Digital sales consultation through AI-driven dialogue systems is increasingly shaping interactions between companies and customers. A central trend in this context is Conversational Commerce, where AI-based systems provide personalized product recommendations and advice in online shopping. With advancing technological development, these systems are becoming ever more human-like – from text-based chatbots to voice assistants that communicate naturally, and finally to AI avatars whose appearance and behavior are sometimes barely distinguishable from real people.
But how does this increasing human-likeness affect user perception and acceptance? This question was at the heart of the bachelor’s thesis by Lara-Maria Kraft, which examined the influence of different AI representations – chatbot, voice assistant, and AI avatar – on users’ willingness to use such systems within the framework of Conversational Commerce.
Research Objective
The aim of the study was to examine whether a higher degree of human-likeness – that is, moving from a text-based chatbot to a voice assistant and ultimately to an AI avatar – increases users’ willingness to use these systems. In addition, the study analyzed whether this relationship is mediated by perceived human-likeness (anthropomorphism) and moderated by potential discomfort as described by the Uncanny Valley effect.
The findings aim to help companies in digital commerce design AI-based tools that foster high user acceptance and thereby contribute to long-term commercial success in Conversational Commerce.
Method
The research was conducted as an experimental online study with 176 participants. Subjects were randomly assigned to one of three interaction formats differing in their level of human-likeness:
- Chatbot – text-based interaction
- Voice Assistant – auditory communication
- AI Avatar – audiovisual representation
All groups were shown an identical video scenario simulating an AI-based product consultation in an online store. Using the example of choosing a running shoe, participants were guided through the consultation process by a virtual assistant. Afterwards, they evaluated their perception of anthropomorphism (the attribution of human qualities), their feeling of eeriness in line with the Uncanny Valley concept, and their willingness to use the presented AI system.
Key Findings
The results revealed a generally high willingness to use AI-based consultation systems in the context of Conversational Commerce – regardless of their specific form. Contrary to expectations, the degree of human-likeness had no significant impact on usage intention; no differences were found between chatbot, voice assistant, and AI avatar conditions.
Likewise, neither the mediating role of anthropomorphism nor the moderating effect of eeriness (as explained by the Uncanny Valley effect) were empirically supported. However, a clear positive relationship emerged between perceived anthropomorphism and willingness to use: the more human-like a system was perceived to be, the greater participants’ readiness to interact with it. Thus, anthropomorphism, understood as the subjective attribution of human qualities, proved to be a central factor influencing user perception and acceptance of AI tools – independent of their visual or technical design.
Practical Implications
Based on the empirical results, several practical recommendations can be derived for the design of AI-driven systems in Conversational Commerce. The findings show that users generally display a high openness to interaction with AI systems, regardless of their external form. This overall acceptance can serve as a strategic impulse for companies to systematically integrate AI-based consultation tools into digital shopping processes.
Especially in e-commerce, where personal consultation is often lacking, such systems can enhance the customer experience by supporting users throughout the information, selection, and decision-making phases. However, companies should not primarily invest in costly or hyper-realistic avatars. Instead, the focus should be on creating a natural, empathetic, and user-oriented communication style that builds trust and conveys social closeness – factors that are far more decisive for acceptance than purely visual human-likeness.
Conclusion
This study explored the influence of increasing human-likeness in AI-based systems on user acceptance within the field of Conversational Commerce. Contrary to expectations, the degree of human-likeness – whether chatbot, voice assistant, or AI avatar – did not significantly affect users’ willingness to engage. What truly matters is how human-like the system is perceived.
Perceived anthropomorphism emerged as a key factor: the more human qualities users attributed to a system, the greater their willingness to use it – independent of its actual design. However, this perception is not necessarily triggered by more human-like visual appearance alone.
Future research should therefore focus on identifying the specific factors and interaction elements that foster anthropomorphism in AI systems – and explore how these can be strategically employed to further enhance user acceptance and experience.