Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Dental students trained with AI simulators scored higher, a review of 12 studies finds

Study title: Effectiveness of artificial intelligence-driven simulation in dental education: a systematic review and meta-analysis of learning outcomes

Evidence type
Systematic review or meta-analysis
Sample
n = 557
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2789cite or check it

Study title: Effectiveness of artificial intelligence-driven simulation in dental education: a systematic review and meta-analysis of learning outcomes

A systematic review and meta-analysis searched PubMed, Scopus and Web of Science up to March 2026 and included twelve studies with over 1,400 dental students. Five comparative studies, yielding six comparisons and 557 participants (292 AI, 265 control), were pooled. AI-driven simulation was associated with better overall learning outcomes than traditional training (SMD = 1.20; 95% CI 0.73 to 1.67).

Read the study abstract

BACKGROUND: This systematic review and meta-analysis aimed to evaluate and quantitatively synthesize the effectiveness of artificial intelligence (AI)-driven simulation (ADS) technologies in improving learning outcomes in dental education. METHODS: The PubMed, Scopus, and Web of Science electronic databases were searched up to March 2026. Studies evaluating ADS modalities, with or without a comparator, were included. Comparative studies with appropriate control groups contributed to the meta-analysis, while all…

The gain

A systematic review and meta-analysis of five comparative studies (557 dental students) found AI-driven simulation was associated with better overall learning outcomes than traditional training (SMD 1.20; 95% CI 0.73-1.67).

In brief

The effect was largest for intelligent tutoring systems (SMD 1.86) and VR-based haptic simulation (SMD 1.62), and smaller for generative AI (SMD 0.67). Mobile AI showed a positive but non-significant effect (SMD 0.88). These modality differences are exploratory and rest on few studies.

The pooled figure mixes hand-skill, knowledge and self-reported outcomes, so it is a composite effect rather than a measure of one skill. Heterogeneity was moderate to high (I² = 68.9%), and a trim-and-fill adjustment lowered the estimate to SMD 1.04. No study measured patient outcomes.

Main points

  1. Twelve studies with over 1,400 participants were included; five comparative studies with 557 students were pooled in the meta-analysis.
  2. The pooled effect favoured AI simulation over traditional training (SMD = 1.20; 95% CI 0.73 to 1.67; p = 0.0013).
  3. Intelligent tutoring systems showed the largest effect (SMD 1.86), followed by VR haptic simulation (SMD 1.62); generative AI showed SMD 0.67.
  4. Excluding studies with imputed standard deviations lowered the effect to SMD 0.94 (95% CI 0.42-1.46), still significant.
  5. Trim-and-fill analysis suggested two missing studies and reduced the estimate to SMD 1.04 (95% CI 0.57 to 1.51).
  6. Of the included randomized trials, two were rated high risk of bias and one low; four of the non-randomized studies were rated serious risk.
  7. No included study evaluated patient-related (Kirkpatrick Level 4) outcomes, and most measured short-term results.

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.

  • Overall learning outcomes after trim-and-fill adjustment for possible publication bias

    SMD 1.04 (95% CI 0.57 to 1.51)

    557 people · search to March 2026

    The review’s own words
    The effect size was reduced slightly after adjustment, to SMD = 1.04 (95% CI: 0.57 to 1.51; p = 0.0013), although this was still significant, suggesting potential publication bias; however, given the small number of studies, these findings should be interpreted with caution and do not allow reliable conclusions regarding the presence or impact of publication bias.

  • Learning outcomes with intelligent tutoring system simulation compared with traditional simulation (network meta-analysis)

    SMD 1.91 (95% CI 1.12 to 2.69)

    557 people · search to March 2026

    The review’s own words
    ITS-based systems recorded the highest effect compared to traditional simulation (SMD = 1.91; 95% CI: 1.12 to 2.69), followed by VR-haptic (SMD = 1.53) and VR adaptive (SMD = 1.49).

  • Overall learning outcomes of dental students with AI-driven simulation compared with traditional training

    SMD 1.20 (95% CI 0.73 to 1.67)

    557 people · search to March 2026

    The review’s own words
    ADS, considered as a heterogeneous group of AI-based educational interventions, was associated with improved overall learning outcomes compared to traditional methods (SMD = 1.20; 95% CI: 0.73-1.67).

  • Overall learning outcomes excluding studies with imputed standard deviations (sensitivity analysis)

    SMD 0.94 (95% CI 0.42 to 1.46)

    search to March 2026

    The review’s own words
    When performing sensitivity analysis excluding studies with imputed standard deviations [ 28 , 34 ], the pooled effect was slightly reduced but remained statistically significant (SMD = 0.94, 95% CI: 0.42–1.46, p = 0.0102), indicating no material impact on overall findings.

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

The review was prospectively registered in PROSPERO and followed PRISMA 2020. Two reviewers screened 121 records after de-duplication, assessed 21 full texts and included 12 studies from China, Egypt, Pakistan, Thailand, Australia, Turkey, Germany, Spain and Brazil. Four were randomized trials, and sample sizes ranged from 30 to 300.

Interventions were mostly VR systems with machine-learning haptic feedback for procedural training, and chatbot or large language model tools for clinical reasoning. Comparators included phantom-head training, lectures, videos and workshops. Skills covered endodontics, operative skills, cavity preparation, history taking and diagnosis.

In the narrative synthesis, psychomotor gains were most consistent with VR-haptic and tutoring systems measured by objective metrics. Generative AI was more often tied to perceived gains in confidence and satisfaction than to validated improvements in knowledge. Higher-quality studies tended to report objective improvements, while lower-quality studies relied on subjective outcomes.

Meta-regression found that AI modality explained all the between-study heterogeneity, although the overall moderator test was not significant (p = 0.0773). Compared with generative AI, tutoring systems (β = 1.20) and VR simulation (β = 0.82) showed larger effects. The authors call these findings exploratory.

Sources

1 on file
  1. BMC Medical Educationdoi:10.1186/s12909-026-09390-6
    Peer-reviewed2026-05-13

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

Dental students trained with AI simulators scored higher, a review of 12 studies findshttps://truvace.com/item/effectiveness-of-artificial-intelligence-driven-simulation-in-dental-education-a

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

    The gain A systematic review and meta-analysis of five comparative studies (557 dental students) found AI-driven simulation was associated with better overall learning outcomes than traditional training (SMD 1.20; 95% CI 0.73-1.67).

    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 557, 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 The pooled estimate combines randomized and non-randomized studies, mixes psychomotor, cognitive and subjective outcomes, shows moderate-to-high heterogeneity, and rests on few, mostly short-term studies with moderate-to-serious risk of bias.

    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