aceModerate evidence
More
Health students using AI chatbots learned more on their own, review of 14 studies finds
Study title: The Impact of Artificial Intelligence-Powered Chatbots on Self-Directed Learning in Health Sciences Students: A Systematic Review and Meta-Analysis
- Evidence type
- Systematic review or meta-analysis
- Sample
- n = 1,182
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2748cite or check it
Study title: The Impact of Artificial Intelligence-Powered Chatbots on Self-Directed Learning in Health Sciences Students: A Systematic Review and Meta-Analysis
A systematic review and meta-analysis pooled 14 studies (11 randomized trials, 3 quasi-experimental; 1,182 students) testing AI chatbots in health professions education. In the five studies (365 participants) that measured self-directed learning, chatbots were linked to a large improvement (Hedges's g=0.85).
Read the study abstract
Objective To evaluate the effectiveness of AI-powered chatbots on self-directed learning (SDL; primary outcome), satisfaction, confidence, and critical thinking in health professions students. Methods We systematically searched five databases Library up to July 2026. Two reviewers independently performed study selection, data extraction, and risk-of-bias assessment (RoB 2 for RCTs; ROBINS-I for quasi-experimental studies). Random-effects meta-analyses using REML estimation were performed, with Hartung-Knapp-Sidi…
The gain
A meta-analysis of five studies with 365 health professions students found AI chatbots may improve self-directed learning (Hedges's g=0.85, 95% CI 0.50-1.21), though certainty was low.
In brief
Satisfaction also rose in five studies with 320 participants (g=1.05), but results varied widely between studies (I2=75.9%) and the prediction interval ran from -0.67 to 2.77. The evidence for both outcomes was rated low certainty.
The confidence result did not hold up: using the stricter Hartung-Knapp-Sidik-Jonkman method, the interval for three studies (227 participants) crossed zero. Critical thinking and self-efficacy could not be pooled because the studies differed too much.
Main points
- Fourteen studies were included: 11 randomized trials and 3 quasi-experimental studies, with 1,182 health professions students in total.
- Self-directed learning: g=0.85 across 5 studies and 365 participants, with a prediction interval of 0.21-1.50.
- Leave-one-out sensitivity analyses kept the self-directed learning effect between 0.75 and 0.93, all with p<0.05.
- Satisfaction: g=1.05 across 5 studies and 320 participants, with high heterogeneity (I2=75.9%).
- Confidence: the HKSJ interval of -0.37 to 1.98 crossed the null, while the conventional Wald-type interval (0.27 to 1.35) suggested significance.
- Critical thinking and self-efficacy were not pooled because of extreme heterogeneity.
What the review found when it combined the studies
These are the review’s own pooled results, copied from its abstract. Truvace did not recalculate them.
Satisfaction of health professions students using AI chatbots
Hedges' g 1.05 (95% CI 0.36 to 1.75)
5 studies · 320 people · search to July 2026
The review’s own words
Chatbots may improve satisfaction (g=1.05; 95% CI: 0.36-1.75; p=0.014; 5 studies; I 2 =75.9%; prediction interval: -0.67 to 2.77; low certainty).
Code checks every number above against the review’s text before it is shown. The short labels are an AI model’s reading of that text, so the review’s own sentence sits under each one. A pooled figure is only as strong as the studies behind it. How Truvace uses pooled results
The full story
Researchers searched five databases up to July 2026. Two reviewers independently selected studies, extracted data and assessed risk of bias, using RoB 2 for randomized trials and ROBINS-I for quasi-experimental studies.
Random-effects meta-analyses used REML estimation with Hartung-Knapp-Sidik-Jonkman confidence intervals and prediction intervals where feasible. Certainty of evidence was graded with GRADE and rated low for self-directed learning and satisfaction.
The authors attribute the low certainty to indirectness, because studies measured self-directed learning with different tools, and to imprecision, because the sample was below the optimal information size. They call for longer-term evaluations and validated multi-item scales such as the Self-Directed Learning Scale and SRSSDL.
Sources
1 on file- Journal of Medical Education and Curricular Developmentdoi:10.1177/23821205261491865
Talk it over
Who gets this benefit, and who was left out of the study or story?
Health students using AI chatbots learned more on their own, review of 14 studies findshttps://truvace.com/item/the-impact-of-artificial-intelligence-powered-chatbots-on-self-directed-learning
Who gets this benefit, and who was left out of the study or story?
The gain A meta-analysis of five studies with 365 health professions students found AI chatbots may improve self-directed learning (Hedges's g=0.85, 95% CI 0.50-1.21), though certainty was low.
Answer in the threadHow do we know? What can this kind of evidence show, and what can’t it?
The evidence This rests on a systematic review or meta-analysis with a sample of 1,182, from Journal of Medical Education and Curricular Development.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat Certainty of evidence was low because studies used different self-directed learning measures and the sample was smaller than optimal; the confidence finding was fragile and critical thinking could not be pooled.
Answer in the threadHave you seen AI used this way where you study, work or live? Did it go the same way?
Answer in the thread
The questions are the same on every record. The details under them are taken from this record as written.
What readers say