The user experience of voice assistants in retailing: a qualitative comparative study Open Access

Drawing on the theories of planned behaviour (Ajzen, 1991) and reasoned action (Azjen, 1980), this paper examines the influence of voice assistants (VAs) on the shopping experience of online shoppers in Nigeria and the UK. Despite recent advances in VAs and voice recognition devices in the customers’ decision-making process, we still lack research examining the impact of artificial intelligence (AI) on customers’ decision-making processes in diverse cultural contexts. For instance, we need to understand if and how cultural and institutional contexts influence the effectiveness of VAs in retailing. We view this as a significant omission in the theoretical development of the AI and consumer behaviour literature streams, which our research is set to address in the unique contexts of Nigeria versus the UK.

The application of VAs (also known as “predictive chatbot” or “voice assistant”) in service marketing has increased. For instance, smart devices such as Google Home, Google Assistant or Mycroft on Android smartphones, Apple’s Siri on iPhones and iPads, Bixby on Samsung phones, Amazon’s Alexa, Amazon Echo and Microsoft’s Cortana have been on the rise. Proponents of AI in the field of service marketing argue for high information processing speed (De Bellis and Johar, 2020), high capacity to perform some mundane or repetitive electronic tasks (Shih, 2020), simulate dialogue with the human user (Klaus and Zaichkowsky, 2022), and thus an improved relationship with customers (Troshani et al., 2021) as the motivation for investments in these advanced gadgets.

VAs’ capacity to influence consumers’ decision-making (Mari, 2019), revolutionise the shopping process (Shi et al., 2020), enhance firms’ marketing/business strategies (Samiee, 2020) and facilitate smart working and 24-h customer support with minimised language barriers (Figueiredo et al., 2022) has also been highlighted. VA has been argued a timely response to the changing business–customer relationship in service firms (Troshani et al., 2021; Del Giudice et al., 2022), especially since the COVID-19 pandemic (Hassounah et al., 2020; Rautela and Agrawal, 2020; Sharevski et al., 2022; Shetty and Pai, 2021).

Furthermore, apart from retailing, the use of VAs to fulfil varied needs by individuals in their daily lives (e.g. social and utility) is also increasing (Jain et al., 2022). Given other benefits of VAs, e.g. in traffic (re)directions, hotel and travel bookings (Olson and Kemery, 2019), researchers (e.g. Tabassum et al., 2019) have suggested that the industry 5.0 VAs could recognise when families discuss their meal plans and could make recommendations around restaurants, takeaways, or inviting loved ones for dinner. Despite these benefits, the issue of customers’ dissatisfaction (Heitmann et al., 2007; Sharma et al., 2022), discomfort, insecurity, uncertainty (Blut and Wang, 2020; Sia et al., 2004; Vimalkumar et al., 2021) and low self-esteem (Usta and Häubl, 2011) have been under-researched. Notwithstanding such a burgeoning body of VA literature, research examining the application of VAs in diverse cultural contexts is, at best, limited. We view this as a significant gap in the marketing literature.

Moreover, another key benefit of using VAs in retailing is their capacity to mimic humans in conversations (aka anthropomorphic behaviour) (Delbaere et al., 2011). Through anthropomorphism, VA technology facilitates customer service and operational capacity and reduces costs (Buhalis and Moldavska, 2022) and thus is a key to customers’ satisfaction and competitive advantage (Mendes Ferreira et al., 2022). By allowing customers to delegate tasks (Troshani et al., 2021), anthropomorphism in VAs has gained widespread acceptance and has transformed the 21st-century retailing landscape (Sharma et al., 2022).

Consequently, Apple’s Siri, Amazon’s Alexa, Microsoft’s Cortana and Google Assistant (for instance) have been integral parts of marketing strategies to build consumer-brand relationships (Mari, 2019; Shih, 2020). However, some ethical issues – job loss (Cassell et al., 1994; Epley et al., 2008; Huang and Rust, 2018), lack of interpersonal relationship with the customer (Qiu and Benbasat, 2009) and psychological concerns (Jain et al., 2022) because of our overdependence on AI and voice commands (Han and Yang, 2018) – are also emphasised. However, there is still a lack of research comparing the effectiveness of VAs in environments with significantly varying levels of AI and voice command adoptions, such as the UK versus Nigeria. For instance, we need to understand the factors that might support or limit the adoption of AI and VAs in these diverse cultural contexts. We also need research that compares these customers’ capacity to interact with AI and VAs. We view these as significant research gaps which our paper is set to address.

Besides, despite the benefits of AI and VAs, e.g. improved speed, reliability, efficiency of tasks (Cui et al., 2021) and improved quality of life (De Bellis and Johar, 2020), research examining the adoption of AI in emerging markets is, at best, limited. For instance, we lack empirical data on AI and VAs in the emerging economies’ retail landscapes, their human factor challenges, the socio-cultural, socio-economic, safety and ethical issues involved and the emerging markets customers’ perception of VAs.

