Master-Servant-Partner interaction styles with artificial intelligence chatbots and perceived value co-creation

Among artificial intelligence (AI) applications, AI-enabled chatbots have emerged as a key technology (Shin et al., 2023), fostering value co-creation and enhancing customer experiences in online shopping (Jiang et al., 2022). Chatbot technology is changing and evolving rapidly; as a result, organizations are increasingly deploying AI-powered chatbots to engage customers and enhance service delivery (Brandtzaeg and Følstad, 2018; Sheng et al., 2025). Many recent studies, therefore, have begun focusing on human-AI chatbot interactions (Sheng et al., 2025) with the aim of maximizing customer engagement and satisfaction during interactions with AI chatbots (Paul et al., 2023).

Current human-AI chatbot interaction research predominantly focuses on anthropomorphic design and communication styles. Anthropomorphic studies explore how human-like features in chatbots affect user perceptions. Tsai et al. (2021) investigated how anthropomorphic profiles enhance consumer engagement, while Jin and Youn (2023) examined anthropomorphism's interplay with social presence and continuance intention. Prior research has explored the influence of AI chatbot's verbal anthropomorphic cues on consumer compliance with requests, the impact of chatbot appearance on user trust and chatbot's human-like emotional support on user experience (Bae et al., 2023; Meng and Dai, 2021; Adam et al., 2021). AI chatbots' communication style research examines how chatbots' language and communication shape engagement, perceptions and satisfaction (Choi, 2025; Li and Wang, 2023). Communication-style studies address chatbots' social versus task-oriented styles during service delivery (Cai et al., 2024; Xu et al., 2022), conversational warmth affecting self-disclosure (Choi and Zhou, 2023), cognitive processing impacts (Xiao et al., 2025), credibility perceptions (Jin et al., 2024) and usage intention (Sheng et al., 2025). Collectively, these existing studies to date have largely explored the nuanced ways chatbots' anthropomorphic design and their communication styles shape human-AI chatbot interactions. However, despite considerable research on functional aspects guiding chatbot design to improve customer experience, AI chatbots still fail to fully meet user expectations (Brandtzaeg and Følstad, 2018; Zhang et al., 2024).

One critical reason for chatbots' limited success may be that current human-AI chatbot design as well as research disproportionately focuses on technical and operational aspects (anthropomorphic, communicative and functional attributes) (Pham and Duong, 2025). Existing chatbot-centric studies contribute to understanding efficiency but overlook how consumers actively shape interactions through their engagement styles, expectations, values and power orientations. Previous research largely conceptualizes users as passive recipients rather than active participants. Therefore, human-AI chatbot research must shift from design-dominant, chatbot-centric paradigms toward user and/or consumer-centric approaches.

Although evolving chatbot interfaces are reconfiguring online shopping service encounters with their immense capability to understand human reactions and exhibit human-like behaviours (Shin et al., 2023; Pelau et al., 2021), consumers as agents are equally capable of actively shaping interactions and engaging with AI chatbots to co-create value (Lalicic and Weismayer, 2021; Aw et al., 2022; Jiang et al., 2022). Therefore, we support Paul et al. (2023), Hoffman and Novak (2018) and Lalicic and Weismayer (2021) that research on human-AI chatbot interactions must focus on how consumer interaction styles with AI chatbots shape value co-creation in online shopping environments, a gap that has not been sufficiently explored in the current literature. Online shopping environments offer an ideal setting to study such human-AI chatbot interactions, where consumers search, seek information and make purchasing decisions with chatbot assistance, enabling real-time, dynamic value co-creation.

Moreover, there is limited evidence on how interaction styles with AI chatbots influence consumers' value co-creation experiences. While Ekinci and Dawes (2009) emphasized that “interaction quality” mediates the relationship between consumers and smart objects, we extend this by proposing that understanding of various interaction styles (master–servant–partner) and their impact on interaction quality (IQ) offers deeper insights into the understanding of value co-creation in online shopping.

