Humanizing the customer experience with AI chatbots: a study in the food services industry toward achieving SDG 11 and SDG 12 Open Access

Over the last decade, Artificial Intelligence (AI) has emerged as a transformative force, revolutionizing the way companies and service providers enhance their marketing strategies and overall performance. Among the various branches of AI, generative AI has gained particular prominence for its ability to create new content – text, images, audio and videos – by leveraging advanced machine learning algorithms (Feuerriegel et al., 2024). Specifically, Natural Language Processing (NLP) allows these AI systems to understand and respond to user inquiries in a conversational manner, making them more intuitive and user-friendly (Yan et al., 2024).

In the increasingly competitive food delivery market, maintaining exceptional customer support is crucial for ensuring user satisfaction. Traditional methods, such as prolonged hold times and repetitive call center interactions, have become less effective, leading to a shift toward digital customer support solutions. Machine learning-powered chatbots have emerged as a key innovation in this area. These intelligent virtual assistants utilize NLP to recognize the context, intent and sentiment of customer inquiries, enabling more personalized and efficient interactions (Kolasani, 2023). Chatbots also enhance transparency by providing real-time updates on order statuses, delivery times and unforeseen delays, ensuring a seamless experience for users of food delivery apps.

These emerging technologies allow for the exchange of text and other media through hosting platforms, enhancing customer service through interactive experiences with dynamic, humanoid content available anytime and anywhere (Smith et al., 2024). By introducing chatbots, businesses aim to provide more natural and accessible information, thereby enhancing the content and richness of user interactions.

Across various industries, brands are increasingly transitioning from traditional customer service models to digital chatbot solutions. This trend has influenced the hospitality sector, where AI-based technologies and mainly robotics and are being adopted to improve strategies, customer relationships and overall experiences (Limna, 2023). Hotels and restaurants, for example, can now handle customer inquiries more effectively, provide tailored recommendations and fulfill customer needs and expectations through chatbots (Ukpabi et al., 2019). A notable example is Domino's, which uses a Twitter chatbot to streamline pizza orders (Cheng and Jiang, 2021). Chatbots today can manage reservations, take orders, promote deals, offer recommendations, facilitate feedback processing and simplify delivery and takeaway services (Susanto et al., 2024).

Chatbots share operational advantages with other digital ordering methods, including enhanced ROI, advanced productivity and decreased labor costs for restaurant managers. They significantly enhance customer experiences by anticipating needs, delivering quick and personalized responses, offering fast ordering assistance and streamlining payment processes (De Cicco et al., 2021). Additionally, chatbots guide customers through interactions, provide helpful information, support decision-making and reduce order errors.

Since the introduction of chatbots into the restaurant sector, numerous studies have focused on various aspects of their impact, including restaurant performance, service quality, customer comfort and satisfaction, chatbot trust and perceived enjoyment, privacy concerns, chatbot appeal, control, perceived self-efficacy and personalization, human resemblance, perceived efficacy, information overload, user intention, ease of use and perceived usefulness (Chung et al., 2020; Ashfaq et al., 2020; Xiao et al., 2020).

Chatbot efficiency (CE), defined as the ability to handle user queries promptly, accurately and seamlessly, is critical to improving the customer experience (Nwokedi and Nwafor, 2024). Efficient chatbots reduce response times, leading to better user experiences, particularly when customers have urgent inquiries or issues. This efficiency contributes to enhanced support, time savings, cost reduction and better decision-making (Aslam, 2023). Traditional customer support methods often resulted in long hold times and frustrated customers, but the transition to digital support – and specifically to chatbots – has greatly improved accessibility.

Chatbots provide seamless communication channels, reducing friction for users. They analyze user profiles and data to deliver tailored assistance, including personalized recommendations, order suggestions and relevant information. This personal touch creates a sense of familiarity and friendliness, making users feel like they are interacting with a helpful companion rather than a robotic system. A friendly chatbot can turn a potentially frustrating situation into a positive experience, enhancing satisfaction (Shams et al., 2024; Chatterjee and Karmakar, 2023). Moreover, chatbots that acknowledge user emotions and respond empathetically build trust and make users feel understood. The perceived hospitality (PH) of chatbots, characterized by their welcoming demeanor, friendliness and ability to create a sense of familiarity, has the ability to modernise consumer experiences in the food industry (Paliszkiewicz and Cusumano et al., 2023).