To address these gaps, we take a cross-disciplinary approach, drawing from psychology, human–computer interaction and consumer behaviour literature, and we make four complementary but distinctive contributions to the marketing management literature. First, drawing on the theories of planned behaviour (Ajzen, 1991) and reasoned action (Azjen, 1980), we investigate shoppers’ motivation for adopting VAs. Second, we investigate the environmental factors that might influence the application (and effectiveness) of VAs. Third, we examine customers’ attitudes, behaviours and perceptions regarding VA usage. Finally, we examine several salient issues ignored in previous VAs and AI research. To achieve these objectives, this paper addresses the following research questions:

The rest of this paper is structured as follows. The next section reviews the relevant literature. Section 3 presents and justifies the methodology. Section 4 presents and discusses the findings, while Section 5 concludes with theoretical and practical implications, limitations of the study and future research directions.

The development in retailing through the introduction of multichannel and omnichannel has revolutionised the modern retail contexts (Gao et al., 2022). Most recently, the evolution and practical relevance of AI in retailing, such as VAs, have enhanced shopping experiences with more personalised services (Grewal et al., 2022; Kamoonpuri and Sengar, 2023). With the changing retail environments, the expectations of customers are growing, which has influenced their self-expression, captivity and addictive behaviour (Ramadan, 2021). Indeed, VA in retailing is reshaping the future through multi-faceted relationships between retailers and customers (Ramadan, 2021) and functional and social attributes (Pitardi and Marriott, 2021). Alternatively, retail brands are struggling with the optimal implementation of VA-empowered conversational commerce with minimum algorithmic biasness and customer-personalized constraints with retail offerings (Rabassa et al., 2022; Kautish et al., 2023).

In recent years, the diffusion of VAs in retail has significantly influenced consumer brand engagement (McLean et al., 2021; Gupta and Mukherjee, 2025) and consumer-brand relationships (Huh et al., 2023). Consumer intentions play a pivotal role in shaping these interactions, as they determine not only initial adoption but also continued usage and brand loyalty. Research on VAs’ emotional attachment and satisfaction remains limited (Kumar et al., 2024), yet Singh (2022) highlights that emotional connections with VAs significantly impact purchase satisfaction and repurchase intentions. Given the complexity of consumer behaviour in accepting and integrating VAs into their shopping experiences, future research must explore the interplay between consumer intentions and the utilitarian benefits of VAs (McLean and Osei-Frimpong, 2019). Moreover, advancing VA research requires examining cognitive, affective, social, cultural and behavioural branding outcomes, including trust, loyalty and engagement (Huang et al., 2021; Vernuccio et al., 2023; Rohit et al., 2024; Gupta and Mukherjee, 2025). Understanding how consumer intentions drive interactions with VAs in retail will be critical to enhancing user experience, fostering deeper brand connections and optimizing the role of VAs in the evolving retail landscape.

Table 1 synthesises the recent developments in VA research within different retailing contexts. According to Table 1, in the earlier years, research dominantly focused on acceptance models (Moriuchi, 2019; Lee et al., 2021; Pitardi and Marriott, 2021; Ramadan, 2021; Aw et al., 2022). Gradually, the research on VAs has transitioned toward experiential value (Kautish and Khare, 2022; Khare et al., 2023), brand attachment (Prentice et al., 2023), AI–human relationships (Guerreiro and Loureiro, 2023) and psychological dependency (Xie et al., 2023). Although earlier research focused on different facets of VAs’ acceptability within the retail context, research still fails to unleash the sociocultural dimensions of consumers’ behavioural intentions toward using VAs in emerging and developed markets.

The rapid growth of AI in retail has transformed consumer interactions, particularly through AI anthropomorphism, where VAs exhibit human-like attributes to enhance user experience (Ling et al., 2021; Aw et al., 2022; Blut et al., 2024). Given this evolution, understanding consumer adoption of VAs requires a theoretical foundation that accounts for attitudinal and behavioural intentions. The theory of reasoned action (TRA) and its extension -the theory of planned behaviour (TPB)- provide a critical lens for analyzing how consumers form attitudes and behavioural intentions toward AI-driven retailing. These theories explain how psychological and social factors drive consumer acceptance and usage of VAs, making them highly relevant for the current study. 

The TRA advances that an individual’s behaviour is determined by two key constructs: attitude (the individual’s evaluation of performing the behaviour) and subjective norm (perceived social pressure to engage in the behaviour) (Ajzen and Fishbein, 1980). This theory is particularly useful in AI anthropomorphic retailing, where consumer perceptions of VAs as social and interactive agents shape their willingness to use them (Whang and Im, 2021). The TPB, an extension of TRA, further refines this understanding by introducing perceived behavioural control (PBC), the extent to which an individual perceives the ease or difficulty of performing the behaviour (Ajzen, 1991). This addition is crucial in VA adoption, as consumers may perceive control constraints, such as lack of technical knowledge, privacy concerns or trust issues, that impact their adoption decisions.

