Built by academic experts in AI and educational learning.
ChatTutor comes out of the computer science department at the Technical University of Denmark, DTU Compute. Features and learning activities are co-created with experts in didactics, pedagogics, learning science and psychology — so the platform progressively rests on established research rather than on what is easy to build.
Agents are easily configured to question before they conclude — drawing on established work in Socratic questioning, spaced retrieval and metacognitive prompting.
Custom file parsing of complex elements — equations, derivations, figure captions — so answers in STEM courses are context-specific rather than paraphrased prose.
Deployment across institutions in Europe, Asia and the United States — higher and secondary education alike — with educational researchers running studies on the platform and publishing the results in peer-reviewed journals.
Evidence, not enthusiasm.
An agent built for active thinking rather than passive output produced measurable gains in conceptual knowledge, trust and enjoyment — and, four weeks on, in what students still remembered. Not every measure moved, and the paper says which ones did not.
What two controlled experiments actually found.
Published in Educational Psychology Review (Makransky, Shiwalia, Herlau & Blurton, 2025). Two between-subjects experiments compared ChatTutor against generic ChatGPT — same interface, students blind to condition — and against ordinary teaching.
Study 1, measured 1–5 against generic ChatGPT.
Replicated in Study 2: trust 3.56 vs 3.35 (d = 0.36) and enjoyment 3.49 vs 3.14 (d = 0.49), both significant.
Mean score out of 9.
Ahead of ChatGPT (p = .040) and re-studying (p = .003).
Ahead of teaching as usual (p = .001); the gap to ChatGPT was not significant.
Four ways to work with us.
The studies above are not ours alone to run. If you want to study how students learn alongside AI, there are four routes in — two you can start now, two that begin as a conversation.
Both give you the platform in full for research purposes, including the pedagogical configuration layer — and early access to unreleased features, by agreement.
Include ChatTutor as a partner or co-applicant; we can co-write the application. For large studies, or studies that need engineering support to be methodologically sound.
Features built for your research question. Usually funded by the institution, with development hours contributed in kind by ChatTutor. Describe the project and we will tell you what is feasible.
We are a research-focused company. We want to build features that demonstrably work, and to support a growing community of researchers on AI in learning.
- A stated research design before you begin — the question, the method, what you will measure.
- Permission to reference the collaboration publicly — a short testimonial or a summary of the research. Optional, and always your call.
- Findings shared back with us, null results included, so what we build next is evidence-led.
Peer-reviewed work on ChatTutor.
Press coverage of this work sits on the press page. Publishing on ChatTutor? Tell us and we will add it here.