AI-Driven Medical Misinformation Detection and Health Guidance System

This paper presented a web application that brings health claim verification, symptom guidance, medical report reading and personal record keeping into one place. A TF-IDF and logistic regression classifier gives the first label, Sentence Transformers, FAISS, Tavily and the Google Fact Check API supply evidence, and a validator that is independent of the language model scores the result. OCR and PDF parsing turn reports into searchable records. The system can say UNCERTAIN, shows its sources, and never presents a possible condition as a diagnosis. It is meant to support readers and not to replace clinical advice. Much remains to be done. The first task is a proper evaluation, with a larger and more varied claim set and measured accuracy for the classifier, the retrieval step and the validator. We also want to understand what drives the validation score, widen the set of trusted evidence sources, and improve retrieval indexing for complex questions. Report parsing needs the most care: more report formats, better handling of poor scans, and checks that catch values like the ones in Fig. 7. Personal records could gain trend views over longer periods. The LoRA training pipeline can be explored further while keeping the independent validator, and additional validation rules can be added for conflicting evidence.

1.   V. Suarez-Lledo and J. Alvarez-Galvez, “Prevalence of health misinformation on social media: Systematic review,” J. Med. Internet Res., vol. 23, no. 1, Art. no. e17187, 2021, doi: 10.2196/17187. 2.  Y. J. Li, J. J. Marga, C. M. K. Cheung, X. L. Shen, and M. K. O. Lee, “Health misinformation on social media: A systematic literature review and future research directions,” AIS Trans. Hum.-Comput. Interact., vol. 14, no. 2, pp. 116–149, 2022, doi: 10.17705/1thci.00164. 3.   H. R. Saeidnia, S. Jahani, N. Ghiasi, and H. Keshavarz, “Generative AI and health misinformation: Production, propagation, and mitigation—A systematic review,” BMC Public Health, vol. 26, no. 1, p. 693, 2026, doi: 10.1186/s12889-025- 26148-9. 4.  Y. Francis, F. J. Lantry, M. Echeverry, and M. Siddiq, “Effective countermeasures to health misinformation and disinformation for global health engagement practitioners: A rapid review,” Mil. Med., vol. 190, no. Suppl_2, pp. 478–482, 2025, doi: 10.1093/milmed/usaf237. 5.   S. van der Linden, “Misinformation: Susceptibility, spread, and interventions to immunize the public,” Nat. Med., vol. 28, pp. 460–467, 2022, doi: 10.1038/s41591-022-01713-6. 6.  Q. Wei, Z. Yao, Y. Cui, B. Wei, Z. Jin, and X. Xu, “Evaluation of ChatGPT-generated medical responses: A systematic review and meta-analysis,” J. Biomed. Inform., vol. 151, Art. no. 104620, 2024, doi: 10.1016/j.jbi.2024.104620. 7.  T. Han et al., “Medical large language models are susceptible to targeted misinformation attacks,” npj Digit. Med., vol. 7, Art. no. 288, 2024, doi: 10.1038/s41746-024-01282-7. 8.  I. Skafle, A. Nordahl-Hansen, D. S. Quintana, R. Wynn, and E. Gabarron, “Misinformation about COVID-19 vaccines on social media: Rapid review,” J. Med. Internet Res., vol. 24, no. 8, Art. no. e37367, 2022, doi: 10.2196/37367. 9.   X. Feng, J. Luo, Y. Yang, D. El Baz, and L. Shi, “Health misinformation detection: Approaches, challenges and opportunities,” INQUIRY, vol. 62, 2025, doi: 10.1177/00469580251384784. 10.    V. Papanikou, P. Papadakos, T. Karamanidou, T. G. Stavropoulos, E. Pitoura, and P. Tsaparas, “Health misinformation in  social  networks:  A  survey  of  information  technology approaches,” Future Internet, vol. 17, no. 3, Art. no. 129, 2025, doi: 10.3390/fi17030129.