Perceived authenticity (PA) refers to the “notions of realness, trueness to origin, uniqueness, and originality” (Fritz et al., 2017). It is fundamental to the experience of sharing the essence of an item and its true outcome. Research indicates that attributing human characteristics to non-human entities, known as anthropomorphism, can significantly impact consumer satisfaction (Xie et al., 2023). When chatbots display human-like traits, such as using natural language, understanding context and showing empathy, they create more engaging and personalized interactions (Jiang et al., 2023). This PA fosters positive experiences by making customers feel heard and understood. Moreover, PA builds trust. Consumers are more likely to trust a chatbot that behaves like a knowledgeable, helpful human (Pizzi et al., 2023). By providing accurate information and resolving queries promptly, chatbots establish credibility. Mimicking human conversation makes interactions smoother and more efficient, which enhances customer convenience and increases the likelihood of repeat orders. Thus, PA in chatbots represents a significant advancement in personalizing service in the food sector, leading to greater acceptance of these human-like tools.

This study also aligns with SDG 11 by highlighting the potential of AI chatbots to foster sustainable urban service delivery. As cities grow denser and the demand for food services increases, integrating intelligent chatbot technologies into food delivery systems enhances operational efficiency, reduces errors in placing and contributes to smarter, more resilient urban infrastructures (Rane et al., 2024). By automating and optimizing food ordering, AI chatbots support inclusive and sustainable urban experiences by reducing systemic inefficiencies, minimizing overcrowding at physical locations and mitigating resource-intensive service models. Chatbots contribute to the digital infrastructure necessary for smart urban environments through enhancing urban resilience, promoting service accessibility and facilitating sustainable consumption habits aligned with SDG 11 (Das, 2024).

Despite recent advancements in chatbot research, there remains a gap in the literature concerning aspects such as CE, PH and attitude toward PA, which are still relatively underexplored (Yilmaz et al., 2024). Additionally, limited research has examined the importance of CE on consumer behavior, experience and the willingness to reuse chatbots for future orders. Similarly, while PH and chatbot friendliness are recognized as important, few studies have linked personalized and friendly chatbot interactions to positive attitudes and satisfaction. Although PA is crucial in shaping users' perceptions of chatbots, with users tending to value those that exhibit human-like qualities and provide authentic interactions, there is a scarcity of research measuring the impact of PA on consumer experience. Therefore, further studies on these aspects of chatbots, particularly in the context of food ordering, are essential. As more food delivery services adopt chatbots for customer service, exploring these new dimensions and their interrelationships is critical. To address this research gap, it is valuable to study CE, PH and attitude toward PA and link these factors to other variables, such as positive impression (PI), satisfaction and acceptance.

Based on the foregoing, we formulated the following research questions (RQ):

To address the gap in the research, the paper is structured as follows: Section 2 outlines the theoretical background and presents the conceptual framework underlying the study. Section 3 details the methodology, including study design, measurement instruments and data collection. Additionally, Section 4 presents the empirical results of the analysis, followed by Section 5, which offers a discussion of the findings and their theoretical and managerial implications. Moreover, Section 6 outlines the study's limitations and proposes directions for future research. Finally, Section 7 concludes the paper by summarizing key insights and contributions to theory and practice.

The advancement of AI tools has catalyzed enhancements across diverse sectors of industry. One notable application among these advancements is generative AI, which relies on language comprehension and question-answering capabilities (Sengar et al., 2024). Moreover, generative AI demonstrates the ability to comprehend the context of customers' queries, enabling the generation of personalized responses (Kumar et al., 2025).

Numerous organizations are investigating methods to identify and leverage the capabilities of Generative AI and comparable technologies within their operations, aiming to improve performance, satisfy customers and ensure sustainable growth (Agrawal et al., 2023). Companies such as DeepMind, Google Cloud and Walmart have discussed that generative AI tools, including chatbots, can enhance customer service (Kalota, 2024; Razzaq and Shah, 2025). This is due to their ability to analyze customer data and understand the context of each inquiry through NLP techniques, enabling the creation of personalized responses.

