ChatGPT for diabetes education: potential, accuracy, and accessibility in patient support
Open accessAbstract
Background: Diabetes mellitus is a chronic metabolic disease with rising global prevalence. Adequate patient education is essential to encourage self-management and reduce complications. Artificial intelligence applications such as ChatGPT have emerged as potential supplementary resources for patient education alongside the broader integration of technology in healthcare. Methods: A cross-sectional evaluation was conducted using ten frequently asked questions (FAQs) on diabetes, selected from the Diabetic Association of India and the International Diabetes Federation. ChatGPT-4o (accessed via the web interface in March 2025) generated responses to each question in separate, stand-alone chat sessions to simulate typical patient interactions. Five board-certified endocrinologists (diabetologists) with a mean clinical experience of ≥10 years independently evaluated the responses using a 4-point Likert scale across five domains: overall quality, content accuracy, clarity, relevance, and trustworthiness. Final domain scores were computed as the mean of all five raters’ scores. Readability was assessed using the Flesch Reading Ease Score (FRES) and Flesch-Kincaid Grade Level (FKGL). All readability analyses apply exclusively to the English-language outputs generated in this study. Results: The mean FRES was 38.19 and the mean FKGL was 16.87, indicating a reading level appropriate for college-educated individuals and substantially above the recommended sixth-grade benchmark for patient health materials. Mean response length was 300 ± 100 words across the ten prompts. Expert ratings were generally high: aggregated mean scores (±SD) were 4.0 (±0.0) for content accuracy and overall quality, 3.98 (±0.10) for relevance, and 3.9 (±0.20) for clarity and trustworthiness. No clinically inaccurate statements were identified by the raters; however, the high scores and narrow score range indicate a potential ceiling effect that limits discrimination between responses. Raters expressed concern about linguistic complexity, which may impede comprehension among patients with limited health literacy. Conclusions: ChatGPT-4o generated generally accurate and relevant diabetes education content, suggesting potential as a supplementary tool in diabetes care. However, the high reading-level complexity, small evaluation scope (ten prompts, one model, one session), and English-only assessment limit the generalisability of these findings. AI-generated content should supplement, not replace, clinician-led education. Future work should address language simplification, multilingual evaluation, and longitudinal assessment of patient outcomes.
AI snapshot
Written from the paper’s full text.
TL;DR
ChatGPT-4o produced generally accurate and relevant answers to ten diabetes FAQs, with high expert ratings but readability far above recommended patient-education levels. The authors conclude it may supplement, not replace, clinician-led diabetes education, while noting small scope, single-run generation, and English-only assessment.
Key findings
- Five board-certified endocrinologists rated ChatGPT-4o responses highly across five domains: aggregated mean ± SD was 4.0 ± 0.0 for content accuracy and overall quality, 3.98 ± 0.06 for relevance, and 3.9 ± 0.1 for clarity and trustworthiness; mean across all domains was 3.96 ± 0.04.
- No clinically inaccurate statements meeting the pre-specified error definition were identified by raters, but the narrow score range and proximity to the scale ceiling indicated a potential ceiling effect limiting discrimination between responses.
- Readability was consistently college-level and above the recommended sixth-grade benchmark: mean FRES 38.19 ± 8.42, FKGL 16.87 ± 2.31, SMOG 17.24 ± 1.98, and Gunning Fog 19.63 ± 2.54.
- Mean response length was 312 ± 91 words (range 198–487 words) in Table 4; the abstract reports 300 ± 100 words across the ten prompts.
Methods
- Cross-sectional evaluation of ten diabetes FAQs selected from the Diabetic Association of India and the International Diabetes Federation, covering symptoms, management, diet, complications, monitoring, prevention, medications, reversal, and insulin therapy.
- ChatGPT-4o responses were generated via the web interface in March 2025 in separate, stand-alone chat sessions with web browsing and plugins disabled; no prompt engineering or response regeneration was used, and one response was generated per prompt.
- Five board-certified endocrinologists with ≥10 years’ clinical experience independently rated all responses on a 4-point Likert scale across overall quality, content accuracy, clarity, relevance, and trustworthiness; readability was assessed with FRES, FKGL, SMOG, and Gunning Fog using textstat Python library v0.7.0.
