Table 2.

Challenges encountered and solutions explored during the development and implementation of chatbot microintervention.

Challenge			Solutions explored
Development
	Limits to how many consecutive messages or media files could be sent	Integrating more user-response questions Splitting up message clusters into smaller message groups Condensing and simplifying text messages
	Limits to the formats and dimensions of media files were displayed on social media platforms	Choosing more square-like images as opposed to very wide/rectangular images Converting all images into PNG or JPEG format
	Limits to the length and number of preset answer choices	Choosing “quick replies” (which allows for more answer choices) over “buttons” for user-response questions Condensing and simplifying preset answer choices (limiting to at most 6 answer choices) Integrating more open-response questions
	Inability to provide preset answer choices on Desktop Instagram Messenger and Facebook Lite	Providing a notice in the chatbot welcome message on these specific incompatibilities and links for alternative versions of Facebook or Instagram to converse with the chatbot
	Inability to return midway through a lesson after a problem was resolved	Creating checkpoints in the middle of lessons with keywords that, when texted, would allow participants to return to that checkpoint
	Inability to manually launch a chatbot flow without going through the chatbot platform	Along with connecting flows to preset answer choices, each flow was set up to also initiate if a specific keyword was texted (eg, “get started” and “return”)
Implementation
	Chatbot stopped speaking due to server overburden	Splitting lessons into two separate, sequential chatbot flows Not forcing the chatbot to wait indefinitely for each user response without mechanisms to continue to the next message Reducing the size and resolution of media files Informing participants of the potential of chatbot disruptions, providing multiple self-administered troubleshooting strategies, and regularly monitoring conversations to provide manual guidance if needed
	Chatbot stopped speaking due to sporadic social media or chatbot platform glitches	Checking updates from the chatbot company website and emails on temporary systemwide outages or glitches Joining a Facebook group of other users of the chatbot company to check on system outages or glitches and learn strategies to improve the chatbot experience Informing participants of the potential of chatbot disruptions, providing multiple self-administered troubleshooting strategies, and regularly monitoring conversations to provide manual guidance if needed
	Chatbot stopped speaking due to participant not selecting a preset answer	Enforcing response validation for “quick reply” questions, which provided a reminder to choose from preset answer choices Informing participants of the potential of chatbot disruptions, providing multiple self-administered troubleshooting strategies, and regularly monitoring conversations to provide manual guidance if needed
	Participant is unable to start a conversation with the chatbot using the link provided.	Providing backup instructions for manually locating the chatbot’s Facebook or Instagram page and initiating a chat Regularly monitoring conversations for chatbot initiation failure (eg, “get started” was not texted) and providing manual guidance if needed
	Participant speaking with the chatbot from multiple accounts	Ensuring participant self-reported their social media handle in both the pretest and posttest surveys Messaging participant and using time-stamped message and profile information to confirm discrepancies
	Participant accidentally starts a flow by entering a flow-initiating keyword	Ensuring flow-initiating keywords were both easy to remember and unlikely to be used otherwise during chatbot conversations (eg, “grapes1” instead of “grapes”)