Generative artificial intelligence is increasingly helping people find health information, but it also carries the risk of inaccuracies and unverified recommendations. A new research study by authors from the IBA at the Faculty of Medicine, Masaryk University, evaluated a prototype AI assistant based on educational and popular science content from the National Health Information Portal (NHIP) and showed that linking generative AI with verified health content can significantly enhance the credibility of the information provided.
The study compared the NHIP AI assistant with ChatGPT, Microsoft Copilot, and Google Gemini using a set of 67 health-related questions. The evaluation was conducted by 11 subject matter experts, who primarily assessed the credibility, clarity, user-friendliness, and relevance of the responses. Among all the systems evaluated, the NHIP AI assistant achieved the best result in terms of credibility, while general chatbots were rated higher in terms of clarity, user-friendliness, and relevance.
A significant difference was the linking of responses to verified sources: the NHIP AI assistant provided links to source content in 100% of the evaluated responses, while citations appeared in only 0–3% of the responses from the general chatbots being compared. It also included explicit safety warnings in 85% of responses and, in 97% of cases, recommended additional relevant sources to users for more detailed information.
Another important aspect of the design is the system’s limitation to verified NHIP content and its ability to withhold a response in situations where a query exceeds its knowledge base or requires individualized medical advice. This approach can be particularly important in the healthcare setting, where, in addition to the comprehensibility of the response, its traceability, transparency, and security are also essential. From the perspective of health literacy, the design and implementation of the AI assistant represent a completely new way to provide users with professionally verified information in a natural conversational format while simultaneously referring them to additional verified NHIP materials. The study also shows that when using artificial intelligence in healthcare, it is necessary to strike a balance between credibility and user convenience. The results therefore provide an important foundation for the further development of AI tools within public health information portals; however, they do not prove that the mere use of an AI assistant leads to increased health literacy or improved health outcomes. The authors emphasize that these effects will need to be verified in follow-up research directly involving real users.
Researchers from the IBA at the Faculty of Medicine, Masaryk University, achieved success at the 11th International Digital Public Health Conference 2026 in Barcelona. Graduate student Hlib Aleksandrenko and his advisor, Martin Komenda, received an award for the best scientific abstract.
Professor Ladislav Dušek received the annual award from the Association of Social Service Providers, primarily for his contribution to the ELFis (End of Life Care Information System) project—an information system designed to help optimize social, health, and palliative care services.
