Consumer research on text-based chatbots and voicebots: positive, negative and comparative effects
| Study | Theoretical approach | Methodology | Key findings |
|---|---|---|---|
| Chatbots – positive effects | |||
| Cheng et al. (2022) | Stimulus–organism–response (SOR) | Survey | In a service failure setting, empathy and friendliness of a chatbot have a positive effect on trust, which, in turn, has a positive effect on customer reliance. One moderator is task complexity: when a task is more complex, the positive effect of friendliness on trust is weaker |
| Liu et al. (2023) | Emotions expression | Experimental | The use of funny emojis by a chatbot has a positive effect on customer attitudes after a service failure. The effect is mediated by the perceived intelligence of the chatbot and is stronger for people with a growth mindset (incremental theory) |
| Agnihotri and Bhattacharya (2024) | Computers as social actors, information asymmetry | Mixed | Perceived empathy and anthropomorphism of the chatbot have a negative impact on negative word of mouth (nWOM) and a positive impact on customer forgiveness after a service failure. In contrast, privacy concerns when interacting with the chatbot have a positive impact on nWOM and a negative impact on customer forgiveness |
| Cai et al. (2024) | Mind perception | Experimental | A social-oriented (instead of task-oriented) communication style reduces negative emotions triggered by service failures, which in turn improves satisfaction with the service agent and company |
| Chatbots – negative effects | |||
| Mozafari et al. (2020) | Attribution theory | Experimental | When customers are aware that they are interacting with a chatbot, trust decreases, particularly when the conversation revolves around a critical (vs. routine) service issue. Reduced trust leads to customer churn and negatively affects loyalty |
| Crolic et al. (2021) | Emotional appraisal | Field and experimental | Customers entering an interaction with a chatbot in an angry state experience greater dissatisfaction with the service and company if they perceive the chatbot as anthropomorphized |
| Huang and Dootson (2022) | Emotional coping | Experimental | After a service failure, a chatbot informing a customer of the possibility of interacting with a human at a later stage of their interaction can trigger aggressive responses from the customer. This effect is particularly stronger for those who perceive low customer participation in the interaction |
| Voicebots – positive effects | |||
| Cuadra et al. (2021) | Conversational repair | Experimental | A voicebot capacity to self-correct (after, e.g. giving incorrect information) improves customers’ assessment of the robot and interaction. This feature was more strongly valued by participants who had had recent experiences with errors |
| Lv et al. (2021) | Appraisal theory, performance expectancy | Experimental | A voicebot using a cute tone of voice leads to a higher tolerance for service failures. The effect is mediated by higher tenderness and lower performance expectancy and moderated by time pressure and failure severity |
| Huang and Sénécal (2023) | Computers as social actors | Experimental | Warm verbal content used by a voicebot increases repatronage intention after a service failure. This positive effect is mediated by decreased frustration |
| Voicebots – negative effects | |||
| Luo et al. (2019) | Negative biases towards Machines | Field | The disclosure of voicebots’ identities as machines led to a drastic decrease in their effectiveness in outbound sales calls, as reflected in purchases. The effect was driven by a judgment of voicebots as less empathetic and less knowledgeable |
| Sun et al. (2022) | Cognitive load | Survey | Negative technical features (e.g. lack of perceived control resulting from cumbersome design and response delays) affect customer satisfaction and usage continuance intentions of voicebots (specifically, personal assistants) in the context of service failures. The effects are mediated by cognitive load |
| Li et al. (2023) | Self-attribution and performance expectancy | Field | A service failure by a voicebot increases the likelihood of customer complaint behavior. The effect is mediated by negative emotion |
| Chatbots versus voicebots – comparative effects | |||
| Rzepka et al. (2022) | Technology fit theory | Experimental | Voicebots lead to higher satisfaction than chatbots in goal-directed search tasks. This effect is driven by greater perceived efficiency, lower cognitive effort and higher enjoyment |
| Zierau et al. (2023) | Experimental | Voicebots create more flow-like experiences than text-based chatbots, leading to better service evaluations and outcomes. These benefits diminish when interactions are semantically disfluent or involve too many conversational turns | |
| Rohit et al. (2024) | Media richness | Experimental | In a retail setting, voicebots are more effective than chatbots at enhancing cognitive and affective engagement, especially for experiential products. The effect is moderated by localization or product customization according to location |
| Schindler et al. (2024) | Matching, congruence effects | Experimental | Text-based interactions (vs voice-based interactions) with a conversional agent encourage a reason-based (vs emotion-based) focus, leading to a greater preference for utilitarian (vs hedonic) products. This matching effect is mediated by processing fluency and is stronger for low-equity brands |