In their study, Gupta et al. (2024) discussed how chatbots, a type of generative AI tool, have been adopted by businesses to facilitate communication with customers through numerous channels, such as websites and messaging applications. They also defined chatbots as user-friendly applications that respond to clients' requests in a significant and engaging way. This suggests that users can interact with chatbots in an innate and human-like approach, simplifying the process of obtaining information and completing tasks.

Humanness is a key aspect of human–AI interaction, ultimately aiming for AI to be perceived by humans as comparable to a human being (Cheng et al., 2022). This chatbot's humanness is the object of a theoretical debate opposing two research streams. While some studies highlight chatbots' advantages and successes, others focus on consumers' reluctance to use them.

The first research stream explains that customer comfort and satisfaction levels while chatbots increase while chatbots' language becomes more human-like (Schanke et al., 2021). Chatbot humanness is an important property affecting consumer perceptions and attitudes (Al Shafei, 2025). The more the chatbot is human-like, the more it generates acceptance, positive attitudes, satisfaction, behavior and intention outcomes (Mende et al., 2019; Al Shafei, 2025). Chatbot humanness is crucial in creating a customer relationship as it influences consumer engagement and loyalty to the brand.

The second research mainstream shows that, despite advancements in AI-driven chatbots and their sophisticated capabilities, many consumers continue to favor interactions with real humans over chatbots that imitate human behavior. Many online consumer service users prefer to communicate with a human instead of an AI chatbot (Haugeland et al., 2022). Many consumers believe that a human would have a better understanding of their needs and prefer to engage with a live representative, especially when their inquiries become more complex (Pham Thi and Duong, 2025). In addition, several studies did not support the positive attitude toward the human aspect of AI chatbot services. Consumers remain more open, comfortable and pleasant when interacting with humans than chatbots (Jin et al., 2025a, b). They also tend to have fewer privacy concerns. Moreover, computer-illiterate consumers are usually reluctant to use chatbots even when their aspect is similar to human beings (Jin et al., 2025a, b). On top of that, some recent essays and writings present AI chatbots as a source of dehumanizing services and open the path to a more profound theoretical debate about AI, chatbots and their marketing applications (El Bakkouri et al., 2022).

Beyond technical performance and human-like interaction, the integration of chatbots into food service delivery reflects broader sustainability imperatives (Hwang et al., 2025). Specifically, the deployment of intelligent agents aligns with SDG 11 by supporting the optimization of urban food logistics and minimizing overcrowding at service touchpoints (Mohsen, 2024). In dense urban environments, chatbots contribute to smart service that aims at enhancing responsiveness and promoting equitable access to food. Through digital infrastructure and automation, cities become more inclusive and resilient. Moreover, chatbots play a vital role in advancing SDG 12 by fostering transparency, reducing resource waste and enhancing order accuracy (Ziemba et al., 2024). Specifically, digital ordering systems equipped with AI reduce overproduction and streamline inventory by capturing precise customer preferences. This reduces food waste and encourages behavior that aligns with sustainable consumption. In the context of this study, chatbot-mediated interactions are not only efficient but also instrumental in nudging consumers toward more ethical and environmentally conscious choices in food services.

In the following developments, we propose a theoretical model liable to bring insights and highlights while contributing to this emerging theoretical debate on using human-like chatbots in the specific context of restaurants, while aligning with SDG11 and SDG12.

In frontline service contexts, efficiency is termed as the capability to deliver services effectively and in a manner that is perceived as efficient by consumers (Chong et al., 2021). In today's landscape, restaurants and other hospitality businesses increasingly rely on AI chatbots to enhance frontline efficiency (Alawami et al., 2025). CE plays a critical role in various activities, including facilitating human–chatbot interactions, handling customer inquiries, providing assistance, resolving issues and addressing problems autonomously without human intervention (Al Shafei, 2025; Chakraborty et al., 2025; Alhashmi et al., 2025). Chatbots, equipped with strong analytical capabilities and the ability to interact efficiently with users, are particularly effective in managing complex service requests (Pantano and Pizzi, 2020). In relevance to sustainability, Mutambara (2025) and Bathla et al. (2023) discuss how efficient chatbot systems reduce operational strain and optimize digital service delivery, supporting SDG 11 by enhancing urban infrastructure responsiveness.

