Large Language Model Authorship in Ophthalmic Publications

To assess for the likely presence of artificial intelligence (AI)-generated text in the published ophthalmology literature.

Abstract text from 27 142 research articles published in 22 journals between May 2020 and May 2025 was evaluated for changes in word-frequency usage with a focus on stylistic words previously found to be associated with large language model (LLM)-generated text. Four commercial AI-detection services (ZeroGPT, Writer.com, Winston AI, GPTZero) were first validated against control articles with GPTZero showing the best performance, which was then used to detect the presence of AI-generated text in 50 full articles from each journal. For the large-scale screening, research articles and commentary publications (e.g., editorials) were scored at the section and sentence level and compared in the pre- versus post-ChatGPT publication time periods.

Since the release of ChatGPT in 2022, a marked increase in previously rarely used stylistic words was observed with at least a 2-fold usage increase observed in 20% of ophthalmology abstracts. With full article text evaluation, GPTZero scores increased after the release of ChatGPT across all research article sections (e.g., abstract, introduction) and commentary articles. By 2025, 25.7% of sampled research articles and 21.6% of commentary articles contained AI-likelihood scores of more than 2 standard deviations above the baseline. Sentence-level analysis showed that among those publications containing outlier scores, 22.3% of sentences in research articles and 90% of sentences in commentary articles were likely written by AI. Use of AI was not disclosed among any of the publications with outlier scores.

Artificial intelligence brings significant promise in its ability to facilitate both scientific and medical advances. As these tools become more powerful, disclosure regarding the manner of their use becomes increasingly important. We show that LLM-generated text is increasingly present in the ophthalmic literature and is rarely disclosed. Without disclosure requirements and editorial oversight, there is a significant risk that undisclosed LLM usage will continue to increase and may jeopardize authorship integrity and long-term reliability of published findings.

Proprietary or commercial disclosure may be found after the references.