The moderating role of AI chatbots on brand loyalty, customer experience and service quality in online fashion retail

The rise of the internet and advancements in technology have drastically transformed the world, leading to a global shift in how individuals conduct their daily lives (Yeo et al., 2022). The advancement of internet technology has led to the increased use and acceptance of social media as an integral part of everyday life in society. The rapid development of internet technology and social media has allowed consumers to meet their wants and needs through virtual means, driving the accelerating of electronic commerce (Sandunima and Jayasuriya, 2024). The world’s population reaching 7.91 billion by early 2022, and over 58% of the global population engaging on social networks, these platforms have become an ideal space for brands to connect with consumers (Vo Minh et al., 2022). With the increasing presence of social media in society’s daily life, electronic commerce has begun to adapt by incorporating social media into its marketing strategies and sales processes. Ibrahim and Nasr (2025) noted that “many virtual business opportunities have gradually evolved from internet-based trading platforms into social commerce (SC).” According to Kaluarachchi and Nagalingam (2024) SC is a business concept that utilizes social media networks to facilitate business-to-consumer (B2C) and consumer-to-consumer (C2C) e-commerce transactions.

Artificial intelligence (AI) refers to the simulation of human intelligence by machines, particularly computer systems or computer-controlled robots, enabling them to perform tasks typically associated with intelligent behavior (Kang and Choi, 2024). AI is capable of performing various cognitive tasks typically carried out by humans, such as learning, problem-solving and decision making. In recent years, the retail sector has undergone a significant transformation, with online shopping seeing substantial growth. As a result, retailers are faced with navigating a constantly evolving landscape that demands ongoing adaptation and flexibility. The rise of digital technologies has led to a fundamental shift in retail practices, with various touchpoints being integrated to create a seamless Omni channel customer experience (CE). The adoption of AI has significantly accelerated since the COVID-19 pandemic (Ibrahim and Nasr, 2025), as businesses increasingly turned to digital solutions to maintain operations and meet changing consumer demand. The advantages of AI in the fashion industry include enhanced customization, improved customer service and communication, reduced product returns and the ability to automate repetitive tasks and manage inventory more efficiently.

AI chatbots are a type of virtual assistant software designed to engage in conversations with users through text or voice-based interactions (Manzo et al., 2025). Several major fashion brands such as H&M, Tommy Hilfiger, Louis Vuitton, Burberry, American Eagle Outfitters and Levi’s have implemented AI-based chatbots, which have become increasingly sophisticated and influential within the fashion industry (Gooljar et al., 2024). This has made customers more knowledgeable and raised their expectations, demanding more from sellers than what traditional advertising methods could offer. According to Hu et al. (2025), the cost of acquiring new customers is five to ten times higher than that of retaining existing ones, and a 5% increase in customer retention can boost profits by 25% to as much as 95%. Driven by AI and a data-centric approach, involves analyzing previous buying patterns (predictive analytics) and customer sentiments at the time of purchase. This enables businesses to monitor purchasing intentions, leading to more targeted and effective promotion of products and services.

Moreover, AI-powered real time messaging bots for customer service provide firms with valuable opportunities to engage both new and existing customers deliver instant support with tailored solutions and product recommendations and reduce customer abandonment rates and complaints (Gooljar et al., 2024). Furthermore, fashion brands are leveraging AI chatbot applications through social media messaging platforms to support an “online-to-offline” shopping approach, enabling consumers to browse and complete purchases entirely through chat based interaction (Singh et al., 2025). Thus, the use of AI in customer service enhances self-service interfaces by making them more intuitive and interactive, thereby optimizing personalized CEs.

Sri Lanka represents a distinctive context for studying AI-enabled customer services due to several factors. First, digital adoption and familiarity with AI technologies among consumers remain lower than in developed economies, which may influence trust and acceptance of AI chatbots. Second, consumers in Sri Lanka often prefer human interaction in service encounters, making AI-mediated experiences potentially more challenging. Third, infrastructural and technological constraints, such as internet reliability and payment system limitations, may affect the efficiency and perception of AI chatbots. These factors suggest that the effectiveness of AI chatbots in shaping CE and brand loyalty (BL) may differ from findings in developed market contexts.

This study contributes to the literature in several ways. First, although AI-based customer service technologies, particularly chatbots, are increasingly used in online fashion retail, empirical research examining their integrated impact on CE, service quality and behavioral outcomes remains limited. Prior studies have mostly focused on isolated aspects of chatbot functionality, such as usability, responsiveness or communication quality, without providing a comprehensive framework for understanding their influence on BL. Second, existing research has primarily focused on the direct effects of AI-enabled chatbot interactions on customer satisfaction and relationship management outcomes (Chen et al., 2021; Chung et al., 2020; Youn and Jin, 2021), as well as the development of affective or human-like chatbot capabilities (Banchs, 2017). However, relatively little attention has been paid to the moderating role of AI chatbots in the relationships between CE, PSQ and BL. As a result, it remains unclear how and under what conditions AI-driven service interactions strengthen or weaken key behavioral outcomes.

Third, the majority of prior studies have been conducted in developed or technologically advanced economies, implicitly assuming high levels of digital readiness and consumer trust in AI-enabled services. In contrast, developing economy contexts such as Sri Lanka present distinct institutional, technological and cultural conditions that may alter consumers’ responses to AI-based customer services. The limited empirical evidence from such contexts restricts the generalizability of existing theoretical frameworks related to AI-enabled service quality and loyalty formation. By examining Sri Lanka’s online fashion retail sector, this study addresses this gap by extending existing service and technology adoption theories to an understudied context and identifying potential boundary conditions. Finally, by empirically testing the moderating role of AI chatbots, this study offers novel insights into how AI-enabled service mechanisms influence CE and BL in online fashion retail. In doing so, it advances the theoretical understanding of AI-driven service interactions and provides a more nuanced explanation of customer behavior in emerging digital markets.

