Abstract
Study Design Cross-sectional survey study
Background Artificial intelligence (AI) tools are increasingly integrated into various aspects of medicine, including medical research. However, the scope and manner in which early-career surgeons utilize AI tools in their research remain inadequately understood.
Objective This study aimed to investigate the frequency and specific applications of AI tools in medical research among early-career surgeons, including their perceptions, concerns, and outlook regarding AI in research.
Methods A survey comprising 25 questions was distributed among members of an international club of early-career spine surgeons (<10 years of experience). The survey assessed demographics, AI tool utilization, access to AI training resources, and perceptions of AI benefits and concerns in research.
Results Sixty early-career surgeons participated, with 86.7% reporting AI tool use in their research. ChatGPT was the most frequently utilized tool, with a usage rate of 93.1%. AI tools were primarily used for grammatical proofreading (69.6%) and rephrasing (64.3%), while 26.8% of participants used AI for statistical analysis. While 80.4% perceived improved efficiency as a key benefit, 70.0% expressed concerns about reliability. None of the participants had received formal AI training, and only 15.0% had access to AI mentors. Despite these challenges, 91.6% anticipated a positive long-term impact of AI on research.
Conclusion AI tools are widely adopted among early-career surgeons for various research tasks, extending from text generation to data analysis. However, the absence of formal training and concerns regarding the reliability of AI tools underscore the necessity of training for AI integration in medical research.
Clinical Relevance This study provides timely insights into AI adoption patterns among early-career surgeons, highlighting the urgent need for formal AI training programs to ensure responsible research practices.
Level of Evidence 4.
Introduction
A June 2024 headline in The Economist proclaimed that “At least 10% of research may already be co-authored by artificial intelligence (AI).”1 The rise of AI in medicine has ushered in a new era, and many orthopedic surgeons are optimistic about the future integration of AI in their field.2 With the rapid expansion of AI across various aspects of medicine, the field of medical publishing is no exception.3 In the context of medical research, generative AI is considered a useful tool for synthesizing previous studies, identifying knowledge gaps, and generating hypotheses.4 When looking at scientific publishing, generative AI tools have shown the capability to produce scientific content comparable to that written by humans, which was only correctly identified in 68% by reviewers.5,6
Despite this impressive skill set, generative AI comes with a set of limitations. One major issue is that AI generates outputs that sometimes deviate from factual knowledge, known as hallucinations.7 These hallucinations significantly undermine the reliability of generative AI tools in real-world scenarios or medical settings.8 Furthermore, medical AI tools can be biased due to the presence of biases in the training data. If biases exist in the data used to develop the AI tool, these biases will inevitably be reflected in the AI-generated output.9 Consequently, the presence of biases in training data will inevitably influence the accuracy and reliability of the AI’s performance in clinical applications.
Due to these unaddressed limitations, many scientific journal publishers have updated their guidelines.10 For example, some journal publishers such as Springer Nature and the JAMA Network disallow AI tools as credited authors, arguing that authorship implies accountability that AI cannot assume.11,12 Furthermore, the Science family of journals has taken a stricter stance, prohibiting not only AI authorship but also any AI-generated text and figures.10
Hence, it can be assumed that out of the more than 180.5 million users of generative AI, research is not be an expectation.13 Despite numerous studies exploring the capabilities of generative AI in medicine and research, no research has yet investigated the frequency and specific tasks for which researchers currently utilize generative AI in the scientific publishing process. Thus, the purpose of the present study was to investigate the frequency and specific tasks for which young surgeons utilize AI tools to support their medical research.
Materials and Methods
This is a descriptive and cross-sectional study that encompasses an electronic survey of 25 questions.
Participants
A survey was created and distributed among members of the AO Spine Youth Club, an international organization of young spine surgeons who have been out of medical school for fewer than 10 years, all of whom are medically licensed and are either in orthopedic, traumatology, or neurosurgical training.
Survey
The survey was created using Google Forms (Google LLC, Mountain View, CA, USA) and was distributed in electronic form, through a link, and was available in the English language. The survey was developed based on a literature review of AI adoption in health care settings, with input from senior spine surgeons, research methodologists, and AI experts. Questions were designed to assess both quantitative usage patterns and qualitative perceptions through structured and semistructured formats.