Limitations
- Stated: the evaluation was restricted to ten prompts, one model (ChatGPT-4o), and a single session; a single response was generated per prompt, so LLM output variability was not assessed.
- Stated: only five raters were used, formal inter-rater reliability statistics were not computed, raters were not blinded to the AI-generated nature of the content, and no formal adjudication framework or external guideline reference-check was applied to error assessment.
- Stated: all readability analyses were English-only and cannot be extrapolated to other languages; outputs may vary across sessions, model updates, or access configurations.
- Evident from the text: the study assessed no patient comprehension or clinical outcomes, did not compare multiple LLMs, and its hypothesis-generating design limits generalisability beyond the evaluated prompt set.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
5 of 6 cited sources checked · 3 supported · 1 flagged · accuracy not yet measured
- Findingown data
Mean response length was 312 words (range 198-487) across the ten responses.
“Mean response length was 312 words (range: 198–487 words).”
Quote in the paperNot a substantial claim - Findingown data
Mean SMOG Index across the ten ChatGPT-4o responses was 17.24 (SD 1.98).
“Mean (±SD) values were: FRES 38.19 (±8.42), FKGL 16.87 (±2.31), SMOG Index 17.24 (±1.98), and Gunning Fog Index 19.63 (±2.54).”
Quote in the paperSubstantial claim - Findingown data
Mean FKGL across the ten ChatGPT-4o responses was 16.87 (SD 2.31).
“Mean (±SD) values were: FRES 38.19 (±8.42), FKGL 16.87 (±2.31), SMOG Index 17.24 (±1.98), and Gunning Fog Index 19.63 (±2.54).”
Quote in the paperSubstantial claim - Findingown data
Mean Gunning Fog Index across the ten ChatGPT-4o responses was 19.63 (SD 2.54).
“Mean (±SD) values were: FRES 38.19 (±8.42), FKGL 16.87 (±2.31), SMOG Index 17.24 (±1.98), and Gunning Fog Index 19.63 (±2.54).”
Quote in the paperSubstantial claim - Findingown data
No clinically inaccurate statements meeting the pre-specified error definition were identified by the raters in any of the ten responses.
“Raters did not identify any clinically inaccurate statements meeting the pre-specified error definition (factual inaccuracy, clinically unsafe recommendation, or significant omission of safety-critical information) in any of the ten responses”
Quote in the paperSubstantial claim- Rests on an assumption: The operational definition of an 'error' adequately captures clinically significant inaccuracies.
- Findingown data
Mean trustworthiness score across the ten responses was 3.9 ± 0.1.
“Mean ± SD | 4.0 ± 0.0 | 4.0 ± 0.0 | 3.9 ± 0.1 | 3.98 ± 0.06 | 3.9 ± 0.1 | 3.96 ± 0.04”
Quote in the paperNot a substantial claim - Findingown data
Mean FRES across the ten ChatGPT-4o responses was 38.19 (SD 8.42).
“Mean (±SD) values were: FRES 38.19 (±8.42), FKGL 16.87 (±2.31), SMOG Index 17.24 (±1.98), and Gunning Fog Index 19.63 (±2.54).”
Quote in the paperSubstantial claim - Findingown data
Mean content accuracy score across the ten responses was 4.0 ± 0.0.
“Mean ± SD | 4.0 ± 0.0 | 4.0 ± 0.0 | 3.9 ± 0.1 | 3.98 ± 0.06 | 3.9 ± 0.1 | 3.96 ± 0.04”
Quote in the paperSubstantial claim - Premisecitation
In nursing education, a randomised controlled trial demonstrated that AI-assisted preparation of patient education materials produced clearer and more actionable content compared with student-generated materials.
“In nursing education, a randomised controlled trial demonstrated that AI-assisted preparation of patient education materials produced clearer and more actionable content compared with student-generated materials (16).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [16] Kocak, 2024, The effect of the use of artificial intelligence in the preparation of patient education materials by nursing students: a randomized controlled study, Nurse Educ Pract source
- Interpretationown data
The high expert ratings and narrow score range indicate a potential ceiling effect that limits discrimination between responses.