As technology advances, rising consumer expectations can influence interaction styles and outcomes. Lalicic and Weismayer (2021) emphasize that expectations of AI shape interactions with AI-based services. For example, consumers perceiving smart objects as masters may place excessive demands on chatbots when expectations are high, potentially degrading IQ and value co-creation. This aligns with Paul et al.’s (2023) call for research on how preferences and expectations enhance experiences. Accordingly, this study examines how consumer AI expectations moderate the relationship between interaction styles, IQ and perceived value co-creation (PVCC).

Based on these literature gaps, this study addresses three research questions:

1. How do consumer interaction styles with AI chatbots (master, servant or partner) influence their PVCC in online shopping contexts?

2. To what extent, if any, does IQ mediate the relationship between consumer interaction styles and PVCC?

3. How do consumer expectations of AI technology moderate the relationship between interaction styles, IQ and value co-creation?

This study is novel in three ways. First, it shifts AI research focus from a chatbot design-centric to a user-centric perspective, examining consumer interaction styles with AI chatbots in online shopping contexts. Second, moving beyond utilitarian efficiency, it reveals deeper relational schemas such as master, servant and partner styles, using assemblage theory to explain value co-creation. Third, by focusing on user engagement and interactions, the study recommends developers consider user diversity and relational dynamics to enhance engagement beyond technical functionality.

Value co-creation involves multiple actors integrating resources to create value (Vargo and Lusch, 2016). Human actors and AI-enabled chatbots are seen as assemblages of skills and knowledge capable of co-creating value (Chandler and Lusch, 2015). When consumers interact with AI chatbots, varying levels of assemblage occur, leading to value co-creation. During these interactions, AI chatbots influence consumer expectations, perceptions, preferences and actions (Jaakkola and Alexander, 2014). Building on the assemblage process of value co-creation (Aw et al., 2022; Payne et al., 2008), the study argues that PVCC emerges from IQ shaped by the interaction styles of consumers (master, servant and partner) and their expectations of AI chatbots.

Assemblage theory examines socio-material relationships forming collective entities or assemblages (Deleuze et al., 1987; Hoffman and Novak, 2018), offering a framework to study consumer–smart object interactions. It highlights collaborative, ontologically equivalent human–smart object relationships and emergent outcomes, enabling nuanced understanding of these complex interactions (DeLanda, 2016; Harman, 2017; Hoffman and Novak, 2018). Assemblage theory provides a theoretical underpinning by acknowledging subjective, bidirectional consumer–smart object relationships where interaction styles shape relationship quality and value co-creation (Hoffman and Novak, 2018). Drawing on the assemblage theory, Hoffman and Novak (2018) suggest that consumers' relationships with smart objects are reflected in their interaction styles such as master, servant or partner. These styles reflect consumers' tendencies to act as agentic (independent) or communal (relational) actors to achieve their goals (Barbopoulos and Johansson, 2016; Hoffman and Novak, 2018; Ruan and Mezei, 2022). While prior research identifies these interaction styles with smart technologies (Schweitzer et al., 2019), limited attention has been paid to understanding the extent to which these interaction styles (master-servant-partner) influence value co-creation with AI chatbots in online shopping contexts (De Santis, 2019; Zhang et al., 2024).

Consumer interaction styles with smart objects such as voice assistants or AI chatbots refer to the different ways consumers interact, treat and engage with chatbots. The consumer-centric view of human-AI assemblages identifies three interaction styles: master, servant and partner (Hoffman and Novak, 2018). According to Barbopoulos and Johansson (2016), Hoffman and Novak (2018) and Ruan and Mezei (2022), consumer interaction styles with AI chatbots such as master-servant-partner stem from consumers' agentic (independent), passive (dependent) or communal (relational) tendencies.