After an initial email invitation, eligible participants who did not complete the questionnaire were reminded once after 7 days. Participation in the survey was voluntary, and refusal had no negative consequences. The survey was open to new participants for 21 days starting in July 2024, and all data were collected anonymously. By providing each respondent with a unique link, multiple responses from a single participant were avoided.
The survey consisted of the following 4 sections: (1) demographics, including age, sex, native language, and career stage; (2) utilization of AI in personal research; (3) training resources for generative AI utilization; and (4) benefits and concerns regarding generative AI in research.
Descriptive statistics were performed to report frequencies and percentages. Data were analyzed using the SAS System software version 9.2 (SAS Institute, Inc., Cary, North Carolina, United States).
Results
Participants
A total of 60 early-career surgeons participated in this survey, with the majority aged between 25 and 34 years (73.3%, n = 44). Male participants constituted the majority (63.3%, n = 38). Regarding specialization, orthopedic training was most represented (70.0%, n = 42), followed by trauma surgery (16.7%, n = 10) and neurosurgery (13.3%, n = 8).
The majority of respondents (71.7%, n = 43) were in the early stages of their careers, having graduated from medical school less than 8 years prior. The majority were currently in residency programs (65.0%, n = 39). Most of the respondents were nonnative English speakers (88.3%, n = 53). The comprehensive demographic and professional characteristics of the participants are presented in the Table.
Demographic data of survey respondents with percentage of total respondents and number of respondents in parentheses.
Utilization of AI Tools
In terms of research context, a substantial majority of surveyed surgeons (86.7%, n = 52) reported using AI tools. The frequency of use varied, with the largest group (38.3%, n = 23) employing these tools rarely (defined as one-quarter of their research activities). Other usage patterns included occasional use (half the time) by 21.7% (n = 13), frequent use (three-quarters of the time) by 15.0% (n = 9), and constant use by 11.7% (n = 7; Figure 1).
Frequency of artificial intelligence (AI) usage for research among early-career surgeons who employ AI in their studies.
The adoption of AI tools outside of research showed that nearly all participants (99.0%, n = 59) reported use in daily life. The most common frequency of use was once per week (30.5%, n = 18), while equal proportions (18.6%, n = 11 each) reported using AI tools multiple times daily or once monthly.
AI Tool Utilization and Application in Research
Among the surveyed surgeons, ChatGPT (OpenAI, San Francisco, USA) emerged as the predominant AI tool for research purposes, with a utilization rate of 93.1% (n = 54). The translation tool DEEP-L (Cologne, NRW, Germany) was the second most frequently used, employed by 46.6% (n = 27) of respondents. Grammarly (San Francisco, CA, USA), a grammar correction tool, was utilized by 22.4% (n = 13), while the image generation tool DALL-E (OpenAI, San Francisco, CA, USA) was used by 15.5% (n = 9) of participants. Other generative AI tools, such as Gemini (Google, San Francisco, CA, USA) and Claude (Anthropic, San Francisco, CA, USA), were employed by 13.8% (n = 8) and 5.4% (n = 3) of respondents, respectively.
Within the context of scientific research processes, AI tools were primarily employed for grammatically proofreading (69.6%, n = 39) and rephrasing (64.3%, n = 36). Additional applications included literature searches (30.4%, n = 17), paper summarization (37.5%, n = 21), and abstract writing (32.1%, n = 18). Notably, 26.8% (n = 15) of respondents reported using AI tools for statistical analysis, while 19.6% (n = 11) utilized them for image generation. The distribution of AI integration across various research project stages is illustrated in Figure 2.
Research tasks for which early-career surgeons incorporated artificial intelligence (AI) tools.
When queried about the research phase most positively impacted by AI, a majority of respondents (70.4%, n = 38) identified the writing and publishing stage. Data analysis (29.6%, n = 16) and idea generation (25.9%, n = 14) were also noted as significantly benefiting from AI integration. A smaller proportion (7.4%, n = 4) reported that AI most positively influenced their peer review process.
The perceived benefits of AI tool usage in research were primarily associated with improved efficiency (80.4%, n = 45) and reliable proofreading capabilities (33.9%, n = 19). Respondents reported that AI integration resulted in enhanced time management (65.0%, n = 39) and increased output quality (33.3%, n = 20). However, 23.3% (n = 14) of participants indicated that AI implementation had no discernible effect on their research processes (Figure 3).