| This study | Emotion regulation | Experimental | In a service failure setting, chatbots lead to more positive customer reviews via increased perceived relief. The effect is stronger for low-severity failures and when outcomes are worse than expected |
| Study | Theoretical approach | Methodology | Key findings |
|---|---|---|---|
| Stimulus–organism–response ( | Survey | In a service failure setting, empathy and friendliness of a chatbot have a positive effect on trust, which, in turn, has a positive effect on customer reliance. One moderator is task complexity: when a task is more complex, the positive effect of friendliness on trust is weaker | |
| Emotions expression | Experimental | The use of funny emojis by a chatbot has a positive effect on customer attitudes after a service failure. The effect is mediated by the perceived intelligence of the chatbot and is stronger for people with a growth mindset (incremental theory) | |
| Computers as social actors, information asymmetry | Mixed | Perceived empathy and anthropomorphism of the chatbot have a negative impact on negative word of mouth (nWOM) and a positive impact on customer forgiveness after a service failure. In contrast, privacy concerns when interacting with the chatbot have a positive impact on nWOM and a negative impact on customer forgiveness | |
| Mind perception | Experimental | A social-oriented (instead of task-oriented) communication style reduces negative emotions triggered by service failures, which in turn improves satisfaction with the service agent and company | |
| Attribution theory | Experimental | When customers are aware that they are interacting with a chatbot, trust decreases, particularly when the conversation revolves around a critical (vs. routine) service issue. Reduced trust leads to customer churn and negatively affects loyalty | |
| Emotional appraisal | Field and experimental | Customers entering an interaction with a chatbot in an angry state experience greater dissatisfaction with the service and company if they perceive the chatbot as anthropomorphized | |
| Emotional coping | Experimental | After a service failure, a chatbot informing a customer of the possibility of interacting with a human at a later stage of their interaction can trigger aggressive responses from the customer. This effect is particularly stronger for those who perceive low customer participation in the interaction | |
| Voicebots – positive effects | |||
| Conversational repair | Experimental | A voicebot capacity to self-correct (after, e.g. giving incorrect information) improves customers’ assessment of the robot and interaction. This feature was more strongly valued by participants who had had recent experiences with errors | |
| Appraisal theory, performance expectancy | Experimental | A voicebot using a cute tone of voice leads to a higher tolerance for service failures. The effect is mediated by higher tenderness and lower performance expectancy and moderated by time pressure and failure severity | |
| Computers as social actors | Experimental | Warm verbal content used by a voicebot increases repatronage intention after a service failure. This positive effect is mediated by decreased frustration | |
| Negative biases towards Machines | Field | The disclosure of voicebots’ identities as machines led to a drastic decrease in their effectiveness in outbound sales calls, as reflected in purchases. The effect was driven by a judgment of voicebots as less empathetic and less knowledgeable | |
| Cognitive load | Survey | Negative technical features (e.g. lack of perceived control resulting from cumbersome design and response delays) affect customer satisfaction and usage continuance intentions of voicebots (specifically, personal assistants) in the context of service failures. The effects are mediated by cognitive load | |
| Self-attribution and performance expectancy | Field | A service failure by a voicebot increases the likelihood of customer complaint behavior. The effect is mediated by negative emotion | |
| Technology fit theory | Experimental | Voicebots lead to higher satisfaction than chatbots in goal-directed search tasks. This effect is driven by greater perceived efficiency, lower cognitive effort and higher enjoyment | |
| Experimental | Voicebots create more flow-like experiences than text-based chatbots, leading to better service evaluations and outcomes. These benefits diminish when interactions are semantically disfluent or involve too many conversational turns | ||
| Media richness | Experimental | In a retail setting, voicebots are more effective than chatbots at enhancing cognitive and affective engagement, especially for experiential products. The effect is moderated by localization or product customization according to location | |
| Matching, congruence effects | Experimental | Text-based interactions (vs voice-based interactions) with a conversional agent encourage a reason-based (vs emotion-based) focus, leading to a greater preference for utilitarian (vs hedonic) products. This matching effect is mediated by processing fluency and is stronger for low-equity brands | |
| This study | Emotion regulation | Experimental | In a service failure setting, chatbots lead to more positive customer reviews via increased perceived relief. The effect is stronger for low-severity failures and when outcomes are worse than expected |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.