Uses and Gratifications Theory (UGT) provides a framework for understanding why individuals actively engage with media and technology, emphasizing that users select tools and platforms that satisfy specific functional, informational or emotional needs (Korhan and Ersoy, 2016). In digital service environments, motivations such as convenience, efficiency, problem-solving and enjoyment drive users toward technologies that can deliver the gratifications they seek. Within UGT, the determinants of behavior are conceptualized as “gratifications sought,” while the outcomes of engagement are termed “gratifications obtained” (Kang and Choi, 2024). Motivation thus shapes technology use as well as subsequent evaluations, attitudes and decisions.

Applying UGT to AI-based customer service, such as chatbots, suggests that consumers actively engage with these technologies to fulfill functional and emotional needs, including immediate assistance, personalized guidance and efficient problem resolution (Ray et al., 2019). When these gratifications are attained, users are likely to develop positive perceptions of the service, enhancing CE, strengthening PSQ and ultimately influencing BL and purchase intentions (PIs) (Wei et al., 2024). According to Yu (2024), UGT provides a lens to explain individual differences in technology engagement, as the intensity and type of gratifications sought may vary based on consumer expectations, prior experiences and situational context. In the context of online fashion retail, chatbots that deliver tailored, timely and interactive support are likely to satisfy these user gratifications more effectively than traditional service channels, thereby reinforcing BL and driving behavioral outcomes.

In developing market contexts such as Sri Lanka, the application of UGT is particularly relevant. Lower digital adoption, varying consumer trust in AI and a cultural preference for human interaction may influence the type and intensity of gratifications sought and obtained. Consequently, the effectiveness of AI chatbots in shaping CE and BL may differ from that observed in developed markets. Within this context, AI chatbots are expected to moderate the relationship between service quality, CE and BL, as their ability to satisfy functional and emotional needs may be contingent upon contextual factors such as technological infrastructure and consumer readiness.

By integrating UGT with these contextual considerations, this study provides a theoretical lens to understand how AI-enabled service interactions influence consumer behavior in Sri Lanka’s online fashion retail sector, offering insights into both the mechanisms and boundary conditions of AI chatbot effectiveness.

Author conducted a comprehensive review of the existing literature on chatbots and AI-enabled customer service in online retailing contexts. To conduct this research, the author used specific search terms including “BL”, “CE”, “PSQ” and “AI-based customer service” to identify relevant literature. This initial search on Google Scholar using these terms yielded approximately 52,200 results. The selection was then refined by focusing on peer-reviewed empirical studies published in A and A* journals, as classified by the Australian Business Dean Council (ABDC) journal list, to ensure the inclusion of high-quality and reputable academic sources. Additionally, the author focused on studies that examined chatbots and their drivers as measurable constructs, rather than treating them as vague phrases or simple keywords. The rigorous filtering process led the author to identify five key studies that offer valuable insights into the drivers of chatbots in the context of online fashion retail. These studies were further categorized into “fashion” and “other contexts” based on the service environment in which they explore and analyze chatbots. Studies examining other contexts highlighted CE, PSQ, BL and automated responsiveness as significant predictors of chatbot-related outcomes. In contrast, only one study examined online fashion contexts directly, highlighting a clear contextual gap in the literature.

The rapid adoption of AI-based customer service technologies, particularly chatbots, has attracted growing scholarly attention in the service, marketing and information systems literature. AI chatbots are conversational agents powered by AI and natural language processing that provide real-time assistance, personalized recommendations and problem resolution across the customer journey. Prior research demonstrates that chatbots influence both functional outcomes, such as efficiency and responsiveness, and experiential outcomes, such as enjoyment, trust and emotional engagement (Zhou et al., 2026). These dual effects highlight the importance of understanding AI chatbots not only as operational tools but also as experience-enhancing service interfaces.

A dominant theoretical lens used to explain consumer interaction with AI-based services is Uses and Gratifications Theory (UGT). UGT posits that consumers are active agents who deliberately select technologies that satisfy specific functional, informational and emotional needs. In service contexts, these gratifications include convenience, immediacy, problem-solving, personalization and reassurance (Duy et al., 2026). When these gratifications are successfully obtained, users develop favorable evaluations of the service, which shape their attitudes, loyalty and behavioral intentions. Building on this perspective, AI chatbots can be understood as platforms through which such gratifications are delivered in digital service environments.

Although extensive research examines AI chatbots in banking, hospitality, travel and general e-commerce settings, comparatively fewer studies focus on online fashion retail, particularly in emerging markets. This limitation is important because fashion retail is highly experience-driven and involves higher perceived risk compared to other service contexts (Chong et al., 2026). Fashion retail presents unique challenges due to high product involvement, perceived risk related to fit and quality and the need for interactive guidance (Sandunima and Jayasuriya, 2024). Therefore, the effectiveness of AI chatbots in this context depends not only on functional efficiency but also on their ability to simulate interactive and supportive human-like service experiences. These characteristics make AI chatbots a particularly relevant service interface, while also highlighting the need to examine how CE, PSQ and BL translate into PI within specific cultural and technological contexts such as Sri Lanka.

Despite the growing body of research on AI chatbots, existing studies are largely concentrated in developed service sectors such as banking, hospitality and travel, where digital readiness and trust in automation are relatively high. Accordingly, theoretical generalizability to emerging markets remains limited (Kaluarachchi and Nagalingam, 2024). This focus limits theoretical generalization to contexts where consumers may exhibit lower AI familiarity and stronger preferences for human interaction. Moreover, prior studies tend to emphasize direct effects of chatbot use on satisfaction or adoption, offering limited insight into how AI chatbots interact with experiential and relational constructs to influence behavioral outcomes (Talebi et al., 2026). In particular, limited attention has been given to how AI chatbots shape the relationships between key consumer behavior constructs such as experience, service quality and loyalty. These gaps highlight the need for context-sensitive research that examines AI chatbots as enabling mechanisms rather than standalone service tools.

BL is a central construct in consumer behavior research and reflects a consumer’s emotional attachment, preference and commitment toward a brand, resulting in repeat purchases and positive word-of-mouth (Foroudi et al., 2018). Loyalty is commonly conceptualized as comprising both attitudinal loyalty, which reflects positive feelings and brand preference, and behavioral loyalty, which manifests as repeated purchasing and advocacy (Calvo Porral and Lang, 2015; Wu and Xie, 2026). Together, these dimensions’ highlight that BL is both an emotional and behavioral construct that directly influences consumer decision-making. In the highly competitive online fashion sector, loyal consumers are more likely to repurchase and resist competing offers, even when alternatives are readily available.

