Lower grades (RCT)
A GPT-4o tutor built into a university learning platform was associated with lower grades in a 2,379-student randomized trial.
A University of Maryland working paper (Oct 2026) randomized 30 instructors in fall 2025 to a course-integrated GPT-4o tutor drawing on their own materials; students with the tutor performed worse.
Sources
- primaryEdWorkingPaper 26-1598: The Effects of Course-Integrated AI Tutoring on Student Performance and Engagement — A Randomized University Trialedworkingpapers.com
- secondaryWorld Assessment Council summary of EdWorkingPaper 26-1598worldassessmentcouncil.org
Editor's notes
Primary located: EdWorkingPaper landing page and abstract (Liu, Sweet, et al., Oct 2026); losses larger for first-generation students. Secondary coverage reports only ~15% of students offered the tutor used it, which would make this an intent-to-treat effect of access, not of use. verify: randomization unit (instructor vs section), fall 2025 term dates, and uptake rate from the full PDF; the grade effect is reported for within-course comparisons.
Related
History
- The editor has not yet read the working paper; published unreviewed.
- Weekly run: EdWorkingPaper primary source added (entry now renders); effect size recorded from the abstract.
- Added — needs primary source before it renders.
Cite this entry
Permalink https://becauseofai.org/entry/maryland-ai-tutor-negative
“A GPT-4o tutor built into a university learning platform was associated with lower grades in a 2,379-student randomized trial.” Because of AI, updated Oct 22, 2026. https://becauseofai.org/entry/maryland-ai-tutor-negative@misc{becauseofai_maryland_ai_tutor_negative,
title = {A GPT-4o tutor built into a university learning platform was associated with lower grades in a 2,379-student randomized trial.},
howpublished = {\url{https://becauseofai.org/entry/maryland-ai-tutor-negative}},
note = {Because of AI registry, updated Oct 22, 2026},
year = {2026}
}