A Systematic Review of the Limitations and Associated Opportunities of ChatGPT

Figures & data

Table 1. Inclusion and exclusion criteria.

Criteria	Inclusion	Exclusion
Type of publication	Journal article	Reviews Conceptual papers Editorial Conference proceedings Reports Thesis
Study Focus	Reporting empirical findings Conducting evaluation of ChatGPT’s performance	Conceptual papers Papers that solely report perception and attitude data without an evaluation of ChatGPT’s performance
Language of publication	English	Other languages
Date range	No date constraint

Figure 1. Flowchart of stages of the review.

Figure 2. Distribution of the fields of research evaluating ChatGPT’s limitations.

Figure 3. Limitations of ChatGPT (n = 71 instances of limitations).

Table 2. Summary of the limitations of ChatGPT.

Types of ChatGPT limitations	Frequency (percentage)
Accuracy and reliability	32 (47.06%)
Critical and problem-solving thinking	15 (22.06%)
Ethical, legal, and privacy concerns	9 (13.24%)
Potential adverse effect on users’ learning and development	8 (11.76%)
Input and output technical constraints	7 (10.29%)

Table 3. Opportunities of ChatGPT identified in the included studies (n = 44 instances of opportunities).

The opportunities of ChatGPT	Frequency	Percentage (%)
Human like interaction and assistance	6	13.64
Education support and skill development	16	36.36
Task automation and workflow enhancement	11	25
Content creation and ideation	4	9.09
Information retrieval and application	7	15.91

Figure 4. Opportunities for ChatGPT application.

Figure 5. The cyclical evolution of ChatGPT by researchers, practitioners, and developers.