Schweitzer et al. (2019) describe agentic roles as dominant and status-driven, where technology serves users' commands (Sun et al., 2023). Tschopp et al. (2023), applying Fiske's (1992) relational model theory to analyse human-AI dynamics, emphasized authority ranking in digital assistant interactions, reflecting hierarchical relationships where humans command obedience. We refer to this in our study as servant interaction style, wherein consumers are masters and AI chatbots are servants. For instance, when purchasing a portable speaker online, a consumer using an AI chatbot to locate the item may treat it as a servant engaging in a directive, task-focused exchange that quickly fulfils specific requests. Hoffman and Novak (2018) observed that some consumers perceived themselves as dependent on AI chatbots, often following their suggestions and instructions, a phenomenon Mick and Fournier (1998) referred to as the freedom/enslavement paradox. In our study, we refer to this as the master interaction style (consumer – servant; object – master), wherein consumers might approach chatbots with command-and-control expectations. For example, an online customer unsure about which speaker to buy may rely entirely on the AI chatbot, treating it as a master that takes control, guides decisions and asserts authority throughout the interaction. Tschopp et al. (2023), applying Fiske's (1992) relational models theory to human-AI relationships, referred to the equality matching mode as peer-like, egalitarian relationships where chatbots work as valuable companions and co-producers of benefits (Chang et al., 2023). Building on this, we suggest that some consumers may engage with AI chatbots as equal partners, a style we term in this study as partner interaction style. For example, when shopping for a portable speaker, a customer may engage collaboratively with an AI chatbot acting as a partner, listening, empathizing and providing personalized recommendations to the customer, fostering a friendly, interactive experience.

Moreover, whether customers exhibit a single or overlapping interaction style with AI chatbots depends on their needs, context, the nature of their prompts and the chatbots' cues. A two-week Nielsen Norman Group study (2023) identified six chatbot conversation types: search query, pinpointing, funnelling, exploring, chiselling and expanding, showing interaction styles shift with customer goals. Simple tasks may require single styles, while complex tasks may trigger longer, adaptive interactions with chatbots, enabling consumers to move between master and partner interaction styles. Xu et al. (2022) confirmed that socially oriented chatbot communication enhances perceived warmth and satisfaction. This suggests that warmth and friendly communication can prompt users to move from servant and/or master to a partner role. Similarly, Haugeland et al. (2022) showed that conversations exploring user interests can heighten perceptions of anthropomorphism, encouraging a move from servant or master styles to more collaborative interaction styles.

Prior studies have examined how users' personal values, perceived benefits and risks involved in smart objects (Lalicic and Weismayer, 2021), conversational and visual cues of smart objects (Huang et al., 2021) and task characteristics (Castelo et al., 2019) shape consumers' perceived value experiences with AI-supported services. Extending this line of inquiry, assemblage theory posits that consumer interaction styles with smart objects are closely related to value co-creation experiences (Hoffman and Novak, 2018). This relationship stems from the fact that interaction styles themselves are shaped by a range of factors critical for value co-creation, which include perceived emotions (Crolic et al., 2022), credibility and usefulness (Flavian et al., 2023), trust (Bae et al., 2023) and motivations for efficiency and information from smart objects (Choi and Drumwright, 2021).

More importantly, perceptions of chatbots can vary between individuals based on the personal roles they assume (such as agentic, passive or communal). These differences shape distinct interaction styles with AI chatbots, which in turn can lead to varying experiences in PVCC with AI chatbots. Past studies indicate that when users perceive technology as a master (master interaction style with AI chatbots), they tend to feel dependent and unable to critically evaluate its outputs (Tan and Forgasz, 2011). Furthermore, when consumers see AI chatbots as servants (servant interaction style with AI chatbots) using them primarily for basic, mundane tasks, they often feel as though they are merely instructing a tool, leading to a reduced sense of involvement, potentially limiting value co-creation and engagement (McGee and Hedborg, 2004; Tan and Forgasz, 2011). Conversely, past literature suggests that when consumers and AI chatbots engage as partners (partner interaction style with AI chatbots), the increased involvement between both the agents (AI and human) potentially results in enhanced reciprocal positive experiences, thus enhancing perceived value co-creation (Tan and Forgasz, 2011; McGee and Hedborg, 2004; Sun et al., 2023). Hence, we propose the following hypothesis and subsets:

Consumer interaction styles with smart technologies such as master, servant or partner affect the level and intensity of engagement, shaping IQ. IQ refers to customers' perceptions of their interactions during service delivery (Brown et al., 2002). Schweitzer et al. (2019) note that servant-style interactions foster comfort and ease, often resulting in positive perceptions of IQ and relevant information exchange in online shopping. Partner-style interactions involve attributing personality to AI, fostering patience, affection and positive relationships, which enhance IQ (Schweitzer et al., 2019). In contrast, master-style interactions often lead to impatience and reluctance, with consumers perceiving unproductive interactions as untrustworthy and unpredictable (Sun et al., 2023). This mistrust can hinder IQ by reducing comfort and control (Schweitzer et al., 2019). Studies using the master–partner–servant framework emphasize how these styles influence interaction diversity and quality (Tan and Forgasz, 2011; De Santis, 2019). Hence, we propose the second hypothesis and subsets:

In AI-enabled online shopping, chatbots are designed to provide advice, recommendations and answers, fostering high IQ essential for value co-creation. As discussed, consumers interact with smart objects differently based on the roles they adopt, such as agentic, passive or communal. Accordingly, IQ might depend on consumers' interaction styles with AI chatbots (master, servant or partner) and is likely to mediate its impact on the value co-creation. Ekinci and Dawes (2009) found that IQ mediates the relationship between consumers and smart objects, aligning with earlier research identifying it as a precursor to satisfaction. PVCC, like satisfaction, stems from IQ in AI-enabled encounters (Lalicic and Weismayer, 2021). Since value co-creation reflects an experience crucial to satisfaction, we argue that IQ resulting from consumer interaction styles with AI chatbots influences PVCC (Brown et al., 2002; Ekinci and Dawes, 2009). Hence, we propose hypothesis 3:

Ekinci and Dawes (2009) highlight that AI chatbots generate specific consumer expectations, making it crucial to examine their role in value co-creation. Studies identify expectations of personalization, convenience, ubiquity and superior functionality as key drivers for engaging with AI-supported services (Lalicic and Weismayer, 2021; Ukpabi and Karjaluoto, 2017). These expectations influence consumers' decisions to adopt AI technologies and shape their anticipation of desired outcomes. Lalicic and Weismayer (2021) suggest that expectations affect how consumers interact with AI, impacting IQ and value co-creation in AI-enabled assemblages. Consumers' tendencies to view themselves as agentic (independent) or communal (relational) lead them to interact with AI chatbots as a master, servant or partner (Barbopoulos and Johansson, 2016), each shaped by varying expectations. Expectations of AI technology can shape the intensity of interaction styles (Gao et al., 2023). For instance, high expectations may make consumers more subservient in master-style interactions, affecting IQ and value co-creation. Because interactions are bidirectional, these AI expectations either, favourable or unfavourable, may critically influence IQ within AI-enabled assemblages (Hoffman and Novak, 2018; Lalicic and Weismayer, 2021) and PVCC (Ukpabi and Karjaluoto, 2017). Thus, we propose Hypotheses 4 and 5:

Finally, the hypothesized relationships between the variables proposed in this study are illustrated in Figure 1.

This study used 5-point Likert-type scales to measure the constructs. IQ was assessed with 23 items adapted from Pelau et al. (2021), while PVCC was measured using three items from Lalicic and Weismayer (2021). Consumer expectations of AI technology, across convenience, personalization, superior functionality and ubiquity, were measured with 12 items (Lalicic and Weismayer, 2021). Interaction styles (master, servant and partner) were measured using 11 items adapted from Tan and Forgasz (2011) and Geiger (2005). A panel of ten industry and academic experts reviewed and refined the items for clarity and validity. Measurement items are detailed in Table 1.