In online fashion retail, BL plays a particularly important role due to intense competition, low switching costs and limited physical differentiation between retailers. Loyal consumers are more likely to repurchase, tolerate service failures and resist competitive alternatives. However, much of the existing empirical evidence on BL is drawn from developed market contexts, with limited attention to emerging economies such as Sri Lanka (Singh et al., 2025). Moreover, the role of AI-enabled services, such as chatbots, in fostering BL remains underexplored, representing a significant gap in the literature (Xie et al., 2023). This suggests that existing findings may not fully capture how loyalty is formed in AI-mediated service environments. From a UGT perspective, consumers develop BL when repeated interactions with a brand consistently satisfy their functional and emotional needs (Rehman et al., 2019). AI-enabled services, such as chatbots, contribute to this process by offering efficient support, personalized assistance and continuous availability, reinforcing positive brand associations. In this sense, AI chatbots act as continuous interaction points that reinforce both utilitarian and emotional gratifications. In online fashion retail, repeated gratification obtained through AI-mediated interactions is expected to strengthen brand attachment and, in turn, influence PI.

While BL is well established as a predictor of PI, prior research largely treats loyalty as a stable outcome rather than a construct shaped by evolving AI-mediated service interactions. Additionally, existing studies rarely examine how AI-based customer service conditions the loyalty–purchase relationship, particularly in emerging markets. As a result, it remains unclear whether the strength of the loyalty-purchase relationship varies depending on the effectiveness of AI-enabled service interactions. This limits understanding of whether loyalty operates similarly in contexts where digital trust and service expectations differ, underscoring the need to examine BL within AI-enabled retail environments such as Sri Lanka. Based on this reasoning, the following hypothesis was formulated:

CE refers to the cognitive, emotional and behavioral responses that arise from a consumer’s interactions with a brand across the entire customer journey (Hamouda, 2021). Positive CEs enhance satisfaction, shape favorable brand attitudes, strengthen loyalty and increase PIs (Kjeldsen et al., 2023). In digital environments, CE has become a key differentiator, particularly where products are easily substitutable (Rajesh et al., 2025). This highlights the strategic importance of designing engaging and meaningful interactions in online retail settings.

Experiential marketing emphasizes creating engaging and memorable interactions, which are crucial in differentiating online fashion retailers. A successful retailer depends on delivering a superior CE. The strength of a brand resides in the consumer’s mind and is shaped by their awareness and direct experiences with the brand (Chen and Yang, 2021). Accordingly, CE serves as a key mechanism through which brands build competitive advantage in digital markets (Ibrahim and Nasr, 2025). The comprehensive nature of CE plays a crucial role in identifying and fulfilling individual customer needs (Rehman et al., 2025). As a result, firms increasingly focus on enhancing experiential value alongside product differentiation to influence consumer decision-making.

In online fashion retail, CE is heavily shaped by digital touchpoints, including websites, mobile applications and AI chatbots (Guan et al., 2024). Among these, AI chatbots have emerged as an important interface for real-time and interactive customer engagement. AI chatbots enhance CE by providing instant responses, interactive communication and personalized guidance, which reduce effort and uncertainty during the shopping process (Manzo et al., 2025). These capabilities are especially relevant in fashion contexts, where consumers often seek assistance related to sizing, styling, availability and delivery (Yin et al., 2025). Therefore, AI chatbots contribute not only to functional efficiency but also to the overall experiential quality of online shopping.

According to UGT, consumers engage with technologies to fulfill functional needs, such as convenience and problem-solving, as well as emotional needs, such as enjoyment and feeling valued (Jung and Baloglu, 2025). When AI chatbots successfully deliver these gratifications, they enhance the overall CE, which subsequently influences PI. This suggests that the effectiveness of AI-enabled services depends on their ability to simultaneously satisfy both utilitarian and emotional user needs (Hung et al., 2025). In Sri Lanka, variations in digital literacy, trust in AI and prior exposure to automated services may shape how these gratifications are obtained, making CE a critical explanatory variable.

Although CE has been widely studied in digital retail, much of the existing literature assumes homogeneous consumer responses to AI-enabled touchpoints. However, emerging evidence suggests that consumer responses vary significantly depending on trust in automation and cultural expectations of service interaction (Kathuria et al., 2026). Limited attention has been paid to how cultural norms, digital literacy and trust in automation influence experiential evaluations (Filieri et al., 2026). Consequently, existing findings may not fully capture the variability of CE across different market contexts. As a result, it remains unclear whether AI-enhanced CEs translate into PI in the same way across developed and emerging markets. Addressing this gap is essential to better understand how contextual factors shape experience-driven consumer behavior in AI-mediated environments.

Perceived service quality (PSQ) reflects consumers’ evaluation of how effectively a service meets or exceeds their expectations (Silva et al., 2017). It is widely recognized as a key determinant of satisfaction, trust, loyalty and PI in both traditional and digital service environments (Kang and Choi, 2024). In online retail, service quality perceptions are particularly important due to the absence of face-to-face interaction and heightened perceived risk (Liao et al., 2022). In this context, AI-based customer service, particularly chatbots, plays a critical role in shaping service quality perceptions through consistent, real-time and personalized interactions (Foroudi et al., 2026). AI-based customer service, particularly chatbots, strengthens these dimensions by offering instant responses, consistent information and tailored interactions, effectively fulfilling customers’ functional and emotional gratifications as proposed by UGT. Prior research highlights that improvements in service quality whether through traditional or AI-mediated channels positively influence PIs by enhancing satisfaction, trust and overall CE (Qalati et al., 2021). Thus, service quality can be understood as a key mechanism linking service performance to consumer behavioral outcomes (Hu et al., 2025). Therefore, in online fashion retail, PSQ is expected to be a significant predictor of customers’ PI.

Service quality is commonly conceptualized through dimensions such as reliability, responsiveness, assurance and empathy. These dimensions are interrelated and collectively shape consumers’ overall evaluation of service performance. AI chatbots contribute to these dimensions by delivering consistent information, immediate responses, secure guidance and personalized interactions. Reliability is enhanced through accurate and consistent service delivery, responsiveness through real-time assistance, assurance through credible and secure information provision and empathy through personalized and context-aware communication.