The anthropomorphic nature of VAs significantly influences consumer attitudes toward their adoption. Studies show that consumers perceive VAs as pseudohuman agents, expecting them to behave in a human-like manner during interactions (Whang and Im, 2021). When VAs demonstrate human-like voice attributes, they enhance perceived brand anthropomorphism and create a stronger social presence, which positively impacts engagement and adoption (Vernuccio et al., 2023). However, despite these advantages, PBC factors, such as privacy concerns and lack of trust, can hinder adoption (Van Doorn et al., 2017; Liang et al., 2021). These elements reinforce the relevance of TPB, as PBC directly accounts for external constraints that affect consumer willingness to engage with VAs in retail.

Consumer behavioural intentions toward AI anthropomorphic retailing are shaped by perceived usefulness and trust (Lee et al., 2021; Jang et al., 2022; Ashrafi and Easmin, 2023). The TPB is particularly useful in explaining how these perceptions translate into adoption behaviour. Perceived usefulness, a key determinant of attitude, is critical in shaping how consumers view VAs as facilitators of efficient and personalized shopping experiences (Huang and Rust, 2018; Gupta et al., 2020). Additionally, trust plays a central role in subjective norms and PBC, as consumer concerns about data security and privacy directly impact adoption intentions (Pantano and Pizzi, 2020; McLean et al., 2021). Research further suggests that demographic variations influence consumer attitudes and trust toward VAs, highlighting the need for contextualized adoption (Evanschitzky et al., 2015; Mishra et al., 2022; Anshu et al., 2022).

The adoption of VAs in retail is not merely an individual consumer decision but is also shaped by value co-creation, the collaborative interaction between brands, AI technology and consumers (Mele et al., 2021; Grewal et al., 2021). By leveraging AI to enhance user engagement, retailers can strategically shape attitudinal and behavioural intentions through personalized experiences, adaptive interactions and trust-building mechanisms. Therefore, TRA and TPB offer a theoretical foundation to analyze how these co-created experiences influence consumer behaviour, providing a structured approach to studying AI retail adoption in a digitalized economy. The application of TRA and TPB in AI anthropomorphic retailing is critical to understanding consumer adoption of VAs. These theories help explain how attitudes, subjective norms and PBC shape consumer willingness to integrate VAs into their shopping behaviours.

The socio-cultural dynamics within AI-enabled retailing, particularly in the context of VAs, are inherently complex and multifaceted (Grewal et al., 2022; Kamoonpuri and Sengar, 2023). Understanding these dynamics is essential across various retail settings to enhance AI-driven conversational commerce. In marketing, socio-cultural factors significantly influence consumer behaviour and brand loyalty. Despite the growing body of research on AI anthropomorphic retailing, limited attention has been given to these socio-cultural dimensions (Van Doorn et al., 2017; Han and Yang, 2018; Whang and Im, 2021). Therefore, future research should explore the interplay of socio-cultural mechanisms to develop culturally sensitive and ethically responsible AI anthropomorphic retailing strategies, enabling businesses and consumers to co-create value (Dwivedi et al., 2022).

Anthropomorphism, the human tendency to infuse naturally non-human objects with human-like characteristics, is a concept gradually applied in different AI applications and contexts, such as retailing, to generate a higher level of consumer engagement, a sense of familiarity and stronger emotional connections (Epley et al., 2008; Waytz et al., 2010; Eyssel and Kuchenbrandt, 2012). In the context of AI anthropomorphic retailing, this means creating VAs that mimic human behaviour and interaction.

Cultural norms and values play a pivotal role in shaping consumer acceptance and intentions toward AI anthropomorphic retailing, with notable distinctions between non-Western (collectivist) and Western (individualist) societies (Bartneck et al., 2007; Lim et al., 2021). Additionally, socio-demographic factors such as gender, age, education and technological proficiency hold varying degrees of influence across different socio-cultural contexts (Chiang and Trimi, 2020; Blut et al., 2021). While AI anthropomorphic retailing has enhanced consumer engagement, personalized services and operational efficiency (Huang and Rust, 2018), there remains a critical need for deeper exploration of how these socio-cultural factors shape consumer behaviour and attitude toward VAs.

