Table 1. Summary of inclusion and exclusion criteria^a.

Variable	Inclusion criteria	Exclusion criteria
Population	Human participants of any demographics and health conditions.	Studies without human participants (eg, simulations).
Intervention	Studies that manipulated the communication competence of a nonembodied and unconstrained CAb in conversations about health-related topics.	CAs with physical or virtual embodiments (eg, embodied CAs, robots, or virtual avatars). Constrained CAs that restrict user interaction to predefined options. Studies applied CAs in nonhealth topics (eg, marketing and education). Studies manipulated CAs’ features unrelated to communication competence, such as appearance, identity, and gender.
Comparison	Studies with a control group using CAs with no or low levels of communication competence, or without a comparison (eg, pre-post design).	Studies that use standard care or other technologies that did not involve CAs.
Outcome	Measure at least 1 outcome related to: Evaluations of CA Use of CA Psychological outcomes Health outcomes	Studies that report only technical performances of developed methods or algorithms.
Study design	Experimental studies.	Reviews, qualitative studies, survey studies, or other nonempirical studies.

^aIn addition to the PICOS criteria listed in the table, we had an additional eligibility criterion that studies needed to be complete research papers written in English and published in peer-reviewed journals, conference proceedings, dissertations, or theses. For meta-analysis, included studies needed to have a control group using the corresponding conversational agents with no or low levels of communication competence and reported sufficient information to calculate effect sizes.

^bCA: conversational agent.