J Neurosurg Spine. 2026 May 29:1-6. doi: 10.3171/2025.8.SPINE25389. Online ahead of print.
ABSTRACT
OBJECTIVE: With the advent of artificial intelligence (AI), scientific research and writing has benefitted from large language models to generate hypotheses, evaluate data, and draft manuscripts. However, this brings into question the prevalence, impact, and ethics of AI writing assistance on published literature. The purpose of this study was to quantify the extent of AI involvement in published spine articles and establish a statistical threshold for scientific integrity.
METHODS: Spine-focused clinical journals were selected for their impact factor and comprehensive representation of the specialty. All full-length research articles published in 2005 and 2023-2024 in these journals were extracted. ZeroGPT was used to assess AI content in each article. Baseline AI utilization was evaluated on the 2005 data, with 2 standard deviations above the mean serving as the threshold for significant AI usage. Based on pre-AI era articles, a threshold ZeroGPT score of 48.8% was established. Articles exceeding this threshold in the 2023-2024 data were assessed across spine journals and years of publication.
RESULTS: In total, 2790 post-AI articles published across 6 spine journals in 2023-2024 were examined. Among these spine journal articles, 25.7% were considered to have significant AI involvement. AI involvement varied significantly across spine journals, ranging from 20.2% for Spine (Phila Pa 1976) to 31.1% for Journal of Neurosurgery: Spine (p < 0.01). Likewise, AI involvement varied significantly across the years, with peak utilization at 32.0% at the start of 2023 and plateau in utilization at 20.7% by the second quarter of 2024 (p < 0.01).
CONCLUSIONS: AI involvement in drafting manuscripts was observed in 25% of articles in recent spine literature. Although the use of AI has plateaued since mid-2024, likely due to the implementation of clear ethical guidelines and utilization of improved detection tools, continued efforts should be made with the evolving AI landscape to the ensure quality, authenticity, and integrity of spine research.
PMID:42214108 | DOI:10.3171/2025.8.SPINE25389