Automated responsiveness, delivered through AI chatbots, refers to the system’s ability to provide prompt, accurate and accessible support to customers in real time (Golalizadeh et al., 2023). In online fashion retail, responsiveness is a critical component of PSQ, as quick and reliable assistance reduces customer effort, alleviates uncertainty and enhances overall satisfaction. Prior research demonstrates that responsive digital services improve customer perceptions of convenience, build trust and increase engagement, ultimately influencing repeat purchases and PIs (Selter et al., 2023). From the perspective of UGT, customers actively seek tools that fulfill functional needs such as efficiency, problem-solving and timely support, which chatbots can deliver more consistently and at scale than human agents. Accordingly, responsiveness enhances not only operational efficiency but also perceived value in AI-enabled service interactions. In the online fashion context, where consumers often have immediate questions about product availability, sizing, styling or delivery, AI-driven responsiveness ensures these needs are addressed swiftly, enhancing CE and strengthening their intention to purchase.

Assurance, or competence, refers to the credibility, expertise and security conveyed by a service provider, which fosters customer trust and confidence (Saxena and Thakur, 2024). In online retail, assurance mechanisms such as secure payment systems, accurate product descriptions and knowledgeable customer support reduce perceived risk and encourage PIs. Prior research highlights that assurance significantly enhances consumer satisfaction and loyalty, particularly in digital environments where uncertainty and perceived risk are higher due to the lack of physical interaction (Mousavizadeh et al., 2016). From a UGT perspective, customers actively seek reassurance and security when interacting with online platforms to fulfill their need for risk reduction and confidence in decision-making. In this regard, AI chatbots function as trust-enhancing mechanisms by providing consistent and reliable information across interactions (Srinivasa Raja, et al., 2026). AI chatbots can enhance assurance by providing accurate, consistent and credible information, guiding users through purchase decisions, verifying product availability and addressing concerns in real time. In the context of online fashion retail, where product quality, sizing and authenticity are critical, assurance delivered via AI chatbots strengthens PSQ and positively influences PIs, making it a key factor in shaping consumer behavior in e-commerce.

Empathy reflects the degree of personalized attention, understanding and care that a service provider conveys to customers, which plays a critical role in shaping satisfaction, loyalty and PIs (Jing et al., 2022). Empathic interactions make customers feel recognized and valued, which enhances their emotional connection to the brand and motivates future purchasing behavior. In online retail, where physical interaction is absent, empathy becomes particularly important for fostering trust and reducing uncertainty. Prior studies have shown that empathetic digital services, such as personalized responses and attentive support, positively influence customer satisfaction, engagement and repeat purchases (Bae et al., 2024). From a UGT perspective, consumers seek services that fulfill their needs for efficiency, reassurance and emotional support. When AI chatbots satisfy these needs, they enhance PSQ, which in turn positively influences PI (Xiong et al., 2026). Thus, empathy represents a critical relational dimension through which AI-enabled services influence consumer behavior. In the Sri Lankan online fashion retail context, infrastructural limitations and cultural expectations regarding service interactions may influence how service quality is perceived, making this relationship particularly salient.

Prior studies generally report a direct relationship between PSQ and PI; however, these findings are predominantly based on traditional or advanced digital service environments (Gupta et al., 2026). However, emerging evidence suggests that this relationship may differ in AI-mediated contexts, particularly in emerging markets. In AI-mediated contexts, particularly in emerging markets, service quality may influence behavior indirectly through experience and relational mechanisms rather than direct evaluation. This indicates that consumers may rely more on experiential and emotional cues when interacting with AI-based services. This suggests a theoretical gap regarding how service quality operates when delivered through AI interfaces, warranting further empirical examination.

AI-Chatbot customer service has become an integral component of online retail strategies, enabling firms to maintain continuous, personalized and scalable interactions with consumers. Empirical research shows that AI chatbots improve response speed, consistency and service availability, while also enhancing customer engagement and satisfaction (Lo Presti et al., 2021; Tan and Liew, 2022). These capabilities highlight the role of AI chatbots as not only service delivery tools but also as mechanisms that shape the overall CE (Zhou et al., 2026). Well-known fashion brands have successfully integrated chatbots to provide product recommendations, order tracking and personalized styling advice, demonstrating the practical relevance of AI-mediated service encounters.

For example, prior studies illustrate how brands such as Tommy Hilfiger have implemented AI chatbots within social media platforms to enhance interactive customer engagement (Yen and Chiang, 2021). Specifically, the chatbot or AI bot engages viewers by promoting them to interact though a time-sensitive greeting, encouraging immediate involvement (Song and Shin, 2024). Through structured questioning and predefined response options, the chatbot identifies customer preferences such as style and size, enabling personalized recommendations (Jiang et al., 2022). Such interactions demonstrate how AI chatbots facilitate guided decision-making and reduce customer effort during the purchase process (Chong et al., 2026). Customers are then redirected to the brand’s official website to complete their purchases. This example illustrates how AI chatbots integrate functional efficiency with interactive engagement to influence consumer behavior.

Empirical evidence highlights that AI chatbots enhance service quality by reducing response times, providing consistent information and handling high volumes of inquiries efficiently (Chen et al., 2022). They also contribute to BL by fostering timely, personalized and engaging interactions, which encourage repeat engagement and strengthen positive perceptions of the brand (Naeem, 2019). Building on this evidence, AI-based customer service can be conceptualized as a mechanism that influences how key antecedents translate into behavioral outcomes. Rather than influencing PI directly, AI-based customer service can shape how other key determinants translate into behavioral outcomes. From a UGT perspective, AI chatbots enhance the extent to which functional, informational and emotional gratifications are obtained during service interactions. Therefore, the effectiveness of AI chatbots determines the extent to which these gratifications strengthen consumer responses. As a result, effective AI-based customer service is expected to strengthen the impact of BL, CE and PSQ on PI.