Table

Study, Country, ChatGPT version	Aims	Method	Limitations	Opportunities
Alafnan et al. (2023) NA Kuwait	To investigate opportunities and challenges in using ChatGPT for students and instructions of communication and writing courses	ChatGPT was asked to generate responses to 30 theory-based questions (x 5 times for each question). Its responses were checked by Turnitin software and experts.	potential adverse effect on students’ learning and development if used inappropriately and unethically	technology-enhanced teaching and input for learning and discussion
Ali (2023) NA India	To test ChatGPT’s knowledge and opinion on a controversial health topic	21 prompts were inputted into the ChatGPT for responses, which were evaluated by experts (i.e., surgeons)	several factual inaccuracies; generic response to controversial topics lacking evidence support; below average level of accuracy (40%)	N/A
Amin et al. (2023) NA Germany	To evaluate ChatGPT’s capacity to perform text classification	ChatGPT was asked to predict personalities, sentiment analysis and suicide tendency based on prompts crafted from three relevant datasets on these topics	lower performance compared to a specialized language model (RoBERTa-base)	decent performance compared to other text classification models (i.e., Word2Vec and bag-of-words baseline) robustness against noisy data no training by users needed for ChatGPT
Ariyaratne et al. (2023) ChatGPT-3 UK	To compare the article writing of ChatGPT and humans	ChatGPT was asked to write about a radiology topic, which was then assessed on a 5 point scale from being bad and inaccurate to being excellent and accurate by radiologists	4 out of 5 articles written by ChatGPT being factually inaccurate providing risky medical suggestions fictitious references	N/A
Au Yeung et al. (2023) NA UK	To test ChatGPT’s capacity to predict medical diagnoses	ChatGPT was given clinical vignettes and asked to predict diagnoses.	One or more critical diagnoses were missing in 60% of responses of ChatGPT general prediction of diseases only potential bias in clinical diagnosis against Black people “takes the truth of prompts at face-value”, which influences the accuracy of its response	N/A
Cadamuro et al. (2023) NA Austria, Italy, Croatia	To test ChatGPT’s capability to interpret laboratory test results	ChatGPT was asked to interpret 10 simulated laboratory reports, drafted as optimized prompts. Its output was evaluated by experts in terms of relevance, accuracy, helpfulness and safety	superficial interpretations, most of which lack coherence more suitable for test-by-test interpretation	able to recognise all laboratory tests
Cascella et al. (2023) NA Italy	To test ChatGPT’s use in healthcare context	ChatGPT was provided with some input and then asked to: - compose a medical note for a patient admitted to an emergency - write a research conclusion based on some information bout the research method and finding - write an abstract based on csv (comma-separated values) formatted data	lack capability in interpreting or explaining causal relations among components/ conditions no performance of statistical analysis often not aware of limitations unless requested	aiding in the research process by generating hypothesis, exploring literature, extracting important information communicating research findings in a clear and understandable manner
Clark (2023) NA USA	To examine the capability of ChatGPT in answering a chemistry test	ChatGPT was used to answer closed (multiple choice) and opened ended questions for a chemistry test	inadequate in problem solving or answering questions requiring specific skills only able to achieve 44% score in the chemistry test, i.e., well below the class’s average score at 69% providing seemingly logical but flawed explanations not well-equipped for generating sample responses for exam purpose	potential use to create assignments for students to analyze and improve its response
Day (2023) NA Canada	To investigate the accuracy of references generated by ChatGPT	ChatGPT was asked to answer questions on various topics commonly of interest to geographers	References generated through a predictive process rather than facts subject matter knowledge is required to detect incorrect information, a skill students need to develop	a supporting tool for teaching writing
Duong and Solomon (2023) NA USA	To assess ChatGPT performance in answering questions related to biomedical field	The responses from ChatGPT to 85 multiple-choice questions on human genetics were contrasted with human responses.	not particularly adept at answering critical thinking and calculation-based questions but more suitable for memorisation-based questions inconsistency in answers and explanations where one might select the wrong answer but then provide a correct explanation not suitable for clinical or high-stake uses	rapid and accurate responses to genetic questions potential use to support healthcare professionals in treatment and diagnosis and patients in having accessible medical information
Fergus et al. (2023) NA UK	To evaluate ChatGPT’s response to year end exam assessments	ChatGPT was asked to answer exam questions from two modules of a pharmaceutical program	failing to pass year end exams with the total grade on module 1 and 2 being 34.1% and 18.3%, respectively unable to respond to non-text questions ChatGPT-generated texts not being detected by Turnitin	able to provide well-articulated answers to text-based questions catalyst for discussion on academic integrity and assessment design
Giannos and Delardas (2023) ChatGPT-3.5 UK	To test ChatGPT’s performance on several uni admission tests	ChatGPT was asked to respond to 509 multiple-choice questions on various topics. Its responses were evaluated against various skills such as critical thinking, logical thinking, math, problem solving, and reading comprehension.	limited scientific and mathematical skills poor performance on critical thinking and reasoning skills providing more incorrect than correct responses	well-written responses a catalyst for redesigning educational assessment
Gregorcic and Pendrill (2023) NA Sweden	To test effectiveness of ChatGPT in answering basic physics questions	ChatGPT was asked to answer a physics question “A teddy bear is thrown into the air. What is its acceleration in the highest point?”	inaccurate responses with contradictions not yet adequate to be a cheating tool for physics student or as a physics tutor	potential use for generating lesson materials
Hoch et al. (2023) NA Germany	To test ChatGPT’s performance on a board certification exam	ChatGPT was asked to answer 2576 single‑choice and multiple‑choice board certification preparation questions	limited performance depending on the test/ question format and specific domain of knowledge; more accurate in allergology (72% correct responses) compared to otolaryngology (i.e., 71% answers being incorrect) better performance in answering open ended questions rather than multiple choice questions	a supplementary tool for otolaryngology board certification preparation
Ibrahim et al. (2023) NA United Arab Emirates	To evaluate potential risk of plagiarism of ChatGPT	ChatGPT was asked to answer questions from two introductory and two advanced tertiary level courses	failing to reach the passing grade in the advanced course questions	excellent grade on questions from the introductory courses
Kortemeyer (2023) NA Switzerland	To assess whether ChatGPT could successfully complete introductory physics courses	ChatGPT's ability to handle calculus-based physics content was assessed by administering representative assessments from a real course. The model’s responses were then graded using the same criteria applied to student work.	demonstrating beginner-like errors Presenting facts and fiction with similar confidence Probabilistic nature leading to inconsistent results core issues remaining for more newer versions	The necessity to develop epistemologies when ChatGPT assumes the role of subject matter experts.
Lahat et al. (2023) NA United Arab Emirates	To test ChatGPT’s performance in answer patients’ real-life questions	ChatGPT was asked to answer 110 real-life questions from the patients	moderately accurate and reliable only quality of responses depending on question input	a useful source of reference information