We adopted a qualitative methodology in this study to explore motivations and challenges underlying user experiences with VAs. This method was chosen because it allows for a deep exploration of the participants’ experiences and perspectives, which is crucial in understanding the socio-cultural dynamics of AI anthropomorphic retailing. Our interview protocol has four main focuses/sections. Section 1 was designed to help us establish the factors that motivate our participants to use VAs. Part 2 helped us discover their capacity to interact with VAs, the challenges involved and the benefits that accrue from their use of VAs. The third part of our interview questions was designed to determine if and how socio-cultural, environmental and institutional factors impact their VA usage. Section 4 examined the relationship between their VA usage, their level of (dis)satisfaction, as well as their ensuing brand loyalty. Researchers argue that gathering such subjective opinions involves storytelling via in-depth semi-structured interviews (Knight and Jarzabkowski, 2022).

Because interviews offer opportunities to obtain the actors’ side of the story (Liu and Rong, 2015), in our approach to the interview, we acknowledge our participants as the principal actors, as their stories are vital to addressing the critical issues raised in this research. We, therefore, allowed our participants ample time to reflect on their VA usage without interruption (Dick and Collings, 2014; Ezzamel and Willmott, 2008; Liu and Rong, 2015). Given that our participants are discrete across two countries, we used a purposive sampling approach to reach out to the most relevant participants, following the criteria:

• online shoppers in Nigeria and the UK using VAs in their shopping; and

• online shoppers with a strong feeling of companionship with VAs.

We recruited 30 participants with these conditions; their demographic details are in Table 2. So, initially, we had 40 willing participants, but we stopped interviewing at the 30th participant, considering data saturation (Fugard and Potts, 2015). The consent of participants was sought via email and the data collection occurred between April 2022 and February 2023.

Our approach to data collection was in-depth, semi-structured interviews. The conversations were all recorded, and each session lasted approximately 40 min. We started each session by first familiarising ourselves with the participants. This was followed by stating the purpose of our research, the aims and objectives. We also highlighted the ethical issues involved and how these will be managed. We then highlighted the interview structure and process. Then, each participant was asked whether they had used VAs in the decision-making. As they all answered yes, we asked them what motivates them to use VAs. They were also asked to give an overview of their experience with VAs.

Participants were then explicitly asked whether they faced any challenges while using VAs. As every participant answered yes, the follow-up question was, “What type of challenges have you faced when using VAs”? For each challenge mentioned, the participant was asked: “How did you deal with such a challenge”? They were then asked if their use of VAs had impacted their perception of the brand. In furthering the discussions, participants were asked if they preferred a specific type of VA. They were also asked about the frequencies of their VA usage. Each response was probed further with specific questions, including whether and how their VA usage impacted brand loyalty. Participants were also asked to reflect on if and how their adoption of VA in their decision-making has impacted their (a) buying decision-making, (b) ease of purchase, (c) satisfaction and (d) brand loyalty. Finally, participants were asked to reflect on (1) if they think they have been effective in their use of VAs, (2) if they think their shopping habits could be better without VAs and (3) if, and why, they think their VA usages might be out of their control.

Throughout the conversations, participants were asked to provide details of specific examples of stories to support their responses. However, we recognised that our participants might engage in impression management (Rhodes and Brown, 2005). Therefore, we were cautious in wording the interview questions. Also, serving as a social anchor (Brown et al., 2008), the interviewer was quick to spot incidents of impression management when they arose while challenging specific claims made by participants (Knight and Jarzabkowski, 2022). To encourage participants to share their views on using VAs, the interviewer often shared their experience with using VAs. This encouraged participants to narrate their experiences through sensemaking (Gabriel, 2000).

We commenced our data analysis by familiarising ourselves with our data sets before moving on to data coding (Patton, 2015). We followed a detailed process of thematic analysis using guidelines from Gioia et al. (2013). Each member of the research team examined all interview transcriptions separately and independently. Drawing on earlier research on discursive analysis of subjectivity (e.g. Knight and Jarzabkowski, 2022), we constantly went back and forth between the emergent themes (from our interview data) and the extant literature (Laine and Vaara, 2007; Mantere and Whittington, 2021) to ensure correctness. Each member (of the research team) coded an assigned set of raw data using open coding based on the relevant themes before passing it on to other team members for checking. Acting as critical friends (Kember et al., 1997) and research auditors (Filho and Rettig, 2016). Finally, we cross-checked the coding process by integrating the coding structure to unify the interview quotes to guarantee quality control and enhance validity (Alo, 2020). However, reaching such a consensus was not easy, as there were a few disagreements before arriving at a consensus (Patton, 2015). Our rigorous coding process produced a set of first-order codes, second-order themes and aggregate dimensions of user experience with VAs. For quality control and to enhance validity, we shared our data with our participants (Patton, 2015), and their feedback was reassuring.

In addition to the manual coding process, we also concept-mapped the quotes from participants using LexiPortal (Wilk et al., 2019), software used for automated content analysis to provide an unbiased view of data. Similar analysis has been considered more rigorous when combined with traditional manual analysis (Nazir, 2023). Moreover, the concept map revealed hidden nuances and improved our understanding of user experience with VAs.