“The narrow score range and proximity to the scale ceiling indicate a potential ceiling effect, which limits the ability to discriminate between responses of differing quality.”
Quote in the paperNot a substantial claim- Rests on an assumption: The 4-point Likert scale across the five domains is a valid and sensitive instrument for assessing the quality of AI-generated patient education content.
- Interpretationown data
The outputs in their current form are not optimally suited for direct patient use without modification.
“the outputs in their current form are not optimally suited for direct patient use without modification”
Quote in the paperSubstantial claim- Rests on an assumption: A reading level of approximately the sixth grade (FKGL ≤ 6; FRES ≥ 60–70) is the appropriate benchmark for patient education materials.
- Interpretationown data
The readability findings indicate that patients with limited health literacy would likely encounter significant difficulty comprehending the generated content.
“These findings indicate that patients with limited health literacy would likely encounter significant difficulty comprehending the generated content.”
Quote in the paperSubstantial claim- Rests on an assumption: A reading level of approximately the sixth grade (FKGL ≤ 6; FRES ≥ 60–70) is the appropriate benchmark for patient education materials.
- Interpretationown data
ChatGPT-4o may serve as a supplementary tool in diabetes care.
“suggesting potential as a supplementary tool in diabetes care”
Quote in the paperSubstantial claim
- Classified as probably not a substantial claim. 21
- May combine two claims. 8
- The quote only partly states this claim. 2
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationIn musculoskeletal care, ChatGPT-generated guidance for patients with knee osteoarthritis was found to be more comprehensive than clinician-written advice, though more difficult to read.
“In musculoskeletal care, ChatGPT-generated guidance for patients with knee osteoarthritis was found to be more comprehensive than clinician-written advice, though more difficult to read (17).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [17] Singh, 2025, Comparing artificial intelligence–generated and clinician-created personalized self-management guidance for patients with knee osteoarthritis: prospective observational study, J Med Internet Res source
The Discussion’s Principal Findings support the comprehensiveness comparison, but contradict the readability claim: GPT-4 guidance had generally lower reading difficulty than clinician guidance. The Results’ Readability Evaluation likewise report better GPT-4 scores on three of four metrics; clinician-written material had a slight advantage only on the Coleman-Liau Index. Page numbers are not provided in our copy. (LLM, conf=0.90)
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measured
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationIn dentistry, AI-generated materials were associated with improved patient understanding and reduced anxiety.
“In dentistry, AI-generated materials were associated with improved patient understanding and reduced anxiety (20).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [20] Ozan, 2025, Evaluating the impact of AI-generated educational content on patient understanding and anxiety in endodontics and restorative dentistry, BMC Oral Health source
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationA recent systematic review concluded that AI tools consistently improved the readability of patient education materials, while emphasising the continued need for accuracy oversight, particularly for complex topics.
“A recent systematic review concluded that AI tools consistently improved the readability of patient education materials, while emphasising the continued need for accuracy oversight, particularly for complex topics (21).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [21] Raj, 2024, Can artificial intelligence improve patient educational material readability? A systematic review and narrative synthesis, Intern Med J source
The passages returned for this DOI are headed “Functional outcomes of patients with ischaemic stroke with known atrial fibrillation not on therapeutic anticoagulation” (Discussion and Conclusions; pages unavailable), rather than the named AI-readability review. Searches for patient-education readability and accuracy oversight returned only stroke-related text, so our available copy cannot verify the claim. (LLM, conf=0.78)
Experimental — accuracy not yet measured
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationA comparative study of ChatGPT, Gemini, and Copilot for cardiovascular imaging patient education found ChatGPT to perform best on accuracy and completeness.
“A comparative study of ChatGPT, Gemini, and Copilot for cardiovascular imaging patient education found ChatGPT to perform best on accuracy and completeness (18).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [18] Gad, 2025, Evaluating the accuracy and reliability of AI chatbots in patient education on cardiovascular imaging: a comparative study of ChatGPT, Gemini, and Copilot, Egyptian J Radiol Nucl Med source
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationIn oncology, GPT-4-generated prostate cancer education materials scored favourably on understandability and reliability metrics.