Lai (2023) ChatGPT 3.5 Canada	To evaluate ChatGPT's proficiency in managing various question types and difficulty levels related to references in library services.	ChatGPT was assigned to answer questions about references received by McGill University’s library services. Its responses were subsequently assessed using rubrics that considered completeness, accuracy, and the provision of additional references if the user’s inquiry was not fully addressed.	Struggling with factual accuracy in its response Difficulty handling advanced questions Failing to answer questions requiring nuance, additional resources and referrals	Leveraging ChatGPT as a tool for crafting neutral-tone letters and professional responses.
McIntosh et al. (2024) ChatGPT 3.5 and 4.0 Australia & New Zealand	to assess how different GPT models respond to a Culturally Sensitive Test aimed at detecting hallucinations across diverse cultural and linguistic contexts	Different versions of ChatGPT underwent a Culturally Sensitive Test comprising 70 questions spanning real-world contexts. Model responses were scored as 0 for hallucinated and 1 for non-hallucinated answers.	Hallucinations Ethical concerns Inconsistent performance	NA
Nikolic et al. (2023) NA Australia	To examine ChatGPT’s response to assessment prompts	ChatGPT was asked to respond to engineering assessment prompts from 10 subjects across 7 Australian universities	ChatGPT’s response having word limit, often being generic, lacking specific details, fabricating answers and inaccurate calculations requiring pre-training with background information, which can be time consuming	passable responses from ChatGPT with minimised changes to authentic assessment input
Parsons and Curry (2024) ChatGPT-3 USA	To explore ChatGPT's capability in fulfilling graduate-level instructional design assignments.	This research employed a needs, task, and learner analysis to evaluate ChatGPT's capacity to generate instructional materials for a 12th-grade media literacy module. Expert evaluation and grading rubrics were then used to benchmark the quality of the bot’s outputs.	Struggling to adapting their response to specific context Only including knowledge prior to September 2021 Providing generic and superficial responses when asked to contextualise its responses Responses depending on the complexity and format of questions	Input for teacher and curriculum specialist training in integrating AI capabilities.
Prieto et al. (2023) ChatGPT-3.5 USA, United Arab Emirates	To test ChatGPT’s performance to create a construction project schedule	ChatGPT was asked to generate a construction schedule for a simple project	generic responses and fabricating (incorrect) answers quality of response largely depending on the input/ prompt	able to generate coherent schedule to fulfill the task requirements potential for automating preliminary and time-consuming tasks
Puthenpura et al. (2023) NA USA	To explore the benefit of ChatGPT in assisting writing up a case report	Carefully drafted prompts about a case (i.e., based on case presentation, diagnostic test results and treatment results) was inputted into ChatGPT for response.	ChatGPT’s response containing incomplete information, which is difficult to interpret without subject matter knowledge reference hallucination plagiarism concerns	an assisting tool in streamlining the writing process
Rahman and Watanobe (2023) NA Bangladesh, Japan	To evaluate ChatGPT’s performance in assisting students in learning coding skills	ChatGPT was asked to generate codes based on clear or partially clear information as well as correct errors in codes	poor mathematical skills: failing calculation (counting numbers) that elementary children can do codes generated by ChatGPT may including errors that require human check concerns for students’ potential overreliance on ChatGPT	nearly precise responses to technical queries across a diverse array of subjects.
Rozado (2023) NA New Zealand	To evaluate potential political bias in ChatGPT’s response	ChatGPT was asked to answer 15 different political orientation tests	ChatGPT consistently demonstrating bias toward left-wing political viewpoints (14/15 tests)	N/A
Sallam et al. (2023) NA Jordan	To examine pros and cons of ChatGPT in health and public health education	ChatGPT was asked to respond to prompts about medical, dental and pharmacy topics, each with 5 prompts; its responses were then assessed by experts in terms of conciseness, accuracy and clarity	inaccurate and biased content privacy concerns students’ potential deterioration in critical thinking due to overreliance on ChatGPT	enhancing personalized learning, clinical reasoning skills and understanding of intricate medical concepts
Sanmarchi et al. (2023) ChatGPT-3 Italy	To examine the potential of ChatGPT in designing and conducting epidemiological study	ChatGPT was asked to suggest study questions and design based on an existing paper and then evaluate its response in light of coherence and relevance (by 3 senior researchers)	ethical and legal consequences due to inaccurate data reproducibility issue with ChatGPT due to its inconsistent response inadequate to design conceptual and structure of the study/paper	a valuable support for researchers in setting up an epidemiological study with its response brinh most effective in method, data analysis and offering recommendations
Segal and Khanna (2023) NA USA	To investigate both the capabilities and constraints of ChatGPT in aiding the composition of a case report	Asked to compose a text about a case based on crafted prompts with relevant medical information	providing erroneous description of the genetics and and overestimating the condition of the disease hallucinating references	can be used to generate a rough draft for research and writing purposes
Seth et al. (2023) ChatGPT-3 Australia, Denmark	To test the value of ChatGPT’s input in medical field (i.e, thumb arthritis), particularly for research writing	ChatGPT was asked to answer 5 questions about plastic surgery regarding thumb arthritis	superficial information; not creative and cannot be used to generate plastic surgery solutions hallucinating references able to provide accurate and relevant information (albeit superficial)	N/A
Shoufan (2023) NA United Arab Emirates	to assess ChatGPT's effectiveness in aiding students with no prior knowledge in answering assessment questions	Computer engineering students (experiment group: n = 41–56) used ChatGPT to answer previous test questions before learning about the related topics. Their scores were compared with those of previous-term students (control group: n = 24–61) who answered the same questions in a quiz or exam setting.	Struggling with tasks involving code completion, image analysis, and consistency Performance varying depending on the type and format of questions Providing potentially misleading and incomplete responses	Awareness of ChatGPT limitations shapes educational practices, prompting adjustments to assessment tasks.
Stojanov (2023) ChatGPT-3.5 New Zealand	To report on experiences of using and limitations related to ChatGPT	An ethnographic study by the author where he learnt how to use ChatGPT, had conversation with it and reflected on his experiences	responses with superficial and potentially contradictory information	providing good general knowledge of technical topics in a prompt and efficient manner serving as a learning aid or “a more knowledgeable other” (p. 1)
Thirunavukarasu et al. (2023) NA UK	To evaluate strength and weaknesses of ChatGPT in general practitioner setting	ChatGPT was asked to answer questions of Applied Knowledge Test (AKT) (i.e., a medical test)	unable to achieve sufficient scores to pass the test (60.17% vs 70.42% required to pass) performance quality inconsistent with the difficulty levels of the test questions	achieving a level of proficiency comparable to that of a human expert could be used to automate tasks or as an assistant in clinical settings
Wagner and Ertl-Wagner (2023) ChatGPT-3 Canada	To test ChatGPT’s accuracy and reliability in answering radiologist questions	ChatGPT was asked to answer 88 questions and evaluated by radiologists including the authenticity of its answer	providing inaccurate responses (or responses with errors): 33% hallucinating references: 63.8% self-created references	N/A