Despite choosing samples from different cultural backgrounds, a few themes ran through our interview data. Given the similarities in our interview data from both sides of the Atlantic, we decided combining the quotes under some relevant subheadings would make more sense instead of drawing a comparative analysis. For instance, our participants have similar motivations for using VAs. Secondly, they admitted experiencing similar challenges/barriers in their VA usage. Thirdly, they highlighted how VAs are influencing their purchase behaviours and decision-making. Fourthly, they revealed some benefits they derive from their VAs usage. Finally, they linked their use of VAs to their brand loyalty and their (dis)satisfaction levels. To help us make sense of these findings in our data analysis below, we use three key themes termed (i) Motivations for using VAs, (ii) Challenges in VA usage and (iii) The influence of VAs on brand trust and brand loyalty. Figure 1 shows themes of user experience with VA.

The concept map shown in Figure 2 complements thematic analysis findings and provides a deeper understanding of user experience with VAs. Specifically, while it might sound weird that some buyers prefer VAs to human sales assistants, Figure 2 highlights some of the key reasons behind buyers’ adoption of VAs – the capacity to listen, understand, interpret, show compassion, search for items, store information and make recommendations. Although the use of VAs (e.g. Google Assistant and Apple Siri) comes with some issues – trust, privacy and security – by saving time, providing security to human users and assisting in running some repetitive daily functions, the use of VAs in service marketing makes shopping easier and fosters brand–customer relationships.

We begin our data analysis by showing why some shoppers prefer VAs. The interview quotes below will use two key themes (anthropomorphism and convenience) to present and analyse our participants’ motives for using VAs.

Our participants highlighted their passion for VAs’ capacity to personify and imitate humans. The next set of quotes reveals that participants have formed human-like relationships and bonds with such inanimate objects:

In retail marketing, anthropomorphism (the power of modern technology to mimic humans in conversations) helps managers alter consumer purchasing intentions (e.g. Wan and Chen, 2021), changing how individuals purchase and consume goods. However, it is also a key to building effective customer relationships (van Esch et al., 2019). However, anthropomorphism also provides comfort for humans from their homes, as the next set of quotes reveals that our participants also use VAs to perform some routing tasks in their homes to fill some gaps in security while providing comfort for the users:

The above quotes are consistent with earlier research on how people can become used to technological applications such as VAs at home to assist with housekeeping and routine household tasks (Pitardi and Marriott, 2021).

The next set of quotes reveals that because of the interactive capacity of VAs, their accessibility and the high degree of satisfaction that ensues, our participant’s shopping experiences are positive:

The above quotes are consistent with recent research examining VAs’ capacity to detect, interpret, store, retrieve and communicate complex information to human users in real time using cutting-edge AI technologies (Vimalkumar et al., 2021). Similarly, the below quotes are consistent with previous research on different orientations for shopping online and the varying attitudes of online shoppers toward online purchases (Yeo et al., 2017). For instance, apart from the quality of the interactions involved, our participants also revealed that their intentions to purchase online are driven by the level of literacy support involved (Pincott and Branthwaite, 2000) and the convenience when using VAs (Ameen et al., 2021):

Consistent with earlier research on VAs’ capacity to mimic the physical store experience (e.g. Klaus and Zaichkowsky, 2022), the above quotes show participants’ satisfaction with their buying experience because of AI’s capacity to replicate a natural/in-store shopping experience. Although the above quotes support recent research that links VAs with impression management (e.g. Coulthard and Keller, 2012), increased customer well-being and business growth (e.g. Ameen et al., 2021), our participants reveal some dissatisfaction with VAs. Specifically, there are scepticisms around VAs’ capacity to execute specific commands effectively, and consequently, the issue of trust, which results in a somewhat unpleasant experience:

The above quote contradicts a recent study that argues that Amazon’s pre-shipping expertise improves the logistics of their product delivery (Klaus and Zaichkowsky, 2022). Instead, the above quotes highlight the risk of purchasing the wrong or multiple goods by mistake, as the buyer cannot see or edit their basket before making payment. Earlier research also highlights the downsides of VAs and AI in buyer decision-making. These include awkward feelings because of some individuals’ inability to interact with modern technology (Maroufkhani et al., 2022), which can arise because of distress caused by constant interaction with technologies, as well as data misuse (Osiceanu, 2015), because of the fear of mismanagement of personal data. These issues are examined in-depth in the next theme, which explores the challenges of using VAs.

Despite the benefits of VAs and AI in service marketing, participants also revealed their huge concern about adopting these modern technologies in their buying decision-making. Specifically, they highlighted issues around big data, spying technologies, data breaches, data privacy, security risks and logistics issues when multiple shopping is involved. There is also the social disorder associated with the overdependence on AI. This also highlights some shoppers’ reasons for resisting the adoption of VAs. These are explored further under the next set of subthemes.