While there has been substantial research on advanced chatbot features such as trust, enjoyment, appeal, personalization, interaction, human resemblance, ease of use and user intention, gaps remain in understanding CE. While the efficiency of human employees has been extensively studied (Yu et al., 2020), research focusing on virtual AI-based employees, such as chatbots, is limited. As a result, the relationship between CE and other relevant factors is not well understood. To address this gap, this research introduces a new framework that connects CE with key concepts like PH and PA.

In the food service industry, PH is a significant concept. It encompasses attributes such as friendliness, cheerfulness, politeness, responsiveness, patience, communication skills, positivity and the ability to encounter customer needs and expectations. PH is a crucial factor for success in industries like hospitality, tourism and food service, as it significantly enhances customer satisfaction (Sim et al., 2006). It also serves as a key indicator of customer loyalty and satisfaction, being closely linked to perceived comfort, empathy and revisit intention (Khairunisa and Melani, 2025). Pijls et al. (2017) identify three primary dimensions of PH: inviting (openness, appeal, freedom), care (servitude, empathy and acknowledgment) and comfort (ease, relaxation and comfort). Their findings suggest that these dimensions enhance PH and enhance the overall customer experience. Concerning sustainability, Leal Filho et al. (2024) discuss that hospitality cues in chatbots foster inclusive service experiences, contributing to SDG 11's goal of accessible urban services and social cohesion.

Moreover, Srivastava et al. (2020) propose that PH is shaped by four key attributes: modifiability, security/privacy, interoperability and reliability. Their study introduces a PH framework that generates an index based on these attributes, revealing that chatbot-PH is contingent upon these factors. Given these insights, the first hypothesis of this study is focused on the role of PH.

PA defines a tool's honesty, sincerity and commitment to being true to others, encompassing realness, trueness, uniqueness and originality (Fritz et al., 2017). Ma et al. (2025) conclude that anthropomorphic design cues in chatbots significantly elevate users' perceptions of humanness and sincerity, driving deeper engagement and trust. McGuire et al. (2023) demonstrate that transparency features such as revealing decision logic reduce perceived deception and bolster authenticity judgments. Jia et al. (2025) find that consistent, error-free chatbot performance in hospitality contexts enhances continuity and reliability, thereby reinforcing authenticity and improving satisfaction and loyalty.

In the food service industry, PA involves interactions between the object, individuals and society. When applied to AI-based platforms and robots, PA becomes an important factor for brand success. Rodrigues et al. (2022) identify brand PA based on originality, naturalness, reliability and continuity, demonstrating its positive relationship with marketing constructs like brand involvement, brand image and brand satisfaction. According to Nguyen et al. (2025), authentic chatbot interactions promote trust and transparency, aligning with SDG 12 by encouraging responsible digital consumption behaviors.

In the context of chatbots, PA is defined by characteristics such as transparency, anthropomorphism, humanness and coherence (Kumar et al., 2025). These characteristics enhance the quality of chatbots and their interactions with customers. Moreover, chatbots that leverage real-time personalization, tailoring menu suggestions and responses based on individual user preferences have been shown to elevate perceptions of sincerity and uniqueness, as personalized interactions convey attentiveness and reinforce the chatbot's commitment to each customer's needs (Rane et al., 2024). Additionally, systems that consistently recall past orders and remembered preferences across sessions foster a sense of continuity and reliability, further strengthening users' judgments of authenticity and long-term relational trust. Marjerison et al. (2025) further investigate the link between chatbot PA and the quality of chatbot–human interaction, finding that higher PA leads to better interaction quality. Based on these insights, we suggest the following hypothesis:

PI in business and management is the audience's perception of a person, object, idea, situation or event (Schlenker, 1980). In the food service industry, it relates to consumers' impressions of the business, its offerings and its communication strategies. Creating a PI is crucial for establishing strong customer relationships. Nyamekye et al. (2023) show that restaurant atmospherics and PH can enhance customer experiences and overall PIs. Services and PH in restaurants also lead to better customer first impressions, perceptions and behaviors (Khetjenkarn and Agmapisarn, 2025).

Building on this, recent empirical work highlights how specific service cues translate into PIs in automated contexts. Lee and Eastin (2021) demonstrate that personalized greetings, such as addressing customers by name and referencing previous orders, boost favorability scores as they convey attentiveness and care (Blümel, 2024). Li et al. (2024) find that adaptive response timing tailoring messages to match customers' interests, enhances customers' perceptions of warmth and responsiveness and leads to stronger first-impression ratings. Moreover, Onat et al. (2025) report that consistent empathetic language using friendly, conversational phrasing across all messages reduces feelings of impersonality and increases likability.