In emerging markets such as Sri Lanka, where trust in automation and preferences for human interaction may vary, the effectiveness of AI-based customer service becomes particularly important. In such contexts, consumers may rely more heavily on the quality and reliability of AI interactions when forming purchase decisions (Talebi et al., 2026). When AI chatbots are perceived as reliable, responsive and helpful, they are more likely to amplify the positive effects of loyalty, experience and service quality on consumers’ purchase decisions.

Despite growing interest in AI-based customer service, prior research has largely focused on developed markets and treated chatbots as independent predictors of satisfaction or PI. Most studies are descriptive and rarely consider how cultural, technological and relational factors in emerging markets influence consumer behavior (Du et al., 2026). Moreover, limited attention has been given to the moderating role of AI-based customer service in shaping the strength of relationships between key consumer behavior constructs. By examining BL, CE, PSQ and the moderating role of AI chatbots within the Sri Lankan online fashion retail context, the present study addresses these gaps, thereby extending UGT and service quality frameworks by incorporating a context-sensitive and interaction-based perspective.

This study adopted a sequential mixed-methods design, consisting of a qualitative phase (Study 1) followed by a quantitative phase (Study 2), to examine the influence of AI-based customer service on consumer behavior in the Sri Lankan online fashion retail context. This design was chosen to ensure that the quantitative model was grounded in consumers’ lived experiences with AI chatbots rather than being derived from theory alone. The qualitative phase explored consumer motivations and perceptions, while the quantitative phase empirically tested the relationships identified in Study 1. This sequential approach strengthens both theoretical rigor and empirical validity. The integration of qualitative and quantitative findings ensured methodological triangulation and enhanced construct validity.

The qualitative phase employed a purposive sampling technique to recruit 20 college students aged 19–25 (81% female, 19% male) who regularly shop online and have prior experience with AI-based customer services. College students were selected because they are early adopters of digital technologies and frequent users of online retail platforms (Kang and Choi, 2024), making them suitable participants for exploring AI chatbot interactions.

To ensure a shared and realistic service experience, participants first viewed a short demonstration video of Tommy Hilfiger’s AI chatbot and then interacted with Daraz’s AI chatbot for a minimum of five minutes. Following this interaction, participants responded to two open-ended questions focusing on (1) their motivations for using AI chatbots and (2) the perceived advantages and disadvantages of AI-based customer service.

The qualitative data were analyzed using line-by-line coding. Similar responses were grouped into categories and then abstracted into higher-order themes. Two independent coders conducted the analysis, achieving an inter-coder reliability of 95%, which ensured analytical rigor and consistency. The analysis revealed key themes related to functional gratifications (e.g. efficiency, responsiveness), emotional gratifications (e.g. empathy, trust) and informational gratifications (e.g. assurance and accuracy), consistent with the UGT.

Importantly, the findings from Study 1 directly informed the development of Study 2. The identified themes were used to refine construct definitions, justify the proposed hypotheses and adapt measurement items for the quantitative survey. Specifically, the qualitative findings guided the adaptation of measurement scales to ensure contextual relevance to Sri Lankan online fashion consumers. This integration ensured that the conceptual model tested in Study 2 was empirically grounded in consumer experiences within the Sri Lankan context.

The quantitative phase tested the conceptual model using survey data collected from 261 Sri Lankan consumers who had recently engaged in online fashion shopping. Participants were recruited through email invitations and social media platforms. Eligibility was confirmed by prior experience with online fashion retail. Although convenience sampling was employed, efforts were made to ensure diversity in terms of age, gender, occupation, income and online purchase frequency to enhance representativeness.

Before completing the survey, participants viewed the same short video demonstration of Tommy Hilfiger’s AI chatbot and interacted with Daraz’s AI chatbot for at least five minutes. This procedure helped standardize respondents’ recall of AI chatbot experiences and reduced response bias.

The online questionnaire measured PSQ, CE, BL and PI using validated multi-item Likert scales adapted from prior studies (1 = strongly disagree, 7 = strongly agree). PSQ was measured using a 4-dimension scale (reliability, responsiveness, assurance, empathy), adapted from SERVQUAL-based literature (Bhoumik et al., 2026; Foroudi et al., 2026), with 3–4 items per dimension. CE was measured using a 5-item scale capturing cognitive, emotional and behavioral responses (Wan et al., 2026). BL was measured using a 4-item scale covering attitudinal and behavioral loyalty (Lei and Kong, 2026), while PI was measured using a 3-item scale reflecting likelihood of purchase and willingness to recommend. AI-based customer service was measured as a moderating construct using a 3–4 item scale adapted from prior chatbot interaction studies.

All constructs demonstrated satisfactory reliability, with Cronbach’s alpha values exceeding 0.70. Construct validity was assessed using confirmatory factor analysis (CFA). In addition, convergent and discriminant validity were confirmed using AVE and composite reliability (CR), following established SEM guidelines Kaluarachchi and Nagalingam (2024).

The final sample consisted of 54.6% females and 45.4% males, with respondents from Generation Z (15.2%), Generation Y (65.8%) and Generation X (12.4%). Participants represented a range of occupational backgrounds and income levels, and online purchase frequency varied from once a month to once every four months. This diversity reflects a broad cross-section of Sri Lankan online consumers. Table 1 presents these demographic characteristics in detail.

Structural equation modeling (SEM) was employed to test the hypothesized relationships among the constructs. Model fit was assessed using standard indices, including CFI, TLI, RMSEA and SRMR. Multicollinearity diagnostics (VIF) were conducted to ensure that correlations among predictors did not bias the results. The sample size of 261 exceeds the minimum recommended threshold for SEM analysis (e.g. 200 or 10 observations per indicator), ensuring adequate statistical power and model stability (Jayasuriya et al., 2024). The analysis examined the direct effects of BL, CE and PSQ on PI, as well as the moderating effect of AI-based customer service.

Participants reported several functional, informational and experiential gratifications from using AI chatbots in online fashion retail. Functionally, chatbots provided instant responses and facilitated fast navigation, enabling users to quickly access product information without interacting with sales staff. Informationally, chatbots offered useful advice, tailored product recommendations and alternative options based on users’ preferences, supporting decision-making in uncertain or high-involvement purchase situations. Experientially, participants appreciated personalized and interactive experiences, which made online shopping more engaging and enjoyable, akin to having a personal shopper.

