Applications of artificial intelligence-based conversational agents in healthcare: A systematic umbrella review.
Artificial intelligence-based conversational agents (AI-based CAs) have emerged as essential tools for communication and service delivery in the healthcare industry. However, global synthesis regarding the current applications and effectiveness of this technology remains scarce.
This study aims to provide a comprehensive overview of the current applications of AI-based CAs in healthcare and associated health-related outcomes.
A systematic search of six databases and additional sources was conducted for studies published from January 1, 2000 to August 20, 2025. Inclusion criteria were as follows: a) articles must be peer-reviewed literature reviews conducted systematically with or without quantitative analysis, b) at least 70% of primary studies included in the review article reported applications of AI-based CAs, c) included primary studies were conducted in healthcare context, d) review articles must report health-related outcomes, and e) full-text must be available in English.
In total, 44 review articles were included. AI-based CAs were implemented in a wide range of medical functions, including supporting clinical decision making, supporting mental health, providing educational content, etc. Most articles studied text-based CAs that utilize Deep Learning methods. The most common mode of deployment is stand-alone or embedded in currently available applications. Regarding effectiveness, only articles discussing the effectiveness of AI-based CAs to assist addiction such as smoking and substance use reported all positive health-related outcomes. Some domains, such as clinical decision support and mental health support, receive more attention than others, while understudied areas such as promoting healthy lifestyle offer potential results.
It remains challenging to draw conclusions regarding the overall effectiveness of current AI-based CAs applications in healthcare domain, calling for transparency and standardization in CAs development practice. Moreover, some healthcare domains are more heavily studied than the others, causing an imbalance. Scholarly works on the understudied areas are critical to ensure successful implementation of AI-based CAs.
This study aims to provide a comprehensive overview of the current applications of AI-based CAs in healthcare and associated health-related outcomes.
A systematic search of six databases and additional sources was conducted for studies published from January 1, 2000 to August 20, 2025. Inclusion criteria were as follows: a) articles must be peer-reviewed literature reviews conducted systematically with or without quantitative analysis, b) at least 70% of primary studies included in the review article reported applications of AI-based CAs, c) included primary studies were conducted in healthcare context, d) review articles must report health-related outcomes, and e) full-text must be available in English.
In total, 44 review articles were included. AI-based CAs were implemented in a wide range of medical functions, including supporting clinical decision making, supporting mental health, providing educational content, etc. Most articles studied text-based CAs that utilize Deep Learning methods. The most common mode of deployment is stand-alone or embedded in currently available applications. Regarding effectiveness, only articles discussing the effectiveness of AI-based CAs to assist addiction such as smoking and substance use reported all positive health-related outcomes. Some domains, such as clinical decision support and mental health support, receive more attention than others, while understudied areas such as promoting healthy lifestyle offer potential results.
It remains challenging to draw conclusions regarding the overall effectiveness of current AI-based CAs applications in healthcare domain, calling for transparency and standardization in CAs development practice. Moreover, some healthcare domains are more heavily studied than the others, causing an imbalance. Scholarly works on the understudied areas are critical to ensure successful implementation of AI-based CAs.