Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Chinese medical students taught with AI scored higher in exams, a review of 12 studies finds

Study title: Effectiveness of AI-assisted medical education for Chinese undergraduate medical students: a meta-analysis

Evidence type
Systematic review or meta-analysis
Sample
n = 824
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2833cite or check it

Study title: Effectiveness of AI-assisted medical education for Chinese undergraduate medical students: a meta-analysis

A meta-analysis of 12 studies covering 824 undergraduate medical students in China found that AI-assisted teaching beat traditional teaching on exam scores (standardized mean difference 2.06, 95% CI 1.35–2.76). In eight studies with 599 students, satisfaction was also higher with AI (odds ratio 5.80, 95% CI 3.30–10.18).

Read the study abstract

BACKGROUND: In recent years, artificial intelligence (AI) has been adopted as an innovative teaching method. However, no studies have comprehensively evaluated the effectiveness of AI in the context of medical education in China. This study aimed to assess the effectiveness of AI-assisted teaching methods versus traditional teaching methods regarding their impact on medical education outcomes in China. METHODS: We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines…

The gain

A meta-analysis of 12 studies with 824 Chinese undergraduate medical students found AI-assisted teaching raised exam scores over traditional teaching (SMD 2.06) and made students about 5.8 times more likely to report satisfaction.

In brief

The effect was larger in longer programmes: studies running more than 12 weeks showed an SMD of 3.27, against 1.68 for those of 12 weeks or less. AI adaptive learning systems (SMD 3.35) outperformed AI-assisted diagnosis systems (SMD 1.68) in subgroup analyses.

The studies were small, ranging from 18 to 120 students, and the pooled exam-score result showed very high heterogeneity (I² = 94%). Small studies can overstate effects, so the large pooled effect needs confirmation in bigger trials.

Main points

  1. Twelve studies published between 2019 and 2024 were included, drawn from searches of seven English and Chinese databases covering January 2015 to December 2024.
  2. Of the 824 students, 414 were in AI-assisted groups and 410 in control groups; majors were clinical medicine and medical imaging.
  3. Nine studies were randomized controlled trials; the design of the other three was not clearly specified.
  4. Satisfaction data came from questionnaires in eight studies involving 599 students (301 AI, 298 control), with low heterogeneity (I² = 11%).
  5. Long-term studies (more than 12 weeks, five studies) showed an SMD of 3.27 versus 1.68 in four short-term studies.
  6. Egger's test found no significant publication bias (P = 0.79), and a leave-one-out analysis kept all estimates within the original confidence interval.
  7. Tools covered included ChatGPT, AI imaging diagnosis systems, AI-assisted diagnosis systems and adaptive learning platforms.

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.

  • Examination scores in short-term AI-assisted teaching studies (12 weeks or less)

    SMD 1.68 (95% CI 1.40 to 1.97)

    search to December 2024

    The review’s own words
    Additionally, the pooled analysis revealed that AI-assisted teaching significantly increased examination scores in the fixed-effects model in the long-term group (SMD = 3.27, 95% CI [2.84, 3.71], P < 0.00001) and short-term group (SMD = 1.68, 95% CI [1.40, 1.97], P < 0.00001).

  • Examination scores in long-term AI-assisted teaching studies (over 12 weeks)

    SMD 3.27 (95% CI 2.84 to 3.71)

    5 studies · search to December 2024

    The review’s own words
    Additionally, the pooled analysis revealed that AI-assisted teaching significantly increased examination scores in the fixed-effects model in the long-term group (SMD = 3.27, 95% CI [2.84, 3.71], P < 0.00001) and short-term group (SMD = 1.68, 95% CI [1.40, 1.97], P < 0.00001).

  • Examination scores with AI adaptive learning systems versus traditional teaching

    SMD 3.35 (95% CI 2.85 to 3.85)

    search to December 2024

    The review’s own words
    Additionally, a pooled analysis revealed that AI-assisted teaching significantly increased examination scores in the fixed-effects model in the AI-assisted diagnosis system group (SMD = 1.68, 95% CI [1.40, 1.97], P < 0.00001) and the AI adaptive learning system group (SMD = 3.35, 95% CI [2.85, 3.85], P < 0.00001).

  • Satisfaction of Chinese undergraduate medical students with AI-assisted versus traditional teaching

    Odds ratio 5.80 (95% CI 3.30 to 10.18)

    599 people · search to December 2024

    The review’s own words
    All studies indicated that AI-assisted teaching significantly increased student satisfaction (OR = 5.80, 95% CI [3.30, 10.18], P < 0.00001) compared with traditional teaching (Fig. 4 ).

  • Examination scores of Chinese undergraduate medical students with AI-assisted versus traditional teaching

    SMD 2.06 (95% CI 1.35 to 2.76)

    824 people · search to December 2024

    The review’s own words
    The pooled data from the forest plots revealed that compared with traditional teaching, AI-assisted teaching significantly increased students’ examination scores (SMD = 2.06, 95% CI [1.35, 2.76], P < 0.00001) (Fig. 3 ).

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 followed PRISMA guidelines and searched PubMed, Web of Science, Embase, the Cochrane Library and three Chinese databases. Of 588 initial articles, 104 were reviewed in more detail and 12 met the criteria: Chinese undergraduate medical students, randomized or non-randomized comparisons of AI and traditional teaching, and usable quantitative data.

All 12 studies reported significant gains in exam scores for the AI groups, but results varied widely (I² = 94%). Meta-regression pointed to randomization, study duration and type of AI technology as sources of the variation, while institution type, major, curriculum and sample size did not explain it. Heterogeneity fell to 0% to 28% within subgroups split by duration and by AI type.

The authors note that most studies did not report allocation concealment or blinding of participants and assessors. Satisfaction was measured with non-standardized, institution-specific questionnaires whose reliability was rarely reported. Instructor AI proficiency was seldom described, and the authors call for larger randomized trials and standardized outcome measures.

Sources

1 on file
  1. BMC Medical Educationdoi:10.1186/s12909-025-07770-y
    Peer-reviewed2025-08-27

Talk it over

Four questions to start onWho gets this benefit, and who was left out of the study or story?

Start with a question

Chinese medical students taught with AI scored higher in exams, a review of 12 studies findshttps://truvace.com/item/effectiveness-of-ai-assisted-medical-education-for-chinese-undergraduate-medical

  1. Who gets this benefit, and who was left out of the study or story?

    The gain A meta-analysis of 12 studies with 824 Chinese undergraduate medical students found AI-assisted teaching raised exam scores over traditional teaching (SMD 2.06) and made students about 5.8 times more likely to report satisfaction.

    Answer in the thread

  2. How 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 824, from BMC Medical Education.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat Only 12 studies with 824 students were pooled (individual studies had 18 to 120 students), and small studies can overstate effects; most did not report blinding or allocation concealment, and satisfaction relied on non-standardized questionnaires.

    Answer in the thread

  4. Have 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

More from Education