Despite these benefits, participants highlighted limitations of AI chatbots. Common concerns included repetitive pre-programmed responses, lack of empathy and impersonal interactions, as well as occasional difficulties understanding how to use the system. These findings suggest that while AI chatbots satisfy core functional and emotional needs, cultural expectations for human-like interaction and trust influence their perceived effectiveness in the Sri Lankan context.

AI chatbots are accessible 24/7 and offer instant responses to customer inquiries. Participants reported that they could get what they needed in a timely manner, without the pressure of dealing with salespeople in offline stores or the hassle of browsing through online stores. Participants indicated they would use AI-based chatbots to conveniently access real-time information from the brand, especially when uncertain about what to purchase.

Participants indicated that they prefer using AI chatbots to quickly locate desired products, rather than manually navigating through an entire website. Participants expressed that they would use AI chatbots to enhance the speed and convenience of their online shopping experience. Some participants noted a preference for avoiding interactions with sales staff regarding their shopping preferences. They also feel hesitant to email or message customer service, as they prefer not to engage directly with a real person. Most participants, therefore, feel more comfortable interacting with the AI chatbot.

Some participants reported that helpful advice and suggestions from AI chatbots assist them in making purchase decisions. Chatbots provide choices based on users’ preferences, so if users have an idea of what they like but are unsure about trends or other options, chatbots can offer guidance and alternatives.

AI chatbots help users find exactly what they’re looking for by allowing them to narrow down searches and choices, making shopping more personalized and easier. Participants said they use AI chatbots for customized, fun and innovative online shopping experiences. They feel the chatbots pay attention to their specific needs, making it feel like having a personal shopper on their computer or phone, which creates a positive shopping experience.

Chatbots offer automated, pre-programmed responses with limited options. While they can provide general guidance, participants noted that sometimes their needs require help only a real person can offer. For more complex or technical questions, participants prefer to contact a live online support agent who can better understand, adapt to and resolve their issues.

Chatbots may lack the depth of interpersonal communication. Participants often feel the interaction is impersonal since they are talking to a programmed robot that only simulates care for their style preferences. Some even feel uncomfortable engaging with a computer, as chatbots lack genuine human-to-human connection.

Some participants reported difficulties in understanding how to use the chatbot. They suggested that providing a clear description or guidance at the start of the chat would be helpful. Additional participant comments are as follows.

1. I feel somewhat hesitant about the chatbot because it collects large amounts of data, and I worry about the potential consequences of this data collection.

2. I can’t think of many positives unless the chatbot is advanced enough to truly understand my questions, if that happens, then it’s useful. Otherwise, it often feels like a waste of time.

3. Sometimes, the recommendations provided by chatbots aren’t helpful, and it can take a while to search and display the results.

CFA was employed as a key analytical technique to rigorously evaluate item loadings, as well as to assess the constructs’ validity and reliability. CFA confirmed the validity and reliability of the measurement model after removing items with low loadings. Model fit indices indicated good fit (χ²/df = 1.254, CFI = 0.979, RMSEA = 0.078), supporting the robustness of the constructs. Harman’s single-factor test showed the largest factor accounted for 45.29% of variance, suggesting that common method bias was not a significant concern.

The initial iteration of the CFA did not produce an acceptable model fit, leading to the removal of items CE5, BL1, BL2, PSQ1, AI1, PI1 and PI5 due to their low factor loadings. Subsequently, a second iteration of the CFA yielded a model fit that satisfied the established thresholds for adequacy as shown in Table 2. Notably, the chi-square to degrees of freedom ratio was 1.254, indicating a commendable model fit and aligning with the recommended threshold of less than 3 (Kaluarachchi and Nagalingam, 2024). Furthermore, the goodness-of-fit indices supported the robustness of the model, with GFI at 0.908, AGFI at 0.901, CFI at 0.979, TLI at 0.960 and NFI at 0.966, all reflecting strong model fit in accordance with benchmarks established in prior research (Fazal-e-Hasan et al., 2025; Kang and Choi, 2024). Additionally, RMR at 0.007 and RMSEA at 0.078 contributed to the model’s robust fit, with PCLOSE values exceeding 0.05, confirming model adequacy. Notably, all factor loadings were highly significant at P < 0.001.

SEM revealed that BL and CE significantly influenced PI, supporting H1 (β = 0.84, p < 0.001) and H2 (β = 0.771, p < 0.001). These results highlight the importance of emotional attachment and positive experiential interactions in shaping consumer purchase decisions. In contrast, PSQ did not directly affect PI (β = −0.039, p = 0.401), indicating that service quality alone may be insufficient to drive purchase behavior in an emerging market without mediation or amplification through AI interactions. AI-based customer service had a significant positive effect on PI (β = 0.332, p < 0.001), underscoring the role of chatbots in facilitating behavioral outcomes.

Harman’s single-factor test (Kang and Choi, 2024) was conducted to assess common method bias and the extent of method variance in the dataset. As all variables were self-reported, the potential for common method bias existed (Adam et al., 2021). The unrotated principal components analysis revealed that the largest single factor accounted for 45.29% of the total variance, which is below the recommended 50% threshold (Kang and Choi, 2024). Therefore, common method bias does not appear to be a significant concern in this study.

The structural equation model was tested using the statistical software AMOS version 26. The results are summarized in Table 3. The structural model demonstrated a satisfactory fit with the data, as indicated by χ² = 915.54 with 350 degrees of freedom, χ²/df = 1.254, CFI = 0.979, NFI = 0.966, GFI = 0.908, RMSEA = 0.078 and RMR = 0.007.

Regarding H1 and H2, the results showed that BL had a significant positive influence on PI (β = 0.84, p < 0.001), and CE positively impacted PI (β = 0.771, p < 0.001).

BL is composed of various elements, including customer motivations, decision-making processes, repurchase behavior and brand preferences (Foroudi et al., 2018). BL positively influences PI, as loyal customers are more likely to choose and repurchase from brands they trust and feel emotionally connected to, thereby reducing decision-making time and increasing the likelihood of future purchases.

