Artificial intelligence in oncology publishing: a systematic review and policy analysis of high-impact journals

Abstract

Background Generative artificial intelligence (AI) is reshaping scholarly communication, yet guidance for its responsible use remains uneven across biomedical journals. We aimed to systematically assess editorial policies governing AI-assisted writing in high-impact oncology journals. Methods We conducted a systematic review of publicly available editorial and normative documents, operationalized as a cross-sectional policy audit. Oncology journals with a 2023 Journal Impact Factor ≥5 (JCR 2024) were included. Author instructions, editorial policies, and publisher statements issued between January 2020 and March 2025 were analyzed across four domains: authorship, disclosure, permissible uses, and enforcement. Results Sixty journals met inclusion criteria. Most journals prohibit AI systems as authors (58/60, 96.7%), reaffirming human accountability. Disclosure of AI use is mandated by 58/60 journals (96.7%), although reporting requirements vary in placement and specificity. Permissible uses are recognized by 58/60 journals (96.7%), generally limited to language editing and formatting under human supervision, while autonomous content generation or interpretation is discouraged. Enforcement provisions are present in 21/60 journals (35.0%), indicating incomplete standardization. At publisher level, disclosure adoption is universal in Elsevier (17/17), Springer Nature (20/20), AACR (6/6), Wiley (6/6), and AMA (1/1), and present in 8/10 journals in the “Other” category. Enforcement varies widely across publishers. Discussion Editorial policies show strong convergence on core principles but remain heterogeneous in implementation, particularly regarding enforcement. We propose a cross-publisher “AI Policy Minimum Dataset” including standardized disclosures, defined permissible uses, and proportionate enforcement mechanisms, supported by transparent and regularly updated policy frameworks. Greater harmonization is essential to ensure integrity, accountability, and equitable use of AI in oncology publishing.

Ähnliche Arbeiten