design
This paper uses a clever identification strategy: measure the seasonal drop in ChatGPT activity during non-school summer months (2023 and 2024). Areas with larger summer dropoffs have heavier school-related AI use. The author then examines whether higher AI-use areas show different test score trends.
Key Finding: Null Result
No meaningful impact of ChatGPT availability on high school test score averages in either direction. Whether students use AI to cheat, to learn, or both, the aggregate effect on standardized test performance is negligible — at least in the short term and as AI is actually used.
Interpretation
The null result has two non-exclusive interpretations: (1) AI-assisted cheating does not harm test performance in aggregate, or (2) negative effects of AI misuse are offset by positive uses of AI as a learning aid. Either way, this challenges both utopian and dystopian narratives about AI in classrooms. It aligns with the genai-performance-vs-learning distinction — performance boosts from AI may not translate to test score improvements. It connects to over-reliance research showing that AI assistance can produce apparent competence without underlying understanding.
Methodological Contribution
The seasonal variation identification strategy is innovative for causal inference in education technology research, where RCTs are often infeasible at scale.
Related Pages
- over-reliance — AI as shortcut vs. learning tool
- academic-integrity — Cheating and AI use in schools
- k-12-ai-education — High school context
- student-experience — How students actually use AI
- ai-tutor-effectiveness-review — Contrasts with controlled AI tutoring studies
- genai-performance-vs-learning — Performance gains without learning
Citation
APA: Huntington-Klein, N. (2026). Little impact of ChatGPT availability on high school student test score performance. arXiv:2605.08812.