Relevant to chatbots, PIs significantly influence customer experiences. Chatbots that exhibit emotional realism are more likeable and leave better impressions on customers (Kim et al., 2025). Additionally, Chauhan and Mehra (2025) find that the language used by chatbots, including typos and capitalized terms, can affect customer impressions. Thus, we propose the following hypothesis:

From another perspective, PA also influences PIs. Grandey et al. (2005) demonstrate that the PA of employee expressions impacts customers' impressions of friendliness. Deng et al. (2025) reveal that brand PA is a significant factor influencing consumer impressions. Similarly, Markowitz et al. (2022) find that verbal PA leads to PIs, with consumers forming favorable opinions of those they perceive as authentic.

In the context of chatbots, Yan et al. (2025) study how the warmth of an initial chatbot message can shape consumer impressions, establishing a direct relationship between PA and PI. Similarly, in their study across multiple restaurant chatbots, Nguyen et al. (2022) find that those incorporating brief, personalized explanations of menu suggestions boosted PI compared to generic recommendation bots. Moreover, a recent study by Niros et al. (2025) discuss how chatbot offering detailed customization and tailoring orders enhance perceived coherence, which in turn amplifies the overall impression. When users perceive a chatbot as authentically attentive to their individual preferences and transparent about its reasoning, they not only judge the interaction more positively but also report greater willingness to recommend the service to others. These findings reveal that PA is a critical antecedent of PI in AI-mediated food ordering scenarios. In relevance to these conclusions, we imply the following hypothesis:

Consumer attitude is generally defined as a feeling of positivity or negativity that a consumer has toward an object, brand/product, idea, company or marketing practice (Sani and Gbadamosi, 2025). A positive consumer attitude increases the likelihood of purchasing or using a product. In the food industry, a brand often acts as a “guarantee of perceived authenticity,” serving as a significant predictor of consumer attitudes toward food consumption and positive motivations (Maison et al., 2004). Proxies of PA, such as perceived availability, perceived consumer effectiveness and certainty, positively impact attitudes toward products and purchase intentions (Yang et al., 2025).

The PA of chatbots in deliberative discussions can also serve as an indicator of a consumer's positive attitude toward a brand or product (Kim et al., 2021). Al-Oraini (2025) provides empirical evidence that chatbot conversation PA, convenience and enjoyment lead to a positive consumer attitude toward chatbots. For instance, Maar et al. (2023) discuss that chatbots offering personalized, context-aware explanations for their recommendations, such as detailing why a particular dish complements a user's order elevates perceived reliability, strengthen consumer attitude. Moreover, Al -Shafei (2025) finds that sustained authenticity signals, including consistent use of usernames and adaptive empathy, foster deeper psychological engagement, resulting in positive attitudes and subsequent purchase intentions. Recent evidence showed that combining authenticity cues with warmth and empathy, such as consistent use of personalized greetings and acknowledgment of user preferences, increases relational quality and elevates first-impression rating, resulting in a positive attitude toward chatbot (Li et al., 2024). Accordingly, we propose:

On another level, literature suggests that a PI could positively influence consumers' attitudes toward a chatbot. Zhang and Wang (2025) investigated the direct relationship between PIs and attitudes toward virtual agents, showing that favorable PIs may help engage consumers in interactions and enhance their attitudes, leading to long-term acceptance. Dassouli et al. (2025) similarly show that consumers' first impressions, alongside perceived ease of use and usefulness, positively impact attitudes toward mobile messenger chatbots. Roy and Naidoo (2025) further confirm that better impression, such as perceived warmth, responsiveness and likability, positively impact brand attitudes toward chatbots. According to Djatmiko et al. (2025), positive attitudes toward chatbots reflect openness to digital transformation, which supports SDG 11's smart city initiatives and inclusive technology adoption.

