The use of Artificial Intelligence in neurosurgical manuscript writing: Journal specific policies and their...

The use of Artificial Intelligence (AI) has grown dramatically in recent years. In addition to its use for data analysis, its applications have extended to manuscript writing. In this article, we analyze the policies within top neurosurgical journals surrounding AI use for manuscript writing, its implementation, and whether disclosure of this practice affects article citation metrics.

Neurosurgical journals with h-indices ≥ 100 and with "spin*" or "neurosurg*" (including translations in other languages) and no other medical subspecialty within their title were included (n = 9). Each journal's policy surrounding AI use in manuscript writing was assessed for whether disclosure was mandated, and if so, requirements for the disclosure. A search was performed using each journal's respective database to find articles that disclosed AI use. Data extracted from each article included: article acceptance and online publication date, type of article, section containing the AI disclosure, total citations, AI program used, and the stated purpose of AI use. A cohort of non-AI-assisted articles was created to assess whether AI disclosure impacts the total number of citations received after publication.

All nine journals mandated disclosure, however, there were variations in the contents required from each disclosure, where in the manuscript the disclosure must be, limitations for AI use, and whether the journal provided a template for how to disclose AI use. A total of 67 publications were included in this review. The journal with the greatest number of articles was World Neurosurgery (n = 41, 61 %), and the journal with the greatest percentage of articles published disclosing AI use since January 1, 2022, was Neurosurgical Focus (0.68 %). Despite the low prevalence across all journals assessed, the rate of growth for articles written with AI has steadily increased. No significant difference was found in the total number of citations between articles that disclosed AI and a cohort of similar articles that did not (W = 85.5, p = 0.69562).

The number of articles declaring AI use was lower than expected. However, such articles have been growing exponentially. Policies surrounding AI use and its implementation varied across journals. We therefore provide recommendations to promote similarity in guidelines between journals, as this will lessen confusion among authors and promote transparency within the medical research community.