Temporal Changes in Linguistic Revision Burden in English Abstracts of Korean Sports Science Journals: A Pre- and Post-2022 Analysis in the Era of AI-Assisted Writing

다국어 초록 (Multilingual Abstract)

The widespread availability of large language model (LLM)–based AI writing tools since 2022 has raised questions about whether measurable changes in academic writing quality are reflected in published texts. This study examines temporal shifts in linguistic revision burden, conceptualized as an indicator of the cognitive and discursive demands of academic writing, in English abstracts from Korean sports science journals. Forty abstracts from two journals—the Korea Journal of Applied Biomechanics (J1) and the Korean Journal of Sport Psychology (J2)—were analyzed, with 10 abstracts from 2019–2020 and 10 from 2024–2025 selected per journal. All abstracts underwent systematic linguistic auditing using ChatGPT (version 5.2), and correction density (revisions per 100 words) was calculated. Corrections were classified as grammatical accuracy, syntactic clarity, lexical choice, stylistic refinement, or discourse-level coherence. Results show a clear reduction in correction density in 2024–2025 compared to 2019–2020, consistent across journals. Category-level analyses indicate fewer grammatical and syntactic corrections and a proportional shift toward stylistic and discourse-level refinements in later abstracts. Segmented regression shows that revision burden was already declining before 2022 and continued thereafter, without a statistically decisive breakpoint. Overall, the findings document a substantial temporal decrease in linguistic revision burden in the post-2022 writing environment. Although the findings do not support a definitive causal claim regarding the role of LLM-based tools, the results provide system-level evidence consistent with evolving technological and institutional conditions in academic writing, including the normalization of AI-assisted support. By operationalizing linguistic revision burden through correction density at the level of published texts, this study offers a novel empirical framework for examining changes in academic writing across non-Anglophone research contexts.