Participants revealed their fear of modern/sophisticated technologies. While participants were already attracted toward AI for their shopping because of some accruing benefits – anthropomorphism, convenience, companionship and literacy support – concerns around data insecurity have affected their confidence levels. Earlier studies view technophobia as a psychological orientation or attitude toward technology, which when developed can undermine the original attraction toward technology, resulting in limited use, fear of using it, or not using it with confidence (Khasawneh, 2018; Martínez-Córcoles et al., 2017; Ronit, 2011). These are consistent with the below quotes. Specifically, the below quotes consist of three main elements that illustrate our participants’ concerns over adopting modern technologies – mistrust, ignorance and fears about the effects of technology on users (Osiceanu, 2015; Kim et al., 2023):

The above quotes reveal why some shoppers may be sceptical about adopting VAs. Technophobia – the fear of new technology – also highlights why the varying levels of e-commerce adoption in our modern society are linked to our varying levels of uncertainty avoidance, along with our subjective norms and behavioural intention (Akhtar et al., 2019; Huang et al., 2019; Merhi and Ahluwalia, 2017).

Contrary to earlier quotes that show VAs’ capacity to identify, interpret, retrieve and communicate complex information in real-time using advanced AI as the key motivation for our participants’ adoption of VAs in their buying decision-making, our participants also highlighted their fear around data insecurity. They also expressed concerns about extensive data mining:

On the other hand, some other participants had a difference of opinion:

Furthermore, apart from their fears about how their data is managed, our participants also expressed deep concerns over the use of difficult-to-comprehend jargon, technical languages and terminologies, coupled with data overload, communicating excessive information to users, which they argued is a decoy to hide critical information:

The above quotes have revealed some VA users’ concern over potential deliberate attempts by marketers to conceal vital information that should influence their buying decision-making despite the recent enactment of General Data Protection Regulation (GDPR) laws, which should serve as guidelines for data collection and management (Katsikeas et al., 2020). Given such mistrust and lack of protection of consumers’ rights, our interview questions further investigated whether VA users are fully aware of the risk factors involved in their identity verification online and whether this affects their perception and level of satisfaction:

Our interview questions also investigated whether participants feel they have any control over how their data is used or feel they have lost control. Researchers examine the interplay between big data in voice technologies and the Internet of Things (IoT), arguing that it becomes tough for users to exercise/make their own choices (Nguyen et al., 2021). Thus, they are entrapped in authoritarianism, lack of empowerment, autonomy and, ultimately, lack of choice (Dholakia et al., 2021). Despite the high level of risk and the perceived vulnerability, overall, participants revealed their satisfaction and willingness to share their private information via their VAs, provided they are getting the level of satisfaction and service quality they expect:

The above quotes reveal mixed reactions to questions examining issues around data security and shoppers’ trust in the way AI and VA manage customers’ data. This also reiterates our individual differences, our varying levels of uncertainty avoidance, as well as our subjective norms and behavioural intention to adopt AI (Akhtar et al., 2019; Huang et al., 2019; Merhi and Ahluwalia, 2017).

Another recurrent theme across our interview quotes is the apprehension over the listening feature on VAs. Our interview data clearly showed that our participants felt marketers were spying on them. They contend that their VA actively listens to their private conversations and gathers data (on behalf of the marketers) based on what they have discussed without their consent. For instance, participants revealed receiving targeted adverts related to their recent discussions on their respective social media pages shortly after engaging in such a conversation, and therefore a fear of possible intrusion into their privacies without their consent:

Furthermore, our research also found that while participants may not be happy that they are being spied on and that their privacies are being eroded, they must adapt to the situation because of a lack of choice. Consequently, some VA users have found an innovative way of bypassing the electronic surveillance and unauthorised invasion into their privacies:

Participants’ responses above show shoppers’ growing mistrust and apprehension regarding conversations with their VAs because of fear of a potential breach of GDPR laws. The above quotes also show how some VA users have now developed coping mechanisms of concealing some of their data and taking extra precautionary measures in their private conversations to secure their privacies.