“In oncology, GPT-4-generated prostate cancer education materials scored favourably on understandability and reliability metrics (19).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [19] Gibson, 2024, Evaluating the efficacy of ChatGPT as a patient education tool in prostate cancer: multimetric assessment, J Med Internet Res source
The quote only partly states this claim.
Findingown dataChatGPT-4o generated generally accurate and relevant diabetes education content across the ten FAQs, as assessed by five endocrinologists.
“ChatGPT-4o generated generally accurate and relevant diabetes education content”
Quote in the paperSubstantial claim- Rests on an assumption: The 4-point Likert scale across the five domains is a valid and sensitive instrument for assessing the quality of AI-generated patient education content.
- Rests on an assumption: The five board-certified endocrinologists are representative of expert opinion on diabetes patient education.
Probably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataMean clarity score across the ten responses was 3.9 ± 0.1.
“Mean ± SD | 4.0 ± 0.0 | 4.0 ± 0.0 | 3.9 ± 0.1 | 3.98 ± 0.06 | 3.9 ± 0.1 | 3.96 ± 0.04”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataMean relevance score across the ten responses was 3.98 ± 0.06.
“Mean ± SD | 4.0 ± 0.0 | 4.0 ± 0.0 | 3.9 ± 0.1 | 3.98 ± 0.06 | 3.9 ± 0.1 | 3.96 ± 0.04”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Findingown dataRaters expressed concern about linguistic complexity of the responses.
“Raters expressed concern about linguistic complexity, which may impede comprehension among patients with limited health literacy.”
Quote in the paperSubstantial claim- Rests on an assumption: The five board-certified endocrinologists are representative of expert opinion on diabetes patient education.
Classified as probably not a substantial claim.
Findingown dataMean overall quality score across the ten responses was 4.0 ± 0.0.
“Mean ± SD | 4.0 ± 0.0 | 4.0 ± 0.0 | 3.9 ± 0.1 | 3.98 ± 0.06 | 3.9 ± 0.1 | 3.96 ± 0.04”
Quote in the paperSubstantial claimClassified as probably not a substantial claim. May combine two claims.
InterpretationreasoningCollectively, the prior findings suggest promise for LLMs in health communication, while underscoring the necessity of clinician oversight and validation.
“Collectively, these findings suggest promise for LLMs in health communication, while underscoring the necessity of clinician oversight and validation.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
Study design and conductown dataFive board-certified endocrinologists independently evaluated all ten responses.
“Five board-certified endocrinologists (diabetologists) with a minimum of ten years of clinical experience in diabetes management were recruited to evaluate the ChatGPT-4o responses.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
Study design and conductreasoningPrompt engineering and prompt customisation were deliberately excluded to simulate a non-expert user interaction.
“Prompt engineering and prompt customisation were deliberately excluded so that outputs would reflect the experience of a non-expert user interacting with the model in a naturalistic manner”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
Study design and conductreasoningEach prompt was entered as a new, stand-alone chat session with no prior conversation history to avoid contextual carryover.
“each prompt was entered as a new, stand-alone chat session with no prior conversation history”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Study design and conductown dataRaters were not blinded to the AI-generated nature of the content, which may have introduced bias.
“Raters were informed that the content was AI-generated; blinding to the AI-generated nature of the content was not implemented, and this is acknowledged as a potential source of bias.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Study design and conductreasoningAll readability analyses apply exclusively to the English-language outputs generated in this study.
“All readability analyses in this study therefore apply exclusively to the English-language outputs generated by ChatGPT-4o.”
Quote in the paperSubstantial claimMay combine two claims. The quote only partly states this claim.
Interpretationown dataThe high reading-level complexity, small evaluation scope, single-run generation, absence of inter-rater reliability analysis, and English-only assessment limit the generalisability of the findings.
“The high reading-level complexity, small evaluation scope (ten prompts, one model, one session), and English-only assessment limit the generalisability of these findings.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
Study design and conductreasoningWeb-browsing and plugin tools were disabled during the ChatGPT session to ensure responses reflected the model's parametric knowledge only.
