Using AI-based Learning Assistants in Higher Education: A Large-Scale Descriptive Analysis

Synthesis: Presents a large-scale descriptive analysis of an AI learning assistant (Syntea) using objective log data from 77,543 higher-education students, characterizing real usage patterns, adoption, and engagement at scale. The work connects to broader debates about how Generative AI systems reshape Student Experience and the conditions under which AI support scaffolds rather than undermines learning. It has direct implications for The Pedagogy of AI Mistakes: Fostering Higher-Order Thinking and the risk of Over-Reliance when assistants absorb too much of the cognitive load. Findings also bear on AI Literacy and Self-Regulated Learning, and on how institutions should govern Student Experience and Academic Integrity. Practitioners in Higher Education and teachers can use the evidence to calibrate when to deploy Large Language Models (LLMs)-based help and how to pair it with Feedback that preserves learning gains.

What this means for practice

• Learners. Check whether the assistant is even offered in your course before reading your usage as a preference: adoption tracked course coverage, and usage was lowest in Architecture & Construction and Design & Media and highest in Education & Psychology and Marketing & Communication.
• Learners. Put AI-assisted study into the windows in which you already work. Activity peaked in the late morning around 11:00 and on Monday through Wednesday, while part-time students shifted their use to weekends and later hours — aligning tool use with real study rhythms fits it into the routine rather than adding a separate task.
• Instructors. Extend assistant coverage into seminars, project-based courses, and thesis contexts, where use was structurally low because the tool was unavailable rather than unwanted; Master's students' lower rate (54.25% vs. 58.17% for Bachelor's) largely reflects their longer thesis phase.
• Administrators. Read group differences as access conditions first. Female students used the assistant more than male students (59.05% vs. 54.94%) and Gen Z most of all (63.66%), patterns the authors tie to unequal course integration rather than individual disposition.
• Designers. Add multimodal capabilities for disciplines whose learning is visual or practice-based, since text-centered support fits those programs least well.

Limitations

• The study is descriptive, not causal, and covers a single observation month — a choice that improves comparability but rules out any reading of longer-term trends or of why adoption changed.
• Usage is measured as adoption and timing only; the data carry no information about interaction quality or learning outcomes, so nothing here speaks to whether using Syntea helped anyone learn.
• All 77,543 students come from one distance university and its own Syntea deployment, so patterns reflect that institution's course-coverage rules and program structure.
• Several subgroup estimates are small enough that the authors caution against them: the Boomer cohort is 132 students (37.88% usage) and the group recorded as diverse is 109 students (34.86% usage), both subject to substantial sampling variability.

Citation

Schaaff, K., Stierstorfer, Q., & Hekkel, V. (2026). Using AI-based Learning Assistants in Higher Education: A Large-Scale Descriptive Analysis.