India, with an internet penetration rate of 50%, 37% smartphone Internet users and an e-commerce industry projected to reach $200bn by 2026 (Chawla and Kumar, 2022), was chosen as the study context. Data for this study were collected through a data collection agency during the final quarter of 2022 from online consumers residing in Indian states such as Delhi National Capital Region, Maharashtra (Mumbai, Pune), West Bengal (Kolkata), Karnataka (Bangalore), Gujarat (Ahmedabad), Telangana (Hyderabad), Uttar Pradesh (Lucknow) and Madhya Pradesh (Indore). Ethics approval (REC/01.2022–23) was obtained for this research. Purposive sampling was employed to identify and engage individuals with prior experience using AI chatbots in online shopping contexts. A brief explanation of AI chatbots and screening questions ensured adherence to inclusion criteria. A pilot test with 40 respondents evaluated survey clarity, structure and feasibility. Feedback led to minor refinements in question wording. The main study yielded 472 useable responses. Respondent demographics (see Table 2) indicate that 67.8% of participants were male and 32.2% female, reflecting a higher male engagement in technology use and online shopping in India, consistent with prior studies (Jain et al., 2021; Kuruvilla and Ranjan, 2008).

SmartPLS 4 was employed to estimate the structural model. The study integrated mediation and moderation analyses to examine and test hypotheses on the relationship proposed in the study (Hayes, 2018). Scholars support combining mediation and moderation effects within a single model to better understand and predict relationships among variables in structural equation modelling (Edwards and Lambert, 2007; Hayes, 2018). The interaction term approach was used to test the index of conditional mediation (Hayes, 2018). Conditional direct, indirect and interaction effects were assessed using the bootstrapping approach.

A multivariate normality test score of 1.21 (below 1.96) and skewness/kurtosis values between −2 and + 2 confirmed normality. Variance inflation factor scores ranged from 1.37 to 2.49, below the threshold of 4, indicating no multicollinearity issues. Content validity was ensured through literature review and expert validation (Churchill and Iacobucci, 2002). The overall measurement model showed a good fit with χ² = 2548.69, standardized root mean square residual = 0.039 and normed fit index = 0.841. The reliability and validity statistics of the variables and measures used in the study are presented in Table 1.

Composite reliability ranged from 0.76 to 0.96, exceeding the benchmark of 0.7 (Hair et al., 2010), confirming internal consistency (Fornell and Larcker, 1981). Convergent validity was supported by average variance extracted values between 0.53 and 0.71, above the threshold of 0.5 (Hair et al., 2010). All item loadings exceeded 0.6 except one item (“Interaction with AI is intimidating”), which was removed from analysis. Discriminant validity was confirmed using Fornell–Larcker criteria and heterotrait-monotrait ratio of correlations scores ranging from 0.03 to 0.89, below the threshold of 0.90. Predictive validity was established with Q² values of 0.86 for IQ and 0.68 for PVCC, indicating predictive relevance of the measurement model (Geisser, 1975). Harman's (1967) one-factor test showed a single factor accounting for only 24.9% of total variance, confirming no common method variance bias.

The structural model was estimated using the partial least squares method in SmartPLS, with bootstrapping (5,000 iterations). Table 3 presents the results. The direct effects of consumers' interaction with AI chatbots as master (β = 0.11, p > 0.05), servant (β = 0.03, p > 0.05) or partner (β = 0.02, p > 0.05) on PVCC were non-significant, thus rejecting H1(a), H1(b) and H1(c). However, these effects were mediated by IQ. Consumers' interaction with AI chatbots as master (β = 0.16, p < 0.05), servant (β = 0.12, p < 0.05) or partner (β = 0.34, p < 0.05) had positive and significant effects on IQ, supporting H2(a), H2(b) and H2(c). The strongest effect was observed for partner-style interactions, followed by master and servant styles. IQ positively influenced PVCC (β = 0.37, p < 0.05) and the significant indirect effect through the mediator, i.e. IQ, confirms full mediation, thereby supporting H3.