CE significantly influences PI in the online fashion industry. When shoppers engage with intuitive, responsive and personalized online platforms such as AI-powered chatbots they are more likely to develop positive perceptions of the brand, which directly impacts their intent to purchase (Jung and Baloglu, 2025). Features like real-time product recommendations, easy navigation and instant support help recreate the in-store experience digitally, reducing uncertainty and enhancing satisfaction. Personalized experiences that cater to individual preferences make customers feel valued, increasing the likelihood of conversion. These findings indicate that in online fashion retail, behavioral intention is primarily driven by relational and experiential factors rather than purely functional evaluations of service quality.

Regarding H3, PSQ had a negative and statistically insignificant influence on PI (β = −0.039, p = 0.401). This unexpected finding suggests that PSQ alone is not a sufficient direct predictor of PI in the Sri Lankan online fashion retail context.

One possible explanation is that consumers may perceive service quality as a basic expectation rather than a differentiating factor in digital environments, particularly where AI chatbots standardize service delivery. Another explanation is that the effect of PSQ is likely indirect, operating through experiential and relational constructs such as CE and BL rather than directly influencing PI. This aligns with the moderation results of AI-based customer service, which indicate that chatbot interactions may amplify the influence of service quality when embedded in experiential contexts rather than functioning as an isolated determinant.

AI-based customer service proved effective by exerting a positive and statistically significant impact on PI (β = 0.332, p < 0.001). This highlights the importance of AI chatbots as enabling mechanisms that enhance the translation of service perceptions into behavioral outcomes.

To address hypothesis H4, this study examined the moderating effect of AI-based customer service (chatbot) on the relationship between brand-related factors and customer PI in the online fashion retail context. Moderation analysis was conducted using interaction terms (BL × AI, CE × AI and PSQ × AI) to assess whether the effect of independent variables on PI changes depending on the level of AI chatbot effectiveness.

As presented in Table 4, the interaction between BL and AI-based customer service (BL × AI) is statistically significant, indicating that AI chatbot usage strengthens the relationship between BL and PI. This suggests that loyal customers are more likely to convert into actual purchasers when they experience enhanced engagement and support through AI chatbots, which reinforce brand relationships and reduce decision uncertainty.

The results of this study, as presented in Table 5, provide insights into the moderating effect of chatbot usage on the relationship between CE and PI. The interaction effect between CE and AI chatbot usage (CE*AI) is statistically significant, indicating that AI chatbots strengthen the relationship between CE and PI. This implies that when customers experience higher levels of interactive, responsive and personalized chatbot service, their positive experiences translate more strongly into PI. In other words, AI chatbots enhance the emotional and functional value of CE, leading to stronger behavioral outcomes.

The results from the model, as presented in Table 6, highlight the moderating effect of chatbot usage on the relationship between PSQ and PI. The interaction effect between PSQ and chatbot usage (PSQ*AI) is statistically significant, indicating that chatbot presence moderates the relationship between PSQ and PI. This suggests that service quality perceptions alone may not directly translate into PI; however, when AI chatbots are present, they enhance the effectiveness of service quality by improving responsiveness, accessibility and perceived support.

Interestingly, the direct effect of AI-based customer service on PI was not statistically significant in this model, which further supports the argument that chatbots function primarily as enabling or amplifying mechanisms rather than direct drivers of purchase behavior. Instead, their value lies in strengthening existing psychological and experiential factors such as loyalty, experience and PSQ.

The findings confirm that AI-based customer service plays a critical moderating role in online fashion retail by enhancing the strength of key psychological and service-related predictors of PI. This supports the theoretical position that AI chatbots function as contextual enablers that amplify consumer responses rather than acting as standalone determinants of behavior.

This study provides theoretical and contextual insights into the role of AI-based customer service in online fashion retail, with particular emphasis on an emerging market context. Guided by UGT, the qualitative findings (Study 1) reveal that consumers engage with AI chatbots to satisfy functional, informational and emotional needs, including convenience, immediacy and personalized assistance. These results are consistent with prior studies conducted in developed markets, which emphasize responsiveness and personalization as central drivers of positive digital service experiences (Jung and Baloglu, 2025; Selter et al., 2023).

However, this study extends existing research by showing that Sri Lankan consumers simultaneously value efficiency while expressing concerns about limited empathy, repetitive responses and the lack of human interaction. This indicates that AI chatbot acceptance is still in a transitional stage in emerging markets, where consumers compare automated service experiences against traditional human service expectations. This tension suggests that chatbot evaluations in emerging markets are shaped not only by technological performance but also by culturally embedded expectations of interpersonal service (Ray et al., 2019). Unlike consumers in developed economies where AI-enabled self-service is widely normalized Sri Lankan consumers appear to assess chatbot effectiveness through a hybrid lens that combines functional efficiency with expectations of human-like engagement (Naeem, 2019). This finding highlights the importance of contextual sensitivity when applying AI service theories across different institutional and cultural settings. These findings are consistent with prior research in developed markets, which highlights the importance of responsiveness and personalization in digital service encounters.

The quantitative findings (Study 2) demonstrate that BL and CE have strong and significant direct effects on PI. These results align with established literature in online fashion retail, confirming that emotional attachment and experiential satisfaction remain critical drivers of consumer behavior (Foroudi et al., 2018; Kjeldsen et al., 2023; Rehman et al., 2019). Notably, the strength of these effects in the Sri Lankan context suggests that relational and experiential factors may outweigh purely functional evaluations, particularly in markets where trust in AI-enabled services is still developing (Silva et al., 2017). This implies that purchase decisions are more strongly influenced by trust-building and relationship-oriented interactions than by efficiency alone.

In contrast, PSQ does not exert a significant direct effect on PI, diverging from findings commonly reported in developed-market studies (e.g. Tan and Liew, 2022; Lo Presti et al, 2021). This non-significant result suggests that PSQ functions as a baseline expectation rather than a motivational driver in AI-mediated retail environments. In other words, consumers may assume a minimum acceptable level of service quality, making it insufficient on its own to influence purchase decisions.