Note. NA = Not applicable (i.e., the study did not provide sufficient information to determine the version of ChatGPT used).

Alafnan, M. A., Dishari, S., Jovic, M., & Lomidze, K. (2023). ChatGPT as an educational tool: Opportunities, challenges, and recommendations for communication, business writing, and composition courses. Journal of Artificial Intelligence and Technology, 3(2), 60–68. https://doi.org/10.37965/jait.2023.0184

 Google Scholar

Ali, M. J. (2023). ChatGPT and lacrimal drainage disorders: Performance and scope of improvement. Ophthalmic Plastic and Reconstructive Surgery, 39(5), 515–514. https://doi.org/10.1097/IOP.0000000000002418

 PubMed Web of Science ®Google Scholar

Amin, M. M., Cambria, E., & Schuller, B. W. (2023). Will affective computing emerge from foundation models and general artificial intelligence? A first evaluation of ChatGPT. IEEE Intelligent Systems, 38(2), 15–23. https://doi.org/10.1109/MIS.2023.3254179

 Web of Science ®Google Scholar

Ariyaratne, S., Iyengar, K. P., Nischal, N., Chitti Babu, N., & Botchu, R. (2023). A comparison of ChatGPT-generated articles with human-written articles. Skeletal Radiology, 52(9), 1755–1758. https://doi.org/10.1007/s00256-023-04340-5

 PubMed Web of Science ®Google Scholar

Au Yeung, J., Kraljevic, Z., Luintel, A., Balston, A., Idowu, E., Dobson, R. J., & Teo, J. T. (2023). AI chatbots not yet ready for clinical use. Frontiers in Digital Health, 5, 1161098. https://doi.org/10.3389/fdgth.2023.1161098