Distribution of perceived benefits of using artificial intelligence (AI) tools in research among early-career surgeons.
Training and Access to AI Expertise
Among the surveyed surgeons, there were no reports of having received specific instruction in AI tool utilization. Only 15.0% (n = 9) of respondents indicated access to mentors or experts for AI-related guidance. Furthermore, a minority of participants (22.0%, n = 13) reported the availability of AI education opportunities at their respective institutions.
Reliability, Concerns, and Trust
Regarding challenges associated with AI tool usage, reliability issues were identified as the primary concern by 70.0% (n = 42) of respondents. Data privacy concerns ranked second, reported by 50.0% (n = 30) of participants, followed by a perceived lack of training, noted by 38.3% (n = 23). Trust in AI-generated outputs varied among respondents. The majority (63.3%, n = 38) expressed moderate trust, while 20.0% (n = 12) reported slight trust. A smaller proportion (11.7%, n = 7) indicated high levels of trust, and 5.0% (n = 3) reported no trust at all in AI outputs. Ethical considerations surrounding AI tools were divided among the participants, with 52.5% (n = 31) expressing concerns and 47.5% (n = 28) reporting no ethical reservations.
Future Outlook and Adoption
The long-term impact of AI tools in research was predominantly viewed positively by the surveyed surgeons, with 91.6% (n = 55) anticipating a beneficial effect. In contrast, 6.7% (n = 4) foresaw a negative impact, and 1.7% (n = 1) expected no significant impact. Regarding the likelihood of adopting new AI tools, a substantial majority of respondents indicated a positive inclination. Specifically, 36.7% (n = 22) reported being very likely to adopt new AI tools, while 46.7% (n = 28) stated they were likely to do so.
Discussion
This study revealed a high adoption rate of AI technologies, with 86.7% of early-career surgeons reporting the use of these tools in their research activities. The tools were primarily employed for grammatical proofreading and rephrasing, but they were also used for literature research, statistical analysis, and image generation. Notably, none of the surgeons had received formal training in utilizing AI tools.
AI’s capability to generate scientific work for medical experts has been shown and raised concerns. Therefore, many scientific journal publishers have found it necessary to update their guidelines, requiring declarations of AI utilization in the submission process via a check box up to prohibiting not only AI authorship but also any AI-generated text and figures.10–12 However, with 86.7% of early-career surgeons reporting using these tools in a varying frequency, from rarely (38.3%) to constant (11.7%) use, for research, one can assume that there is a relevant number of undeclared AI contributions in medical research.
The predominantly used AI tool for research was ChatGPT (93.1%), which aligns with its popular use as a breakthrough tool after its launch in November 2022.14 This tool has exhibited top performance in various research-relevant tasks such as coherent content and essay generation, chatbot responses, language translation, and question answering.15 The second most common tool utilized was a translation tool, DEEP-L (46.6%), which aligns with the 2 most reported applications of AI tools: grammatical proofreading and rephrasing. This finding reflects the fact that most influential medical journals are written in the English language; however, the majority of the participants of this study were nonnative speakers.16 The widespread adoption of AI for improving grammar and writing style could help overcome language barriers, potentially reducing the amount of “lost science” resulting from research not being published in English.17
It is important to note that the application of AI extends beyond text-based tasks, with 26.8% of respondents utilizing these tools for statistical analysis. This finding suggests that AI is not only assisting in the writing and editing phases but is also beginning to play a role in data interpretation and analysis.18 Nonetheless, due to the fast improvement of these tools, newer versions might be more successful.
The perceived benefits of AI tool usage, primarily improved efficiency (80.4%) and enhanced time management (65.0%), indicate that these technologies are addressing the challenges and opportunities of ever-growing data in terms of data capture, storage, manipulation, management, analysis, knowledge extraction, security, privacy, and visualization.19
The high adoption rate of AI tools (86.7%), coupled with significant reliability concerns (70.0%), raises important ethical considerations specific to spine surgery research. Given that spine surgery research directly influences patient care decisions and surgical techniques, the reliability of AI-assisted research requires careful scrutiny. When AI tools are used for tasks such as literature review or manuscript preparation, potential biases in AI language models could lead to selective interpretation or oversight of critical surgical complications or outcomes. Additionally, with 26.8% of participants using AI for statistical analysis despite limited formal AI training, there is a risk of misinterpreting or over-relying on AI-generated analytical outputs. This is particularly concerning in spine surgery research, where statistical conclusions often inform surgical decision-making and risk assessment. These findings underscore the need for establishing clear guidelines for AI use in surgical research and implementing formal training programs to ensure responsible integration of AI tools.