Building on these findings, recent research highlights several mechanisms linking PI to attitude in AI-mediated services. Yoon and Yu (2022) examined restaurant-menu curation chatbots and found that experience characteristics, including perceived value and usability, significantly shaped users' attitudes toward the chatbot, which in turn influenced their intention to reuse the service. Their findings underscore the importance of first-touch experiences in forming lasting attitudinal responses in food service environments. Moreover, Klein and Martinez (2023) conducted an experimental study in the food e-commerce sector, revealing that anthropomorphic design cues and emotional realism in chatbot interactions not only enhanced customer satisfaction but also mediated the relationship between PI and attitude through enjoyment and trust. These findings suggest that when chatbots exhibit human-like communication traits such as empathy, personalization and conversational fluency, they leave a stronger PI, which in turn fosters a more favorable attitude toward the chatbot.

In the restaurant context, Puerta-Beldarrain et al. (2025) note that while traditional ordering methods still outperform chatbots in perceived social presence, well-designed chatbot interactions can bridge this gap by enhancing cognitive attitudes and perceived competence. Therefore, we hypothesize:

Customer satisfaction refers to the overall evaluation customers make based on their entire purchase experience or interaction over time (Alet Vilaginés, 2023). In the context of AI-mediated services, particularly in the restaurant industry, satisfaction is increasingly influenced by how customers feel about the digital agents they interact with, especially chatbots that serve as the first point of contact. Recent research emphasizes that high customer satisfaction is associated with positive experiences, loyalty and repeat business. Accordingly, attitudes toward chatbots could lead to consumer satisfaction. Oruganti et al. (2025) demonstrate that attitude-related constructs, like perceived trust, performance and corporate repute, remarkably impact customer satisfaction in relevance to the use of chatbot. Premathilake et al. (2025) further clarify that attitude, along with enjoyment and trust, explains the relationship between anthropomorphic design cues and customer satisfaction. Specifically, anthropomorphism leads to higher levels of customer satisfaction when attitudes, enjoyment and trust are high.

In the restaurant domain, Al-Oraini (2025) concludes that attitudes toward chatbots in hospitality settings are shaped by the chatbot's ability to simulate social presence and emotional intelligence. When customers perceive the chatbot as approachable, responsive and aligned with their expectations, they are more likely to report positive emotional outcomes, including satisfaction with the overall dining or ordering experience. Therefore, we propose:

Acceptance norms signify a set of predefined requirements that a product, service or object must satisfy to be approved by the customer, user or other relevant stakeholders (Alexandre et al., 2018). A positive attitude toward a product, service or object can significantly enhance customer acceptance and satisfaction. In the context of AI-driven customer service, such as restaurant chatbots, these norms are shaped not only by technical performance but also by social, emotional and contextual expectations. For instance, users may expect a chatbot to respond promptly, understand nuanced requests and reflect the brand's tone, especially in hospitality settings where warmth and attentiveness are valued.

A growing body of research underscores that attitude formation is a critical contributor to acceptance. When users develop a favorable attitude toward a chatbot – perceiving it as helpful, trustworthy or even enjoyable, they are more likely to integrate it into their service expectations. This is particularly relevant in the restaurant industry, where the emotional tone of service delivery can influence whether customers feel comfortable relying on a digital agent for tasks like ordering, customizing meals or resolving issues.

The Technology Adoption Model (TAM), a model that is remarkably applied to measure acceptance of users, highlights the direct impact of attitude on acceptance (Ma et al., 2025). The TAM model's validation of the positive effects of attitude on chatbot acceptance has been confirmed by Kwangsawad and Jattamart (2022). Recent studies have also highlighted that acceptance is not merely a rational decision but is often influenced by affective and experiential factors. For example, Al-Oraini (2025) found that users who felt emotionally engaged during chatbot interactions through personalized greetings, empathetic phrasing or culturally relevant recommendations were more likely to accept the chatbot as a legitimate service channel. This suggests that attitude acts as a bridge between emotional resonance and behavioral endorsement. Additionally, Anshari et al. (2025) discuss that user satisfaction with chatbots signals successful integration of AI into urban services, advancing SDG 11 by improving quality of life through digital innovation.

In restaurant settings, a chatbot that fosters a positive attitude can effectively lower psychological resistance and increase openness to digital service delivery. Thus, cultivating favorable attitudes through thoughtful design and emotionally intelligent interactions becomes a strategic lever for enhancing acceptance. Therefore, we hypothesize:

Finally, the relationship between satisfaction and acceptance has gained remarkable interest from researchers in the AI-mediated service. Satisfaction when measured in reference to a user's overall evaluation of their interaction experience serves as a key influence on technology acceptance. Research shows that when users are emotionally and functionally satisfied, feeling that their expectations were met with ease, enjoyment and reliability, they are more likely to perceive the technology as worthy of continued use (Sundjaja et al., 2025). According to Al-Oraini (2025), satisfaction is attained when reinforcing perceived value and trust, resulting in behavioral acceptance.