 PubMedGoogle Scholar

Cadamuro, J., Cabitza, F., Debeljak, Z., De Bruyne, S., Frans, G., Perez, S. M., Ozdemir, H., Tolios, A., Carobene, A., & Padoan, A. (2023). Potentials and pitfalls of ChatGPT and natural-language artificial intelligence models for the understanding of laboratory medicine test results. An assessment by the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group. Clinical Chemistry and Laboratory Medicine, 61(7), 1158–1166. https://doi.org/10.1515/cclm-2023-0355

 Web of Science ®Google Scholar

Cascella, M., Montomoli, J., Bellini, V., & Bignami, E. (2023). Evaluating the feasibility of ChatGPT in healthcare: An analysis of multiple clinical and research scenarios. Journal of Medical Systems, 47(1), 33. https://doi.org/10.1007/s10916-023-01925-4

 PubMed Web of Science ®Google Scholar

Clark, T. M. (2023). Investigating the use of an artificial intelligence chatbot with general chemistry exam questions. Journal of Chemical Education, 100(5), 1905–1916. https://doi.org/10.1021/acs.jchemed.3c00027

 Web of Science ®Google Scholar

Day, T. (2023). A preliminary investigation of fake peer-reviewed citations and references generated by ChatGPT. The Professional Geographer, 75(6), 1024–1027. https://doi.org/10.1080/00330124.2023.2190373

 Web of Science ®Google Scholar

Duong, D., & Solomon, B. D. (2023). Analysis of large-language model versus human performance for genetics questions. European Journal of Human Genetics, 32(4), 466–468. https://doi.org/10.1038/s41431-023-01396-8

 PubMed Web of Science ®Google Scholar

Fergus, S., Botha, M., & Ostovar, M. (2023). Evaluating academic answers generated using ChatGPT. Journal of Chemical Education, 100(4), 1672–1675. https://doi.org/10.1021/acs.jchemed.3c00087

 Web of Science ®Google Scholar

Giannos, P., & Delardas, O. (2023). Performance of ChatGPT on UK standardized admission tests: Insights from the BMAT, TMUA, LNAT, and TSA examinations. JMIR Medical Education, 9, e47737. https://doi.org/10.2196/47737

 PubMedGoogle Scholar

Gregorcic, B., & Pendrill, A. M. (2023). ChatGPT and the frustrated Socrates. Physics Education, 58(3), 035021. https://doi.org/10.1088/1361-6552/acc299

 Google Scholar

Hoch, C. C., Wollenberg, B., Lüers, J. C., Knoedler, S., Knoedler, L., Frank, K., Cotofana, S., & Alfertshofer, M. (2023). ChatGPT’s quiz skills in different otolaryngology subspecialties: An analysis of 2576 single-choice and multiple-choice board certification preparation questions. European Archives of Oto-Rhino-Laryngology, 280(9), 4271–4278. https://doi.org/10.1007/s00405-023-08051-4

 PubMed Web of Science ®Google Scholar

Ibrahim, H., Asim, R., Zaffar, F., Rahwan, T., & Zaki, Y. (2023). Rethinking homework in the age of artificial intelligence. IEEE Intelligent Systems, 38(2), 24–27. https://doi.org/10.1109/MIS.2023.3255599

 Web of Science ®Google Scholar

Kortemeyer, G. (2023). Could an artificial-intelligence agent pass an introductory physics course? Physical Review Physics Education Research, 19(1), 1–18. https://doi.org/10.1103/PhysRevPhysEducRes.19.010132

 Web of Science ®Google Scholar

Lahat, A., Shachar, E., Avidan, B., Glicksberg, B., & Klang, E. (2023). Evaluating the utility of a large language model in answering common patients’ gastrointestinal health-related questions: Are we there yet? Diagnostics, 13(11), 1950. https://doi.org/10.3390/diagnostics13111950

 Web of Science ®Google Scholar

Lai, K. (2023). How well does ChatGPT handle reference inquiries? An analysis based on question types and question complexities. College & Research Libraries, 84(6), 974–995. https://doi.org/10.5860/crl.84.6.974

 Web of Science ®Google Scholar

McIntosh, T. R., Liu, T., Susnjak, T., Watters, P., Ng, A., & Halgamuge, M. N. (2024). A culturally sensitive test to evaluate nuanced GPT hallucination. IEEE Transactions on Artificial Intelligence, 1–13. https://doi.org/10.1109/TAI.2023.3332837

