Speaking vs. Listening? Balance Conversation Attributes of Voice Assistants for Better Voice Marketing

• The human-like conversation prompts a new marketing trend of voice marketing. • Relationship among conversation attributes, trust, and voice shopping intention are examined. • Congruency analytics such as polynomial modeling and response surface analysis are employed. • The perceptive incongruency can undermine consumer trust via decreased social presence. Artificial Intelligence (AI) is shaping marketing in an unprecedented way. Empowered by AI, voice assistants are increasingly capable of speaking and listening like humans, offering a great opportunity for a new marketing approach - voice marketing. This research examines how conversation attributes of voice assistants determine consumer trust and intention to engage in voice shopping. Using a sequential mixed-method design, three studies consistently show that consumers perceive the speaking attribute of voice assistants as more human-like than the listening attribute. We find that such incongruency between the two conversation attributes can undermine consumers' trust in voice assistants, leading to reduced willingness to accept product recommendations from voice assistants and shop via voice assistants, which would hamper the development of voice marketing. Accordingly, this research suggests that AI giants with strong technological strength and capital support should distribute more resources to advance the underlying technologies enabling human-like listening (e.g., natural language understanding and voice recognition). But for AI startups with limited financing ability and technical talents, they may consider appropriately reducing investments in the underlying technologies enabling human-like speaking (e.g., natural language generation and voice synthesis) to enhance the congruency level between the conversation attributes of voice assistants.