“Web-browsing and plugin tools were disabled during the session to ensure that responses reflected the model's parametric knowledge only, without real-time internet retrieval.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductreasoningRatings were conducted independently, without discussion between raters.
“Ratings were conducted independently, without discussion between raters, to ensure independence of assessments.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Study design and conductown dataNo formal adjudication framework was pre-specified; the absence of flagged errors reflects unanimous agreement across all five raters.
“No formal adjudication framework (e.g., majority vote or consensus panel) was pre-specified; raters recorded independent assessments, and the absence of flagged errors reflects unanimous agreement across all five raters.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Study design and conductreasoningNo outputs were discarded or regenerated.
“No outputs were discarded or regenerated.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductreasoningFormal inter-rater reliability statistics were not computed.
“formal inter-rater reliability statistics (e.g., intraclass correlation coefficient or Kendall's W) were not computed”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim. May combine two claims.
InterpretationreasoningThe study's findings are hypothesis-generating and should not be generalised beyond the evaluated prompt set.
“These findings are hypothesis-generating and should not be generalised beyond the evaluated prompt set.”
Quote in the paperSubstantial claim
The paper relies on this without support.
The ten FAQs selected from the Diabetic Association of India and the International Diabetes Federation represent the range of educational needs most relevant to diabetic patients and caregivers.
“The questions were selected to represent the range of educational needs most relevant to diabetic patients and caregivers”
The paper relies on this without support.
The 4-point Likert scale across the five domains is a valid and sensitive instrument for assessing the quality of AI-generated patient education content.
“Each rater independently assessed all ten responses using a 4-point Likert scale (1 = lowest; 4 = highest) across five domains: overall quality, content accuracy, clarity, relevance, and trustworthiness”
The paper relies on this without support.
The operational definition of an 'error' adequately captures clinically significant inaccuracies.
“An “error” was operationally defined as: (a) a factual inaccuracy, (b) a clinically unsafe recommendation, or (c) a significant omission of information critical to patient safety.”
The paper relies on this without support.
A reading level of approximately the sixth grade (FKGL ≤ 6; FRES ≥ 60–70) is the appropriate benchmark for patient education materials.
“Patient education materials are generally recommended to target a reading level of approximately the sixth grade (FKGL ≤ 6; FRES ≥ 60–70).”
The paper relies on this without support.
The five board-certified endocrinologists are representative of expert opinion on diabetes patient education.
“Five board-certified endocrinologists (diabetologists) with a minimum of ten years of clinical experience in diabetes management were recruited to evaluate the ChatGPT-4o responses.”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Abstract and Table 3 give different clarity and trustworthiness SDs
MinorThe abstract reports an SD of 0.20 for both domains, whereas Table 3 reports 0.1 for both.
“3.9 (±0.20) for clarity and trustworthiness.”
“| Mean ± SD | 4.0 ± 0.0 | 4.0 ± 0.0 | 3.9 ± 0.1 | 3.98 ± 0.06 | 3.9 ± 0.1 | 3.96 ± 0.04 |”
Cites
Anjana, R. M., Unnikrishnan, R., Deepa, M., Pradeepa, R., Tandon, N., Das, A. K., Joshi, S., Bajaj, S., Jabbar, P. K., Das, H. K., et al. (2023) · The Lancet Diabetes & Endocrinology · doi:10.1016/s2213-8587(23)00119-5 · Cited by 769 · see citing works
Antar, S. A., Ashour, N. A., Sharaky, M. M., Khattab, M., Ashour, N. A., Zaid, R. T., Roh, E. J., Elkamhawy, A., & Al-Karmalawy, A. A. (2023) · Biomedicine & Pharmacotherapy · doi:10.1016/j.biopha.2023.115734 · Cited by 750 · see citing works
Prompt Engineering as an Important Emerging Skill for Medical Professionals: Tutorial
Meskó, B. (2023) · Journal of Medical Internet Research · doi:10.2196/50638 · Cited by 709 · see citing works
Walker, H. L., Ghani, S., Kuemmerli, C., Nebiker, C. A., Müller, B. P., Raptis, D. A., & Staubli, S. M. (2023) · Journal of Medical Internet Research · doi:10.2196/47479 · Cited by 363 · see citing works
Faiyazuddin, M., Rahman, S. J. Q., Anand, G., Siddiqui, R. K., Mehta, R., Khatib, M. N., Gaidhane, S., Zahiruddin, Q. S., Hussain, A., & Sah, R. (2025) · Health Science Reports · doi:10.1002/hsr2.70312 · Cited by 299 · see citing works
Readability of Patient Education Materials on the American Association for Surgery of Trauma Website
Eltorai, A., Ghanian, S., Adams, C., Born, C., & Daniels, A. (2014) · Archives of Trauma Research · doi:10.5812/atr.18161 · Cited by 265 · see citing works
The potential impact of ChatGPT in clinical and translational medicine
Xue, V. W., Lei, P., & Cho, W. C. S. (2023) · Clinical and Translational Medicine · doi:10.1002/ctm2.1216 · Cited by 198 · cited 1× · see citing works
“ChatGPT was selected for the present study because of its widespread public accessibility, its demonstrated performance across healthcare communication tasks in prior literature, and its broad training corpus encompassing medical and educational material (5, 6).”