A key built-in feature in VAs is the device’s bilingual nature. Our interview questions also investigated the role of the bilingual capacity of their VAs in enhancing participants’ satisfaction. British participants using Google Assistant and Apple Siri are happy with their device’s ability to understand or switch between different languages. A common theme running through the responses from Nigerian participants was their frustration with VA’s capacity to switch between various world languages but their failure to understand other accents apart from the Caucasians’. Specifically, participants from Nigeria decried their device’s inability to recognise their accents, as they argued their device was not originally built to recognise accents apart from the British and American accents:

Earlier research on audio bias and disparity argues that while Alexa can switch between French and English in Canada; English and Spanish in the USA; and English and Hindi in India, Google Assistant’s multilingual capacity only covers a combination of 12 major languages. Similarly, the Apple Siri can understand the American, Australian, British, Indian, Irish and (“White”) South African voices (Moussalli and Cardoso, 2020). Researchers also argue that certain ethnic groups were not considered in the original design of these VAs (Baeza-Yates, 2018). Similarly, the British participants also reported issues around invocation inconsistency. Consequently, such a preset nature of VAs limits their usefulness, equitability, enjoyability and usability because of the difficulty in interacting with VAs, which increases the users’ level of frustration:

The above quotes are consistent with earlier research that found auditory challenges facing VA users, including pitch, volume, intonation and speed of conversations (Longo and Padovano, 2020), resulting in VAs’ poor ability to interpret the users’ commands. Consequently, our participants have found some coping mechanisms:

We also asked if participants are facing specific environmental, socio-cultural or socio-economic challenges in using their devices, and a common theme running through the quotes from the Nigerian participants is institutional voids in the continent:

Generally, this section has revealed audio bias and institutional voids as factors impacting participants’ VA usage, which link to the level of satisfaction derived. Participants from Nigeria specifically highlighted a lack of electricity supply. Recent research on the impacts of institutional voids on business development in Africa (e.g. Alo and Arslan, 2021) also highlights a lack of electricity as a significant barrier to business development in the region.

This section examines how the benefits of using VAs can influence customers’ attachments to a brand’s market offering, which are particularly relevant to marketing managers.

Consistent with earlier research on the relationships between brand loyalty, brand performance, value, trustworthiness, reputation, brand name and VAs (Klaus and Zaichkowsky, 2020), responses from both sides of the Atlantic show that shoppers have higher levels of trust when interacting with a specific VA:

The above quotes imply shoppers’ lack of confidence in certain VA brands. The findings above imply that simplifying their privacy and data security policies can enhance customers’ trust in a VA brand. Given the relationship between privacy and trust (Wang et al., 2020) and coupled with an individual’s need to control how businesses use their data, researchers (Klaus and Zaichkowsky, 2022) have suggested that VA brands with a genuine reputation and successful track record (of data management) have a more substantial capacity to convince customers to buy a particular product or service. Similarly, earlier research (e.g. Kamarulzaman, 2011) also argues that consumers’ trust in an e-retailer can influence their intentions to shop online with VAs.

This section links participants’ online shopping experience to their level of satisfaction. A critical feature of VAs is their capacity to recommend products to users. Accordingly, the following quotes examine whether participants are satisfied with their VAs’ recommendations. As can be seen in the quotes below, while some of our participants were unaware that their VAs could make product recommendations, others preferred products recommended by their VAs because of a high level of trust in their VAs:

Although the above quotes support earlier research linking service brands’ relationship marketing to customers’ cognitive psychological comfort attained through differentiation and consistency in VAs (e.g. Hu et al., 2023), our participants were somewhat divided on accepting their VAs’ recommendations. For instance, some participants revealed they prefer conducting independent online reviews to asking their VAs for reviews. Specifically, they argue that their VAs are biased in their recommendations, as they believe that the manufacturers influence such recommendations:

The above quotes indicate that while VAs can recommend goods for human users, the issue of lack of trust must not be ignored. Some participants prefer to carry out reviews themselves because of the fear that the manufacturer can influence such recommendations (Klaus and Zaichkowsky, 2020). This highlights the significance of the interplay between consumer trust, brand reputation and customer satisfaction.

Because of growing customer interaction and engagement with VAs, major brands are keenly leveraging the advantage provided by AI to provide tailored market offerings, gain the customers’ attention and thus expand their market shares (Rzepka et al., 2022; Ameen et al., 2021). Despite such findings, this section shows that the use of VAs is limited to purchasing low-involvement goods, such as candy or ice cream, as previous research (e.g. Kraus et al., 2019) also found that the chances of purchasing high-involvement goods via VAs are low. The next set of quotes reveals participants’ unwillingness to shop for high-involvement goods with VAs because of lack of trust:

The above quotes reveal participants’ lack of trust in VAs’ capacity to handle expensive purchases. Similarly, Labecki et al. (2019) argue that although e-commerce has enhanced voice shopping and is helping customers overcome their initial anxiety about making purchases without first seeing, feeling, or smelling the item, customers’ reliance on voice solutions for their online shopping is limited to acquiring low-involvement goods. Although because of brand trust, a small percentage of customers from both sides of the Atlantic (i.e. 2 participants) revealed that they do not mind purchasing high-involvement goods via their preferred VA – Apple Siri, especially in a secluded place:

The above quotes show that while customers value VAs’ capacity to imitate humans in conversations, there are limits to what type of products shoppers are willing to purchase via VAs. For instance, participants would prefer to shop for low-involvement goods with their VAs and not high-involvement goods. Also, Table 3 presents a comparative view. Overall, this section has shown how the application of VAs in service marketing can influence shoppers’ overall shopping experience. It also shows shoppers’ motivation for using VAs, the challenges and pitfalls that may lie ahead and how VAs can influence customers’ level of trust and their overall satisfaction. The following section discusses our study’s implications for theoretical development and managerial practice.