A further explanation is that in AI-based service contexts, consumers do not evaluate service quality in isolation; instead, they translate service quality perceptions into behavioral intentions only when these perceptions are reinforced by positive emotional or experiential outcomes such as CE and BL. This result suggests a boundary condition for traditional service quality theories in AI-mediated contexts. In emerging markets, service quality attributes such as reliability or responsiveness may not independently trigger purchasing behavior unless they are translated into positive experiences and relational value. In other words, consumers may recognize service quality cognitively, but this recognition influences behavior only when mediated by experiential or emotional mechanisms.

A key contribution of this study lies in identifying AI-based customer service as a moderating mechanism rather than a direct predictor. The findings show that effective chatbot interactions strengthen the relationships between BL, CE, PSQ and PI. This indicates that AI chatbots act as “contextual amplifiers,” meaning they do not directly create PI but enhance the strength of existing psychological and relational drivers. Specifically, when chatbot systems provide real-time support, personalization and interactive engagement, they reinforce customers’ existing loyalty and positive experiences, thereby increasing the likelihood that these perceptions translate into actual PI s.

In the Sri Lankan context characterized by uneven digital adoption and strong preferences for human interaction, AI chatbots enhance PI only when they deliver both functional value and personalized, engaging interactions.

This study advances theory in three important ways. First, it extends UGT to AI-mediated service environments in a developing economy, demonstrating that gratifications are contingent on local technological readiness and cultural expectations. Second, it conceptualizes AI chatbots as moderators that shape how loyalty, experience and service quality translate into behavior, rather than as standalone drivers of PI. Third, it identifies boundary conditions for traditional service quality frameworks, showing that functional attributes alone are insufficient without experiential and relational reinforcement. Together, these insights refine existing theories of digital service quality and UGT, offering a more context-sensitive understanding of AI-driven service interactions beyond developed markets.

This study makes several important theoretical contributions to the literature on AI-mediated services and online retailing. First, it extends UGT by demonstrating that gratifications derived from AI chatbots in online fashion retail encompass not only functional convenience but also emotional and relational dimensions. The findings show that consumers seek immediacy, convenience, personalized assistance and interactive engagement, and that these gratifications are closely intertwined with service expectations traditionally associated with human–brand interactions. By situating AI chatbot use within an emerging market context, the study advances UGT beyond technology adoption and highlights the role of cultural and technological conditions in shaping AI-enabled service experiences.

Second, the quantitative findings refine service quality and digital service frameworks by revealing that responsiveness and empathy enhance CE, which in turn strengthens BL and PI. This indicates that, in AI-driven service environments, PSQ influences consumer behavior primarily through experiential and relational pathways rather than direct functional evaluations. By integrating cognitive evaluations of service quality with affective responses such as experience and loyalty, the study provides a more nuanced explanation of consumer behavior in digital retail contexts. Furthermore, the findings highlight that AI-enabled service encounters are no longer purely transactional but are increasingly relational, where emotional engagement plays a decisive role in shaping consumer decision-making.

From a managerial perspective, the findings offer clear and actionable guidance for online fashion retailers and digital service designers. Consumers expect AI chatbots to deliver personalized, interactive and convenient support throughout the customer journey. Retailers should prioritize chatbot designs that provide tailored product recommendations, real-time assistance and seamless navigation across pre-purchase, purchase and post-purchase stages. In addition, embedding empathetic attributes through adaptive dialogue, personalized responses and context-aware interactions is critical for fostering emotional engagement and BL. Investments in conversational design, sentiment recognition and hybrid AI–human service models can help balance efficiency with reassurance, particularly in emerging markets where trust in AI technologies is still developing. These strategies can also generate tangible commercial benefits, including increased conversion rates, improved customer retention and enhanced competitive advantage in the rapidly growing online fashion retail sector.

Beyond managerial implications, this study also highlights broader societal, teaching and policy relevance. From a societal perspective, effective and empathetic AI-based customer services can enhance consumer trust, reduce uncertainty and improve the overall quality of online shopping experiences. This is particularly significant in emerging economies such as Sri Lanka, where digital commerce adoption is still evolving and consumer trust in AI-enabled systems remains relatively fragile. In teaching and professional training contexts, the findings offer valuable insights for courses in digital marketing, service management and AI-enabled CE, illustrating how theoretical frameworks such as UGT can be applied to real-world service design. From a policy perspective, the results underscore the importance of developing guidelines that promote transparency, fairness and consumer protection in AI-driven service environments, particularly in emerging markets. Policymakers should also consider establishing ethical standards for AI chatbot deployment, particularly regarding data privacy, algorithmic transparency and accountability in automated customer interactions.

The study demonstrates that AI chatbots function not only as technological service tools but also as enablers of experiential and relational value, effectively bridging theory and practice in the evolving landscape of AI-enabled retail services.

This study has several limitations that should be acknowledged. First, the research focused exclusively on the online fashion retail context, which may limit the generalizability of the findings to other product categories. Consumer interactions with AI chatbots may differ across domains such as electronics, home goods or personal care due to variations in product involvement, perceived risk and decision complexity. Future research should therefore examine AI-based customer service across a wider range of retail contexts to assess the robustness of the proposed relationships.

Second, the study relied on purposive and convenience sampling, with the sample consisting primarily of younger consumers and students. While this group represents an important segment of online fashion shoppers, younger users typically exhibit higher digital literacy and greater familiarity with AI technologies, which may influence their expectations and evaluations of chatbot services. As a result, the findings may not fully reflect the perceptions of older consumers or those with lower levels of technological experience. Future studies should employ more diverse and representative samples across age groups, occupations and income levels to enhance external validity.

Third, the study examined interactions with a limited number of chatbot exemplars. Although these chatbots were selected to reflect commonly used AI-based customer service features, differences in chatbot design, intelligence and functionality may influence consumer responses. Future research could incorporate a broader range of chatbot types or compare multiple AI service designs to better capture variability in AI-enabled service experiences.

Finally, the data were collected at a single point in time, limiting the ability to capture changes in consumer perceptions as AI technologies continue to evolve. Given the rapid advancement of chatbot capabilities, longitudinal research designs are recommended to examine how CE, PSQ, BL and PI develop over time with increased exposure to AI-based customer services.

This study was reviewed and approved by the Ethics Committee of the Sri Lanka Institute of Information Technology, Business School.