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-2026-2749
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:18:39.351098Z
status: published
lens: g_space
sector: education
headline: Do AI Writing Tools Improve L2 Writing Performance? A Meta-Analysis of Experimental Evidence, Moderators, and Methodological Quality
dek: Artificial intelligence (AI) writing tools—including large language model (LLM)-based systems, automated writing evaluation (AWE) platforms, and AI-powered writing assistants—have become an increasingly prominent feature of second language (L2) writing pedagogy. Yet whether these tools reliably improve L2 writing performance, and under what instructional conditions, remains an open question. This study reports a systematic meta-analysis of experimental and quasi-experimental evidence on the effects of AI writing…
gain_title: A meta-analysis of experimental studies published 2021 to 2026 found AI writing tools improved second-language writing performance, with a pooled effect of g = 0.714 (95% CI 0.592 to 0.836) across 22 verified effect sizes.
problem_title: (none)
trace_subject: (none)
gain_reading: A meta-analysis of experimental studies published 2021 to 2026 found AI writing tools improved second-language writing performance, with a pooled effect of g = 0.714 (95% CI 0.592 to 0.836) across 22 verified effect sizes.
gain_evidence: The verified model produced a pooled effect of g = 0.714, 95% CI [0.592, 0.836], with low observed heterogeneity (I² = 24.3%).
problem_reading: (none)
problem_evidence: (none)
quick_read: A meta-analysis of experimental and quasi-experimental studies found that AI writing tools improved second-language (L2) writing performance. The main model, built on 22 effects with directly extractable statistics, produced a pooled effect of g = 0.714 (95% CI 0.592 to 0.836), with low heterogeneity (I² = 24.3%).

Searches of Scopus, Web of Science and ERIC, plus citation tracing and hand-searching, identified 34 eligible effect sizes from studies published between 2021 and 2026. A second model using all 34 effects, including converted ones, served as a robustness check.

Exploratory moderator analyses linked larger effects to multidimensional feedback, interventions lasting eight weeks or more, and university settings. Single-session designs and feedback limited to lower-order concerns showed smaller gains. Many studies lacked delayed transfer measures and transparent effect-size reporting.
limitation: Moderator analyses were exploratory, and many included studies lacked delayed transfer measures and transparent effect-size reporting, so lasting effects are not established.
tag: Evidence-backed gain
key_points: Pooled effect across 22 verified effects was g = 0.714 (95% CI 0.592 to 0.836). | Observed heterogeneity was low, with I² = 24.3%. | 34 eligible effect sizes came from studies published between 2021 and 2026. | The databases searched were Scopus, Web of Science and ERIC, with forward citation tracing and hand-searching added. | Interventions lasting eight weeks or more, multidimensional feedback and university contexts were associated with larger effects. | Single-session interventions and feedback on lower-order concerns only showed smaller gains. | Methodological quality was generally adequate but uneven; many studies lacked delayed transfer measures.
rundown: The analysis covered AI writing tools broadly, including systems built on large language models, automated writing evaluation platforms and AI-powered writing assistants, in second-language writing instruction. Effect sizes were expressed as Hedges' g.

The main estimate used restricted maximum-likelihood (REML) with Knapp-Hartung adjustment on 22 effects with directly extractable statistics. A 34-effect model that included conversion-based effects was run as a robustness analysis.

The authors describe the tools as pedagogically contingent rather than inherently effective. They say educational value depends on how feedback literacy, teacher mediation and instructional design shape opportunities for noticing, revision and transferable writing development.
sources:
- peer_reviewed | Journal of Educational Technology and Innovation | https://doi.org/10.61414/awr9xk46 | 2026-09-29
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

00fb0a4e39db432f27c36bcd5d486acbb4f2bba6062cf604ccded84ed90c51df

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000