In the context of food delivery services, satisfaction has emerged as a pivotal determinant of chatbot acceptance, particularly as users increasingly rely on digital interfaces for service fulfillment. Khan et al. (2025) emphasize that satisfaction in AI-mediated food ordering is shaped by the chatbot's ability to deliver seamless, context-aware interactions that align with user expectations. This aligns with findings by Klein and Martinez (2023), who argue that satisfaction mediates the relationship between perceived service quality and acceptance, particularly when chatbots exhibit adaptive communication and emotional intelligence. Furthermore, Hui et al. (2024) highlight that satisfaction in food delivery chatbot interactions is influenced by system usability, linguistic clarity and perceived responsiveness, factors that collectively enhance the user's willingness to adopt and accept the technology.

Based on this, we propose:

Accordingly, the research model has been designed reflecting how CE, being the foundational stimulus, directly enhances PH and PA, which together constitute the primary functional and social cues experienced by users. In turn, these elevated perceptions foster a stronger PI of the chatbot, as hospitable interactions and authentic, human-like responses reinforce favorable first judgments. This PI then translates into a more favorable attitude toward the chatbot, which drives user satisfaction and ultimately influences acceptance of the chatbot for future food-ordering tasks. Accordingly, the research model below has been defined.

The study adopted a quantitative approach to measure the chatbot-related constructs. The constructs – CE, PH, PA, PI, attitude toward the chatbot, satisfaction and acceptance – were drawn from the model presented in Figure 1. The constructs were selected to examine the human–chatbot interaction and provide insight into the human experience while using chatbots for food ordering. The study attempts to improve the knowledge about human chatbot experience by collecting and evaluating data related to chatbot experience with the aim of building up a notion of the chatbot experience concept.

Moreover, the quantitative research design approach was selected to enable rigorous hypothesis testing and to produce generalizable, statistically robust evidence for our proposed model. By employing structured survey instruments and multivariate analyses (e.g. exploratory factor analysis, reliability testing and path modeling), the study can operationalize and measure latent constructs such as CE, PA and hospitality and assess their interrelationships across a sizable, heterogeneous user sample. This methodological choice aligns with best practices in service-quality and human–computer interaction research, where quantitative approaches have been used extensively to validate theoretical frameworks (e.g. Davis, 1989). Moreover, quantitative data aid in estimating effect sizes, controls for potential confounds and tests the statistical significance of each path of the model. While qualitative methods offer rich contextual insights, they are less suited to confirm predefined theoretical relationships at scale; therefore, a quantitative approach was deemed most appropriate for advancing both theory and practice in the rapidly evolving field of AI-mediated service encounters.

A restaurant chatbot allowed participants to place a food order. The goal was to research consumer interactions with the restaurant chatbot. On the Surge platform (https://surge.sh/), an AI chatbot that can help customers order food was created for the current study. Because this chatbot is new, most participants probably have not used it before, and many might not even know what it is. Therefore, it was essential in this study to introduce and explain the built chatbot application to the consumers. The customer was then instructed to interact with the restaurant chatbot and order food. The ingredients and the cost of the restaurant items were visible to customers. The chatbot asked customers if they prefer to alter or change the ingredients to ensure that the place order met the customer's needs and preferences. We made sure that all the ordered food was available. From consumer–chatbot interaction and the placed order, we intended to examine relevant constructs inspired by our theoretical model.

Once the orders were placed using the chatbot, a survey questionnaire was dispatched to understand the relationship between our key variables. To assess the different variables in the model, we have utilized scales sourced from existing literature. Table 1 shows the operationalization of the variables included in the questionnaire.

The subsequent survey consists of the variable sections depicted in Table 1. All variables were assessed using items drawn from the literature on a 5-point Likert scale, alternating from strongly disagree (1) to strongly agree (5). Each construct employed previously validated scales that have been empirically used in service and technology contexts and content relevant to a food-ordering chatbot.

