MediAgent AI: An Agentic AI-Based Healthcare Information Chatbot for Public Health Awareness

Abstract

This paper presents MediAgent AI, a prototype healthcare information chatbot built on an agentic artificial intelligence architecture. Developed on the Botpress conversational AI platform, the system applies natural language processing to interpret user queries spanning symptoms, disease prevention, dietary guidance, and general health topics, drawing exclusively from World Health Organization knowledge repositories to ensure evidence-based responses. Unlike conventional rule-based healthcare chatbots, MediAgent AI employs an agentic design in which the system autonomously navigates query understanding, intent detection, knowledge retrieval, and response synthesis without reliance on fixed decision trees. The chatbot is explicitly scoped as an informational tool: it does not diagnose diseases, prescribe medications, or substitute for professional medical consultation, ensuring full compliance with healthcare AI ethics standards. Empirical evaluation across 200 standardised queries yields an overall weighted accuracy of 87.4%, a mean response latency of 1.8 seconds, and a user satisfaction rating of 4.3 out of 5.0 (n = 20). The paper describes the system architecture, sub-agent design, knowledge integration methodology, and comparative performance analysis, and concludes with a discussion of limitations and prospective research directions. The findings suggest that MediAgent AI represents a cost-effective and scalable complement to public health education initiatives.

Introduction

The text presents MediAgent AI, an agentic AI-powered healthcare chatbot designed to address the growing problem of unreliable and complex online health information. With billions of people relying on the internet for medical guidance, the system aims to bridge gaps in accessibility, accuracy, and usability by providing evidence-based health information sourced from WHO repositories. Unlike diagnostic tools, it is strictly an informational assistant that uses an agentic architecture to handle multi-step reasoning tasks such as query understanding, intent detection, knowledge retrieval, and response generation.

The literature review shows the evolution of healthcare chatbots from rule-based systems to machine learning and large language model-based agents, highlighting improvements in conversational ability but also persistent issues such as hallucination, lack of safety standards, and limited real-world generalization. Across studies, chatbots have been shown to improve patient engagement and health awareness, but many systems still lack robust ethical safeguards, standard evaluation methods, and unified architectures. This motivates the need for a more structured and safety-focused system like MediAgent AI.

The proposed system is built using an agentic AI framework on the Botpress platform, integrating WHO knowledge sources and multi-step reasoning pipelines. Its key contributions include building a functional healthcare chatbot, evaluating performance (accuracy, latency, and user satisfaction), comparing it with existing systems, and addressing ethical considerations.

The methodology involves five stages: requirements analysis, system design, knowledge integration, testing, and deployment. The system is designed to ensure high response accuracy, fast performance, and strict safety constraints, particularly by preventing diagnostic or prescription outputs while still offering reliable health education and guidance.

Conclusion

This study has presented MediAgent AI, an agentic AI-based healthcare information chatbot that demonstrates the substantive potential of advanced conversational AI architectures in democratising access to reliable public health knowledge. Through the integration of WHO-sourced knowledge bases with the Botpress agentic platform and a purpose-designed multi-sub-agent orchestration architecture, the system achieves a balance between informational breadth, response quality, and ethical safety that is not consistently attained by comparable systems in the literature. The empirical evaluation yields three principal findings. First, MediAgent AI achieves an overall response accuracy of 87.4% across five health query categories, with the highest accuracy in symptom information and disease prevention domains. Second, the system delivers 94% of responses within 2.5 seconds, satisfying the operational performance requirements specified at project inception. Third, user satisfaction scores averaging 4.3 out of 5.0 indicate strong user acceptance, tempered by moderate scepticism regarding AI-generated health content trustworthiness — a challenge characteristic of the broader healthcare AI adoption landscape. From a theoretical perspective, this study advances the understanding of agentic AI design in healthcare information contexts by demonstrating that autonomous multi-step reasoning pipelines — traditionally associated with high-complexity AI systems — can be effectively implemented within accessible, low-resource development frameworks. The architectural decomposition of health information retrieval into specialised sub-agents offers a replicable design template for future healthcare chatbot development, particularly in resource-constrained settings where the cost and complexity of bespoke AI development present significant barriers. Several limitations warrant acknowledgement. The usability evaluation was conducted with a relatively small convenience sample (n = 20), limiting the generalisability of satisfaction findings. The knowledge base is currently restricted to English-language WHO content, constraining accessibility for non-English-speaking populations. The performance evaluation relied on standardised queries rather than organic user interactions, potentially underestimating the linguistic variability encountered in real-world deployment. Finally, the system does not currently incorporate user feedback loops that would enable continuous knowledge base refinement based on observed query patterns. Future research directions include: (i) multilingual extension of the knowledge base through integration of WHO regional language repositories; (ii) voice-based interaction capability to serve populations with low digital literacy or physical accessibility constraints; (iii) personalised health recommendation functionality anchored to demographic and health profile parameters; (iv) integration with electronic health record systems and wearable health monitoring devices to enable context-aware health guidance; and (v) deployment of federated learning mechanisms to continuously improve response quality from anonymised interaction data without compromising user privacy. Systematic clinical validation studies involving larger and more demographically diverse user populations represent an important next step towards evidence-based endorsement of MediAgent AI as a component of public health information infrastructure. In conclusion, MediAgent AI exemplifies the manner in which thoughtfully designed, ethically constrained conversational AI systems can serve as valuable adjuncts to public health education efforts — reducing informational inequities, countering the proliferation of medical misinformation, and empowering individuals to make more informed health decisions, while preserving the irreplaceable primacy of professional medical expertise.