Xie, Y., Seth, I., Hunter-Smith, D. J., Rozen, W. M., Ross, R., & Lee, M. (2023) · Aesthetic Plastic Surgery · doi:10.1007/s00266-023-03338-7 · Cited by 198 · see citing works
Empowering Patients: Promoting Patient Education and Health Literacy
Bhattad, P. B., & Pacifico, L. (2022) · Cureus · doi:10.7759/cureus.27336 · Cited by 188 · cited 1× · see citing works
“Accessible and reliable patient education is central to diabetes management, enabling individuals to recognise symptoms, understand disease mechanisms, and adopt strategies to prevent complications (10).”
The Internet as a Source of Health Information and Services
Bujnowska–Fedak, M. M., Waligóra, J., & Mastalerz–Migas, A. (2019) · Advances in experimental medicine and biology · doi:10.1007/5584_2019_396 · Cited by 154 · cited 1× · see citing works
“Nevertheless, the accuracy and reliability of AI-generated medical information require careful, ongoing scrutiny (13–15).”
Gibson, D., Jackson, S., Shanmugasundaram, R., Seth, I., Siu, A., Ahmadi, N., Kam, J., Mehan, N., Thanigasalam, R., Jeffery, N., et al. (2024) · Journal of Medical Internet Research · doi:10.2196/55939 · Cited by 122 · see citing works
Sridharan, K., & Sequeira, R. P. (2024) · BMC Medical Education · doi:10.1186/s12909-024-05365-7 · Cited by 114 · see citing works
Nasra, M., Jaffri, R., Pavlin‐Premrl, D., Kok, H. K., Khabaza, A., Barras, C. D., Slater, L., Yazdabadi, A., Moore, J. M., Russell, J. H., et al. (2024) · Internal Medicine Journal · doi:10.1111/imj.16607 · Cited by 83 · see citing works
The doctor as the second opinion and the internet as the first
Gualtieri, L. N. (2009) · CHI '09 Extended Abstracts on Human Factors in Computing Systems · doi:10.1145/1520340.1520352 · Cited by 81 · see citing works
Abdelhady, A. M., & Davis, C. R. (2023) · Mayo Clinic Proceedings Digital Health · doi:10.1016/j.mcpdig.2023.06.002 · Cited by 55 · see citing works
Yun, J. Y., Kim, D. J., Lee, N., & Kim, E. K. (2023) · International Journal of Medical Informatics · doi:10.1016/j.ijmedinf.2023.105219 · Cited by 45 · see citing works
Islam, S. (2025) · BMC Oral Health · doi:10.1186/s12903-025-06069-0 · Cited by 27 · see citing works
OpenAI’s ChatGPT and Its Role in Plastic Surgery Research
Weidman, A. A., Valentine, L., Chung, K. C., & Lin, S. J. (2023) · Plastic & Reconstructive Surgery · doi:10.1097/prs.0000000000010342 · Cited by 26 · cited 1× · see citing works
“ChatGPT was selected for the present study because of its widespread public accessibility, its demonstrated performance across healthcare communication tasks in prior literature, and its broad training corpus encompassing medical and educational material (5, 6).”