Moderating effects of consumer expectations of AI technology were assessed for H4 and H5. No moderating effect was found for master or servant interactions on IQ (H4a and H4b) or PVCC (H5a and H5b). However, consumer expectations moderated the effect of partner-style interactions on both IQ (β = −0.047, p < 0.05) and PVCC (β = −0.116, p < 0.05), supporting H4c and H5c with significant but negative effects. The interaction model explained substantial variance in IQ (R² = 0.87) and PVCC (R² = 0.72), indicating strong explanatory power for the proposed relationships.

This study confirms that consumers' interaction styles with AI chatbots (master, servant or partner) affect value co-creation through IQ. Key outcomes and insights emerge from the findings. First, although direct effects of interaction styles on PVCC were unsupported, IQ mediated this relationship. This indicates that value co-creation depends on the quality of interactions. Even optimal styles like the partner interaction style may fail to generate value without high-quality engagement. For example, a partner-style interaction may not create value if the chatbot fails to reciprocate cordiality or provide personalized responses. Similarly, servant or master styles will not yield value without fulfilling, satisfactory interactions. This aligns with service-dominant logic (Vargo and Lusch, 2016), which posits value co-creation can be achieved through meaningful interactions between actors. Consistent with prior research (Lalicic and Weismayer, 2021; De Santis, 2019; Sun et al., 2023), our full mediation findings show that IQ functions as a crucial intermediary linking consumer behaviours, such as interaction styles, to their value co-creation experiences in online environments.

Our study found significant positive relationships between all three interaction styles and IQ, consistent with prior research (De Santis, 2019; Schweitzer et al., 2019; Tan and Forgasz, 2011). Partner-style interactions had the strongest effect, supporting Pelau et al. (2021) that human-like, collaborative AI enhances IQ. Additionally, master- and servant-style interactions were found to be positively related to IQ in our study, as consumers perceived chatbots' usage for mundane tasks or superior functionality for complex tasks as likely to trigger quality interactions and fulfilment (Geiger, 2005; Chang et al., 2023).

Next, consumer expectations of AI technology negatively moderated the effects of partner-style interactions on IQ and PVCC. High expectations can lead to unrealistic demands for human-like traits such as voice, empathy or gestures that chatbots cannot fully replicate (Barbopoulos and Johansson, 2016). As Hoffman and Novak (2018) note, human–AI partnerships evoke complex psychological dynamics and unrealistic expectations, which may destabilize interactions and reduce value co-creation. These findings align with expectancy disconfirmation theory (Daradkeh, 2023; Oliver et al., 1985), which posits that negative disconfirmation arises when there is a discrepancy between consumer expectations and the actual performance of AI chatbots. While previous research has shown that expectancy disconfirmation can lead to co-destruction of value (Rheu et al., 2024), this study demonstrates that disconfirmation related to chatbot performance similarly undermines value co-creation.

Finally, contrary to expectations, consumer AI expectations did not significantly moderate the effects of master- or servant-style interactions on value co-creation, likely attributed to the differing expectation structures inherent in hierarchical dynamics. Unlike partner-style interactions involving emotional and relational engagement, master- and servant-style interactions focus mainly on functional performance and task completion (Devansh and Vocca, 2024; Ruan and Mezei, 2022). Research suggests that in master-servant dynamics, users view AI as “helpful professional secretaries”, with clear boundaries and task-focused expectations (Devansh and Vocca, 2024). These utilitarian expectations such as efficiency, accuracy and responsiveness are easier for AI to meet due to advances in technical performance (Choi and Zhou, 2023; Jiang et al., 2022), making them less prone to disconfirmation than the elevated social expectations in partner-style interactions.