References

[1] Aggarwal, A., Tam, C. C., Wu, D., Li, X., & Qiao, S. (2023). Artificial intelligence–based chatbots for promoting health behavioural changes: Systematic review. Journal of Medical Internet Research, 25, e40789. https://doi.org/10.2196/40789 [2] Bhatt, D., Ayyagari, S., & Mishra, A. (2024). A scalable approach to benchmarking the in-conversation differential diagnostic accuracy of a health AI. arXiv preprint arXiv:2412.12538. [3] Botpress. (2024). Conversational AI platform documentation. https://botpress.com [4] Centres for Disease Control and Prevention. (2023). Digital health and disease prevention. https://www.cdc.gov [5] Google. (2023). Advances in conversational AI systems. https://ai.google [6] IBM. (2023). Artificial intelligence in healthcare applications. https://www.ibm.com [7] Jasim, K. M., Malathi, A., Bhardwaj, S., & Aw, E. C. X. (2025). A systematic review of AI-based chatbot usages in healthcare services. Journal of Health Organization and Management. https://doi.org/10.1108/JHOM-12-2023-0376 [8] Kavitha, B. R., & Murthy, C. R. (2019). Chatbot for healthcare system using artificial intelligence. International Journal of Advance Research, Ideas and Innovations in Technology, 5(3). [9] Kim, H. K. (2024). The effects of artificial intelligence chatbots on women\'s health: A systematic review and meta-analysis. Healthcare, 12(5), 534. https://doi.org/10.3390/healthcare12050534 [10] Microsoft. (2023). AI for healthcare transformation. https://www.microsoft.com [11] National Institutes of Health. (2023). AI applications in medical research. https://www.nih.gov [12] Oniani, D., & Wang, Y. (2020). A qualitative evaluation of language models on automatic question-answering for COVID-19. arXiv preprint arXiv:2006.10964. [13] OpenAI. (2024). GPT-based conversational AI systems. https://openai.com [14] Saiteja, K. (2025). Healthcare chatbot using AI. International Journal of Scientific Research in Engineering and Management, 9(6), 1–9. https://doi.org/10.55041/ijsrem49884 [15] Sawad, A. B., Narayan, B., Alnefaie, A., Maqbool, A., Mckie, I., Smith, J., Yuksel, B., Puthal, D., Prasad, M., & Kocaballi, A. B. (2022). A systematic review on healthcare artificial intelligent conversational agents for chronic conditions. Sensors, 22(7), 2625. https://doi.org/10.3390/s22072625 [16] Sri Lalitha, Y., Ganapathi Raju, N. V., Vanimireddy, R. T., Mothe, V. S. K., & Nenavath, A. N. (2023). Conversational AI chatbot for healthcare. E3S Web of Conferences, 391. https://doi.org/10.1051/e3sconf/202339101114 [17] Stanford University. (2023). Artificial intelligence and healthcare innovation. https://www.stanford.edu [18] Wen, B., Norel, R., Liu, J., Stappenbeck, T., Zulkernine, F., & Chen, H. (2024). Leveraging large language models for patient engagement: The power of conversational AI in digital health. arXiv preprint arXiv:2406.13659. [19] World Health Organization. (2022). Ethics and governance of artificial intelligence for health. https://www.who.int [20] World Health Organization. (2023). Fact sheets on diseases and health conditions. https://www.who.int/news-room/fact-sheets [21] World Health Organization. (2023). Global health observatory data repository. https://www.who.int [22] World Health Organization. (2023). Health topics. https://www.who.int/health-topics [23] Yan, N., & Alterovitz, G. (2024). A general-purpose AI avatar in healthcare. arXiv preprint arXiv:2401.12981.

Copyright

Copyright © 2026 Aman Singh, S. Bharat, Shahbaz Alam, Shubham Chauhan, Dr. Arun Khatri. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.