Appropriateness of Artificial Intelligence Chatbots in Diabetic Foot Ulcer Management
Shiraishi, M., Lee, H., Kanayama, K., Moriwaki, Y., & Okazaki, M. (2024) · The International Journal of Lower Extremity Wounds · doi:10.1177/15347346241236811 · Cited by 26 · see citing works
Du, K., Li, A., Zuo, Q.-H., Zhang, C.-Y., Guo, R., Chen, P., Du, W.-S., & Li, S.-M. (2025) · Journal of Medical Internet Research · doi:10.2196/67830 · Cited by 24 · see citing works
Marey, A., Saad, A. M., Tanas, Y., Ghorab, H., Niemierko, J., Backer, H., & Umair, M. (2025) · Egyptian Journal of Radiology and Nuclear Medicine · doi:10.1186/s43055-025-01452-x · Cited by 16 · cited 1× · see citing works
“A comparative study of ChatGPT, Gemini, and Copilot for cardiovascular imaging patient education found ChatGPT to perform best on accuracy and completeness (18).”
Kocak (2024) · doi:10.1016/j.nepr.2024.103933 · Cited by 14 · see citing works
A Quality Assessment of Online Patient Information Regarding Rhinoplasty
Shamil, E., Scenza, G. D., Ghani, S., Fan, K. S., Ragulan, S., Salem, J., Šurda, P., & D'Souza, A. R. (2021) · Facial Plastic Surgery · doi:10.1055/s-0041-1735622 · Cited by 10 · see citing works
Environment, Climate, and Diabetes: An International Topic
Heinemann, L., & Klonoff, D. C. (2025) · Journal of Diabetes Science and Technology · doi:10.1177/19322968251314841 · Cited by 4 · see citing works
Raj (2024) · doi:10.1111/imj.16044 · Cited by 2 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
Song, J., Makhdzir, N., Haron, Z., Qin, C., & Wang, L. (2026) · Frontiers in Public Health · doi:10.3389/fpubh.2026.1902424 · Cited by 1 · see citing works
“Studies using study-specific ratings also reported favorable findings, with Abuzar et al. (26) showing high trustworthiness scores and Hernandez et al. (38) reporting no unreliable responses.”
InfluentialIşıktekin, E., & Hocaoğlu, E. (2026) · Uludağ Üniversitesi Tıp Fakültesi Dergisi · doi:10.32708/uutfd.1949030
Lv, T., Bao, W., Li, S., et al. (2026) · Frontiers in Public Health · doi:10.3389/fpubh.2026.1914167
Health information seeking with generative AI among adults with diabetes or prediabetes
Shinan‐Altman, S., Alon, L., & Levkovich, I. (2026) · Digital Health · doi:10.1177/20552076261487119
- Authors
- Abuzar, M., Rai, S., Pathan, T., Fakki, F., & Syed, S.
- Affiliations
- Anjuman-I-Islam’s Kalsekar Technical Campus School of Pharmacy
- Department of Internal Medicine, Incharge- Diabetes Speciality Centre, MGM Institute of Health Sciences
- Corresponding author
- Abuzar, Mohammad
- Venue
- Frontiers in Digital Health
- Publisher
- Frontiers Media SA
- Published
- 2026-04-13
- Volume
- 8
- Pages
- 1768843-1768843
- ISSN
- 2673-253X (ISSN-L)
- PMC
- PMC13111439
- OpenAlex
- W7154101056
- Semantic Scholar
- ab6511da0f9666eb85248926bb67d8653ff12e2b
- Cited by
- 4
- References
- 23
- Citations by year
- 2026: 4
- FWCI
- 7.57
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Frontiers in Digital Health (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- PubMed Central (submittedVersion, cc-by)
- Topics
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Health Sciences › Health Professions › General Health Professions › Health Literacy and Information Accessibility
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Keywords
- ChatGPT, diabetes mellitus, patient education, artificial intelligence in healthcare, GPT-4o, content accuracy, readability, health literacy, self-management, frequently asked questions, patient support
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-13.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-13.
Cited sources are checked by gpt-6-sol.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