This study sought to examine the factors that influence user experience with VAs. It critically evaluates how AI and institutional factors can influence shoppers’ perceptions of VAs and the effectiveness of voice commerce services in diverse retail environments. Although we chose our samples from diverse cultural contexts, the interview data show strong similarities in our participant’s responses. First, our participants reveal that their motivations for using VAs are its capacity to anthropomorphise humans in conversations, provide comfort and safety for humans and assist with housekeeping. Consequently, some participants have developed trust and formed human-like relationships with such inanimate objects. This is consistent with earlier research linking anthropomorphism to increased consumers’ confidence, satisfaction and brand trust (e.g. Foehr and Germelmann, 2020; Wan and Chen, 2021).

Secondly, because of its language skills, VAs provide learning support for human users, especially those with specific learning difficulties such as dyslexia. The capacity to establish quality interaction and communication with the users with varying degrees of learning needs is argued to enhance the accessibility of VAs (Ameen et al., 2021). Thirdly, despite these findings, some of our participants highlighted some awkward experiences with their VAs because of communication breakdown and distress (Osiceanu, 2015). Other dark sides of VAs highlighted by our participants include data misuse, logistics issues involving multiple shopping, ignorance and technophobia. Although the above challenges may highlight dissatisfaction among VA users, most of our participants are satisfied and trust their VAs at the expense of the specific product/service being advertised/purchased. We also found that most shoppers can only rely on VAs for buying low involvement goods. The following section highlights our study’s implications for marketing management.

Theoretically, the findings of our study enrich the understanding of the user experience of VAs with a granular focus on customer motivations, challenges and the potential tensions of the user experience of VAs with brand trust and brand loyalty. The paper specifically contributes to the ongoing discussion on the challenges of VAs in retail (Grewal et al., 2022; Kamoonpuri and Sengar, 2023), customer motivations for VAs (Kautish and Khare, 2022; Kautish et al., 2023; Khare et al., 2023), brand trust (McLean and Osei-Frimpong, 2019; McLean et al., 2021) and brand loyalty (Huh et al., 2023; McLean et al., 2021). Our findings highlight that in both contexts (Nigeria and the UK), customers are more motivated to use VAs in retailing for their anthropomorphism and convenience. In contrast, user experience highlighted a few challenges: technophobia, data security, suspected espionage, audio bias and disparity. Moreover, the user experience of VAs builds on brand trust and loyalty.

Practically, the findings of our study have important managerial implications for managing VAs in different retailing environments. We discovered that human-like attributes are central to consumer adoption of VAs in retailing contexts (Nigeria and the UK). Additionally, by outlining the flip side of VAs in the retailing contexts, our findings alert retail managers about the potential barriers to enhancing the quality of communication through VAs. Our findings suggest that retail managers should sustain the trust in using VAs while attracting future customers based on the anthropomorphism and convenience attributes of VAs.

Like others, our study also has a few limitations which offer future research directions. First, we explored customer motivations for using VAs in diverse retailing contexts, followed by the challenges of adopting VAs. Finally, we linked the user experience of VAs with brand trust and brand loyalty. However, future research could mainly focus on customer expectations about their experience of VAs. Second, our study focused on retailing in general. However, future research could benefit from the type of products and their value (hedonic vs utilitarian, high vs low involvement products). Third, we stated that the results for both populations were similar, so the data were combined, but we then made a population comparison anyway. We must highlight that this similarity in responses may be because of the selection criteria for participants rather than actual similarities in the broader UK and Nigerian populations. Future research could benefit from doing a broader comparative study of the challenges, outcomes, benefits and prospects of VA usage in two divergent settings. Fourth, we used an exploratory method to explore underlying phenomena. However, future research could use these findings and test the associations using a larger sample and quantitative analyses with various research designs. For example, for customers who are more motivated to shop using VAs, what is the likelihood that they will be satisfied, build trust and remain loyal for a longer period? Finally, our study compared two diverse markets by exploring a cross-market understanding of the user experience of VAs: an emerging market (Nigeria) and a developed market (UK). Future research could bring a more comparative understanding of similarities and differences between such markets with multilevel and multidimensional analysis of user experiences of VAs. In a nutshell, we are confident that our study will motivate future research to develop more generalisable results on user experiences of VAs.