 Google Scholar

Nikolic, S., Daniel, S., Haque, R., Belkina, M., Hassan, G. M., Grundy, S., Lyden, S., Neal, P., & Sandison, C. (2023). ChatGPT versus engineering education assessment: A multidisciplinary and multi-institutional benchmarking and analysis of this generative artificial intelligence tool to investigate assessment integrity. European Journal of Engineering Education, 48(4), 559–614. https://doi.org/10.1080/03043797.2023.2213169

 Web of Science ®Google Scholar

Parsons, B., & Curry, J. H. (2024). Can ChatGPT pass graduate-level instructional design assignments? Potential implications of artificial intelligence in education and a call to action. TechTrends, 68(1), 67–78. https://doi.org/10.1007/s11528-023-00912-3

 Web of Science ®Google Scholar

Prieto, S. A., Mengiste, E. T., & García de Soto, B. (2023). Investigating the use of ChatGPT for the scheduling of construction projects. Buildings, 13(4), 857. https://doi.org/10.3390/buildings13040857

 Web of Science ®Google Scholar

Puthenpura, V., Nadkarni, S., DiLuna, M., Hieftje, K., & Marks, A. (2023). Personality changes and staring spells in a 12-year-old child: A case report incorporating ChatGPT, a natural language processing tool driven by artificial intelligence (AI). Cureus, 15(3), e36408. https://doi.org/10.7759/cureus.36408

 PubMed Web of Science ®Google Scholar

Rahman, M. M., & Watanobe, Y. (2023). ChatGPT for education and research: Opportunities, threats, and strategies. Applied Sciences, 13(9), 5783. https://doi.org/10.3390/app13095783

 Google Scholar

Rozado, D. (2023). The political biases of ChatGPT. Social Sciences, 12(3), 148. https://doi.org/10.3390/socsci12030148

 Google Scholar

Sallam, M., Salim, N., Barakat, M., & Al-Tammemi, A. (2023). ChatGPT applications in medical, dental, pharmacy, and public health education: A descriptive study highlighting the advantages and limitations. Narra J, 3(1), e103. https://doi.org/10.52225/narra.v3i1.103

 PubMedGoogle Scholar

Sanmarchi, F., Bucci, A., Nuzzolese, A. G., Carullo, G., Toscano, F., Nante, N., & Golinelli, D. (2023). A step-by-step researcher’s guide to the use of an AI-based transformer in epidemiology: An exploratory analysis of ChatGPT using the STROBE checklist for observational studies. Journal of Public Health, 1–36. https://doi.org/10.1007/s10389-023-01936-y

 Web of Science ®Google Scholar

Segal, S., & Khanna, A. K. (2023). Anesthetic management of a patient with juvenile hyaline fibromatosis: A case report written with the assistance of the large language model chatgpt. Cureus, 15(3), e35946. https://doi.org/10.7759/cureus.35946

 PubMed Web of Science ®Google Scholar

Seth, I., Sinkjær Kenney, P., Bulloch, G., Hunter-Smith, D. J., Bo Thomsen, J., & Rozen, W. M. (2023). Artificial or augmented authorship? A conversation with a Chatbot on base of thumb arthritis. Plastic and Reconstructive Surgery. Global Open, 11(5), e4999. https://doi.org/10.1097/GOX.0000000000004999

 PubMedGoogle Scholar

Shoufan, A. (2023). Can students without prior knowledge use ChatGPT to answer test questions? An empirical study. ACM Transactions on Computing Education, 23(4), 1–29. https://doi.org/10.1145/3628162

 Web of Science ®Google Scholar

Stojanov, A. (2023). Learning with ChatGPT 3.5 as a more knowledgeable other: An autoethnographic study. International Journal of Educational Technology in Higher Education, 20(1), 35. https://doi.org/10.1186/s41239-023-00404-7

 Web of Science ®Google Scholar

Thirunavukarasu, A. J., Hassan, R., Mahmood, S., Sanghera, R., Barzangi, K., El Mukashfi, M., & Shah, S. (2023). Trialling a large language model (ChatGPT) in General practice with the applied knowledge test: Observational study demonstrating opportunities and limitations in primary care. JMIR Medical Education, 9, e46599. https://doi.org/10.2196/46599

 PubMedGoogle Scholar

Wagner, M. W., & Ertl-Wagner, B. B. (2023). Accuracy of information and references using ChatGPT-3 for retrieval of clinical radiological information. Canadian Association of Radiologists Journal, 75(1), 69–73. https://doi.org/10.1177/08465371231171125

 PubMed Web of Science ®Google Scholar