This study makes three key contributions to the literature addressing human-AI chatbot interactions for value co-creation. First, it provides a novel perspective by examining how consumer interaction styles with AI chatbots as master, servant and partner affect value co-creation through IQ using assemblage theory. This study advances understanding within the framework of assemblage theory, widely used to explore consumer-smart object relationships (Hoffman and Novak, 2018; Schweitzer et al., 2019). It empirically verifies distinct collaborative components within the assemblage, including the AI chatbot's agentic role (AI as master), the consumer's agentic role (AI as servant) and the communal role (AI-human as partner). These roles influence the quality of interaction and experiences of value co-creation.

Second, unlike prior studies, which focus on AI's resource capabilities, this research highlights the active role of consumers as agents in shaping value co-creation. By addressing the overlooked assemblage processes, this study extends the value co-creation literature and offers a comprehensive application of the model in AI-enabled online shopping contexts. It underscores how multiple actors interact and assemble to co-create mutually beneficial value.

Third, the study posits that evolving consumer expectations based on AI's promising future can stabilize or destabilize these outcomes, enriching assemblage theory by demonstrating how consumer expectations shape co-creation dynamics. On these lines, this research examines how consumer expectations of AI moderate the relationship between interaction styles and PVCC. Findings reveal that high expectations can diminish value co-creation for partner-style interactions. Unrealistic demands place strain on partner relationships, an observation consistent with research on hedonic goal orientations (Barbopoulos and Johansson, 2016). This insight is critical for understanding how expectations influence consumer-AI interactions.

This study provides key insights for market researchers and retailers. First, the findings highlight that recognizing consumer interaction styles with AI chatbots as master, servant or partner and their influence on IQ and value co-creation is crucial. This insight helps managers improve marketing effectiveness and achieve a competitive advantage by developing chatbots that adapt to various consumer styles.

Second, most chatbot design focuses on technical features, but our findings emphasize a user-centric approach. Designers should build chatbots that use sentiment analysis to recognize and adapt to different user styles, for example, providing task-focused responses for servant-style users and engaging conversationally with partner-style users to enhance value co-creation.

Third, our study shows that high-quality interaction is essential for value co-creation, regardless of user interaction styles. Chatbots should undergo rigorous quality testing evaluating clarity, responsiveness, accuracy and speed before deployment. For example, designers can use performance measures like response time, feedback mechanisms and regular sentiment assessments to ensure superior IQ. Finally, as evidenced in our study, consumer expectations negatively moderate partner-style interactions. Hence, AI chatbot designers should set realistic AI capability expectations from the outset using clear, user-friendly prompts. Avoid overpromising in marketing and instead educate consumers about effective chatbot use to improve experiences and prevent frustration (Gao et al., 2023).

This study has limitations that suggest avenues for future research. First, it relied on self-reported perceptions, and hence, future studies could replicate the model in field settings using experiments to enhance external validity. Second, while this study focused on consumer interaction styles, additional factors such as personality types, locus of control, self-efficacy and subjective norms could further explain value co-creation in online shopping. Future research should explore other moderators within the assemblage framework to improve predictions regarding the relationship between interaction styles and PVCC. As Ruan and Mezei (2022) suggest, interaction styles may vary depending on product type. Hence, future studies could examine these dynamics in utilitarian, hedonic and luxury consumption contexts. Finally, extending the research to different cultural contexts would enhance the generalizability of the findings.

This study enhances our understanding of consumers' PVCC in AI-enabled online shopping by focusing on the encounter process within the value co-creation model. Drawing on assemblage theory, we demonstrated that value co-creation arises from the combined contributions of multiple actors, highlighting how consumer interaction styles (master, servant and partner) shape both IQ and value co-creation outcomes. The moderation model revealed that high consumer expectations of AI can diminish value co-creation, highlighting the need for companies to avoid overpromising AI capabilities.