Despite the broad application of AI tools in research, this study reveals a critical gap in AI education and training among early-career surgeons. The complete absence of formal AI training among respondents, coupled with limited access to experts, highlights a pressing need for educational initiatives in this rapidly evolving field. Since AI tools have been shown to have many limitations, such as biased training data sets, fabricated outputs, and no scientific references.8–10 The potential for AI-generated hallucinations and biases is a relevant concern. This lack of formal training might be reflected in the reliability concerns most participants have. However, ethical considerations surrounding AI use in research were distributed evenly among participants. Therefore, considering the potential limitations of AI tools, training should be mandatory, especially concerning the broad application of AI in medical research and publishing. To keep up with the ever-increasing sophistication of AI, orthopedic surgeons must be familiar with and able to implement a variety of AI-based approaches and modalities.20
Despite these challenges, the overwhelmingly positive outlook on the long-term impact of AI in research (91.6%) indicates that early-career surgeons view these tools as essential to the future of medical research. Additionally, the high willingness to adopt new AI tools (83.4% likely or very likely) further underscores this positive sentiment.
Limitations of this study include its focus on early-career surgeons from a specific international club, which may not be representative of the broader medical research community. The current study focused exclusively on members of an international club for early-career spine surgeons, which presents inherent limitations regarding generalizability. Club members might represent a particularly engaged and technologically aware subset of surgeons, potentially leading to selection bias in AI adoption rates. Future research should aim to include a more diverse sample of early-career surgeons from various institutional settings, geographical regions, and professional networks to provide a more comprehensive understanding of AI adoption in medical research across the broader surgical community. However, focusing on early-career surgeons might give a better perspective on future developments in research. Furthermore, the majority of respondents were nonnative English speakers, which might have affected their needs for AI tools and fields of application. Additionally, the self-reported nature of the data may introduce some bias in the results. A limitation of this study’s methodology was its exclusive use of descriptive statistics. While these provided valuable insights into AI tool adoption patterns, the sample size of 60 participants limited the ability to conduct meaningful inferential statistics or correlation analyses. Future studies with larger cohorts could explore specific relationships, such as between AI training access and tool usage patterns or between experience levels and AI trust, to provide deeper statistical insights.
Conclusion
The present study demonstrated that AI tools have become an integral part of the research process for many early-career surgeons, despite a lack of formal training. While the potential benefits are evident, challenges related to reliability, ethics, and proper utilization emphasize the need for comprehensive education, such as AI literacy training focusing on appropriate research applications and clear guidelines in this rapidly evolving landscape of AI in medical research.
Footnotes
Funding This study was funded by a research grant from the Rama and Shashi Marda Foundation.
Declaration of Conflicting Interests D.R.L. is a consultant and on the Advisory Board for Choice Spine; is a consultant for DePuy Synthes; has ownership interest from Woven Orthopedic Technologies, Vestia Ventures MiRus Investment LLC, HS2 LLC, and ISPH II LLC; has received research support from Medtronic Sofamor Danek USA, Inc; has received royalties from Nuvasive, Inc; is on the Advisory Board and has ownership interest from Remedy Logic; is a Consultant and has royalties from Stryker; and is a consultant and has ownership onterest from Viseon, Inc.
- This manuscript is generously published free of charge by ISASS, the International Society for the Advancement of Spine Surgery. Copyright © 2025 ISASS. To see more or order reprints or permissions, see http://ijssurgery.com.
References
- 1.↵At least 10% of research may already be co-authored by AI. The Economist. June 26, 2024. https://www.economist.com/science-and-technology/2024/06/26/at-least-10-of-research-may-already-be-co-authored-by-ai.
