The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2817
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:28:38.081785Z
status: published
lens: g_space
sector: education
headline: A Second-Order Meta-Analysis on the Effects of Artificial Intelligence Applications on Student Outcomes
dek: Research on AI in education has expanded rapidly, resulting in numerous first-order meta-analyses, yet these studies are limited by inconsistent findings, heterogeneous methods, variable quality, and overlapping primary studies, requiring a second-order meta-analysis to obtain a more reliable estimate of AI’s effects. This study examined the overall effect of AI applications on student outcomes, including academic achievement and higher-order thinking skills. We included first-order meta-analyses published betwe…
gain_title: A second-order meta-analysis of 19 earlier meta-analyses covering 58,702 students found AI applications raised student outcomes by a moderate average effect size of 0.67 (95% CI 0.55 to 0.78).
problem_title: (none)
trace_subject: (none)
gain_reading: A second-order meta-analysis of 19 earlier meta-analyses covering 58,702 students found AI applications raised student outcomes by a moderate average effect size of 0.67 (95% CI 0.55 to 0.78).
gain_evidence: statistically significant moderate mean effect size (ES = .67, 95% CI [.55–.78]) | Nineteen meta-analyses obtained from electronic databases, involving a total of 58,702 participants
problem_reading: (none)
problem_evidence: (none)
quick_read: A second-order meta-analysis pooled results from 19 first-order meta-analyses published between 2020 and 2025, together involving 58,702 participants. Using a random-effects model, it found a moderate mean effect size of 0.67 (95% CI 0.55 to 0.78) for AI applications on student outcomes, including academic achievement and higher-order thinking skills.

The authors undertook the work because first-order meta-analyses of AI in education give inconsistent findings, use varied methods, differ in quality and draw on overlapping primary studies. Pooling them was meant to give a more reliable overall estimate.

Effects varied by education level, education field and publication bias status, but were robust across AI types and learning outcome types. The authors call for informed AI integration and stronger pedagogical and institutional capacity.
limitation: Effects varied significantly by education level, education field and publication bias status, so the average figure does not apply uniformly across settings.
tag: Evidence-backed gain
key_points: The analysis covered 19 first-order meta-analyses published between 2020 and 2025 with 58,702 participants in total. | Those meta-analyses drew on primary studies published from 1993 to 2024. | The random-effects mean effect size was 0.67 (95% CI 0.55 to 0.78), described as moderate and statistically significant. | Education level, education field and publication bias status helped explain variation in effects. | Effect sizes held across AI types and learning outcome types. | In meta-regression, sample size and publication year did not predict effect size, but the number of primary studies did.
rundown: The study is a second-order meta-analysis, which pools the results of earlier meta-analyses rather than individual experiments. The team searched electronic databases for first-order meta-analyses published from 2020 to 2025, which covered primary studies from 1993 to 2024.

Outcomes included academic achievement and higher-order thinking skills. The random-effects model gave a mean effect of 0.67 (95% CI 0.55 to 0.78) across the 19 included meta-analyses.

Moderator analysis showed that effects differed by education level, education field and publication bias status. Meta-regression found that sample size and publication year did not predict effect sizes, whereas the number of primary studies did.
sources:
- peer_reviewed | Journal of Educational Computing Research | https://doi.org/10.1177/07356331261424767 | 2026-02-13
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

1b86bc9fbcaf385907d6e5b6b454d03fd8ae4b8d3b937f9c333dcabaa1c60d74

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000