aceModerate evidence
More
Meta-analysis finds AI feedback worked about as well as teacher feedback for 4,813 students
Study title: How does artificial intelligence compare to human feedback? A meta-analysis of performance, feedback perception, and learning dispositions
- Evidence type
- Systematic review or meta-analysis
- Sample
- n = 4,813
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2774cite or check it
Study title: How does artificial intelligence compare to human feedback? A meta-analysis of performance, feedback perception, and learning dispositions
An exploratory meta-analysis pooled 41 studies covering 4,813 students and compared learning performance after AI-generated feedback with performance after human-provided feedback. It found no statistically significant difference, with a small pooled effect (Hedge's g = 0.25, CI −0.11 to 0.60).
Read the study abstract
This exploratory meta-analysis synthesises current research on the effectiveness of Artificial Intelligence (AI)-generated feedback compared to traditional human-provided feedback. Drawing on 41 studies involving a total of 4813 students, the findings reveal no statistically significant differences in learning performance between students who received AI-generated feedback and those who received human-provided feedback. The pooled effect size was small and statistically insignificant (Hedge’s g = 0.25, CI [−0.11…
The gain
A meta-analysis of 41 studies with 4,813 students found no statistically significant difference in learning performance between AI-generated and human feedback (Hedge's g = 0.25, CI −0.11 to 0.60), suggesting AI feedback may be as effective.
In brief
A separate analysis limited to language and writing studies reached similar findings, though results varied widely between studies (I2 = 95%). Students' perceptions of the feedback showed a small, negative and statistically insignificant effect (g = −0.20, CI −0.67 to 0.27).
The authors read the results as suggesting AI feedback is potentially as effective as human feedback, and recommend a hybrid approach that combines AI's scalability with the empathetic, contextual qualities of human feedback.
Main points
- The meta-analysis pooled 41 studies with a combined 4,813 students.
- Pooled learning performance effect was small and not significant (Hedge's g = 0.25, CI −0.11 to 0.60).
- A language and writing subset gave similar results.
- Heterogeneity stayed high (I2 = 95%) across studies.
- Perception of feedback leaned slightly negative toward AI but was not significant (g = −0.20, CI −0.67 to 0.27).
- The authors advocate a hybrid model of AI and human feedback.
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.
Student perception of feedback from AI compared with human-provided feedback
Hedges' g −0.20 (95% CI −0.67 to 0.27)
The review’s own words
found a small, negative, and statistically insignificant effect size (Hedge’s g = −0.20, CI [−0.67; 0.27])
Student learning performance with AI-generated feedback compared with human-provided feedback
Hedges' g 0.25 (95% CI −0.11 to 0.60)
41 studies · 4,813 people
The review’s own words
The pooled effect size was small and statistically insignificant (Hedge’s g = 0.25, CI [−0.11; 0.60])
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 analysis, published in Educational Psychology in 2025, is described as exploratory. It compared AI-generated feedback with traditional human-provided feedback across 41 studies and 4,813 students, using learning performance as the main outcome.
The pooled effect was small and statistically insignificant, and its confidence interval spans zero, so the data are compatible with AI feedback being somewhat better or somewhat worse. A language and writing sub-analysis confirmed similar findings.
The authors also looked at how students perceived the feedback and found a small negative but insignificant difference. They argue for hybrid use that keeps human empathy and context alongside AI's scalability.
Sources
1 on file- Educational Psychologydoi:10.1080/01443410.2025.2553639
Talk it over
Who gets this benefit, and who was left out of the study or story?
Meta-analysis finds AI feedback worked about as well as teacher feedback for 4,813 studentshttps://truvace.com/item/how-does-artificial-intelligence-compare-to-human-feedback-a-meta-analysis-of-pe
Who gets this benefit, and who was left out of the study or story?
The gain A meta-analysis of 41 studies with 4,813 students found no statistically significant difference in learning performance between AI-generated and human feedback (Hedge's g = 0.25, CI −0.11 to 0.60), suggesting AI feedback may be as effective.
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 4,813, from Educational Psychology.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat The analysis is exploratory, the confidence interval for the pooled effect spans zero, and heterogeneity across studies was very high (I2 = 95%), so the result is imprecise.
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