- 2.↵Rupp M-C , Moser LB , Hess S , et al . Orthopaedic surgeons display a positive outlook towards artificial intelligence: a survey among members of the AGA Society for Arthroscopy and Joint Surgery. J Exp Orthop. 2024;11(3):e12080. 10.1002/jeo2.12080
- 3.↵Wiechert K , Wang JC , Chapman JR . When will peer-review become A.I.-review? Global Spine J. 2024;14(5):1463. 10.1177/21925682241242892
- 4.↵Khurana D , Koli A , Khatter K , Singh S . Natural language processing: state of the art, current trends and challenges. Multimed Tools Appl. 2023;82(3):3713–3744. 10.1007/s11042-022-13428-4
- 5.↵Gao CA , Howard FM , Markov NS , et al . Comparing scientific abstracts generated by ChatGPT to original abstracts using an artificial intelligence output detector, plagiarism detector, and blinded human reviewers. Scientific Communication and Education. 2022. 10.1101/2022.12.23.521610
- 6.↵Kim HJ , Yang JH , Chang D-G , et al . Assessing the reproducibility of the structured abstracts generated by ChatGPT and Bard compared to human-written abstracts in the field of spine surgery: comparative analysis. J Med Internet Res. 2024;26:e52001. 10.2196/52001
- 7.↵Kaddour J , Harris J , Mozes M , Bradley H , Raileanu R , McHardy R . Challenges and applications of large language models. Computation and Language. 2023. 10.48550/ARXIV.2307.10169
- 8.↵Zhang Y , Li Y , Cui L , et al . Siren’s song in the AI ocean: a survey on hallucination in large language models. Computation and Language. 2023. 10.48550/ARXIV.2307.10169
- 9.↵Ferryman K , Mackintosh M , Ghassemi M . Considering biased data as informative artifacts in AI-assisted health care. N Engl J Med. 2023;389(9):833–838. 10.1056/NEJMra2214964
- 10.↵Loh E . ChatGPT and generative AI chatbots: challenges and opportunities for science, medicine and medical leaders. BMJ Lead. 2023;8(1):51–54:e000797. 10.1136/leader-2023-000797
- 11.↵Tools such as ChatGPT threaten transparent science; here are our ground rules for their use. Nature New Biol. 2023;613(7945):612–612. 10.1038/d41586-023-00191-1
- 12.↵Flanagin A , Bibbins-Domingo K , Berkwits M , Christiansen SL . Nonhuman “authors” and implications for the integrity of scientific publication and medical knowledge. JAMA. 2023;329(8):637. 10.1001/jama.2023.1344
- 13.↵Elad B . OpenAI Statistics 2024 by Demographics, Products, Revenue and Growth . Enterprise Apps Today. May 20, 2024. https://www.enterpriseappstoday.com/stats/openai-statistics.html.
- 14.↵Rahman MdM , Watanobe Y . ChatGPT for education and research: opportunities, threats, and strategies. Appl Sci (Basel). 2023;13(9):5783. 10.3390/app13095783
- 15.↵Kasneci E , Sessler K , Küchemann S , et al . ChatGPT for good? On opportunities and challenges of large language models for education. Learn Individ Differ. 2023;103:102274. 10.1016/j.lindif.2023.102274
- 16.↵Diekhoff T , Schlattmann P , Dewey M . Impact of article language in multi-language medical journals—a bibliometric analysis of self-citations and impact factor. PLoS ONE. 2013;8(10):e76816. 10.1371/journal.pone.0076816
- 17.↵Pascual-Leone N , Liu JW , Beschloss A , Chenna SS , Saifi C . The language of all medical publications and spine publications from 1950 to 2020. N Am Spine Soc J. 2022;10:100118. 10.1016/j.xnsj.2022.100118
- 18.↵Ignjatović A , Stevanović L . Efficacy and limitations of ChatGPT as a biostatistical problem-solving tool in medical education in Serbia: a descriptive study. J Educ Eval Health Prof. 2023;20:28. 10.3352/jeehp.2023.20.28
- 19.↵Gupta D , Rani R . A study of big data evolution and research challenges. J Information Sci. 2019;45(3):322–340. 10.1177/0165551518789880
- 20.↵Fayed AM , Mansur NSB , de Carvalho KA , Behrens A , D’Hooghe P , de Cesar Netto C . Artificial intelligence and ChatGPT in orthopaedics and sports medicine. J Exp Orthop. 2023;10(1):74. 10.1186/s40634-023-00642-8