Item wording was adapted when necessary to ensure clarity and validity without altering core meaning. In CE, three items from Davis's (1989) Perceived Ease-of-Use scale, extended by Venkatesh and Bala (2008), have been introduced to capture responsiveness, accuracy and task-completion speed. Each item was rephrased to reference the food-ordering context, such as “The chatbot processes my order requests without delay,” preserving the original scale's emphasis on system performance. For PH, the reliability and responsiveness dimensions were drawn from the SERVQUAL instrument four items were selected to assess attentiveness and caring tone. Items such as “The chatbot demonstrates genuine concern for my needs” were adapted from face-to-face service encounters to reflect human–chatbot interactions, following precedents in hospitality robotics research (Wirtz et al., 2018).

Moreover, in PA, four-item scale has been adapted based on Fritz et al. (2017) and Rodrigues et al. (2022), targeting sincerity, transparency and human-likeness, such as “The chatbot's responses feel genuine and truthful”. Regarding the PI, and to assess users' overall evaluative judgments, we adapted a three-item measure from Kim et al. (2021), validated in e-service contexts, including “I have a favorable impression of this chatbot.” No substantive modifications were required, as the scale's general wording naturally applies to digital service agents. Additionally, user attitude (ATT) has been measured using the standard two-item semantic differential from Davis (1989), such as “Using this chatbot is good/bad,” contextualizing the referent to “food-ordering chatbot” without altering scale anchors. Finally, for user satisfaction (SAT) and intention-to-reuse (INT), based on Bhattacherjee's (2001a) Expectation–Confirmation Model, three satisfaction items, such as “Overall, I am satisfied with my experience using the chatbot” and intention items, including “I intend to use this chatbot again for ordering food,” were identified. A comprehensive overview of all constructs, their measurement items and corresponding sources is presented in Table A1 Appendix A.

The study data were gathered from participants who had made food orders from Fady's Corner, a Lebanese restaurant that specializes in Lebanese-based cuisine and offers both take-away and dining-in options for platters and sandwiches. The restaurant is located in Jbeil, a coastal city situated 35 Km from Beirut, the capital. Fady's Corner has garnered a strong reputation among local patrons for providing delicious and high-quality meals. Particularly, noteworthy is the restaurant's prompt and superior delivery services during the summer months when the beach, apartments, hotels and beach compounds in the city are bustling with activity. The restaurant has been selected as a single-case site because it uniquely embodies the dual culinary orientation and consumer diversity that characterizes the broader Lebanese market. Unlike restaurants focused exclusively on traditional fare, Fady's Corner offers both Lebanese specialities and international dishes, enabling us to observe chatbot–customer interactions across a full spectrum of menu complexity and cultural expectations. This variation within a single establishment mirrors the heterogeneity of Lebanon's urban dining population, thus ensuring that the findings are contextually grounded and transferable to similar service environments.

As Nickels et al. (2022) argue, a carefully chosen single case can serve as a “critical” exemplar for theory testing and model refinement, allowing for deep, systematic measurement of constructs under naturalistic conditions. By focusing on Fady's Corner, we leverage both its menu diversity and customer mix to validate the hypotheses.

The study involves convenient sampling, and a group of 350 participants have been selected based on age category and their willingness to participate, aged 18 years and above. Data were collected between January 2024 and March 2024. All the participants ordered using the Chatbot. Their phone number and email address were collected after placing their order. The questionnaire included a comprehensive description of the main objectives and parts. The next day, each participant received a phone call to ensure they understood how to complete the questionnaire. All participants were given a free dessert upon completing the questionnaire to thank them for participating. The selected sample covered both males and females of various ages (18 years to above 60 years). Participants were enlisted from various age groups to understand the chatbot experience better.

The participants completed a set of 350 questionnaires. However, after removing the incomplete questionnaires, the sample was limited to 245 individuals, resulting in a response rate of 70%. Accordingly, the sample size encounters the minimum sample size requirement (Blair and Blair, 2015).

In reference to earlier studies, variables like education and age are frequently incorporated as control variables, particularly in the technology context (Cheng and Mitomo, 2017). To ensure that the findings from the empirical research are not influenced by variations from such demographic variables, these control variables were introduced in the questionnaire.

Furthermore, to guarantee the internal validity of research findings, we introduced two potential confounding variables in the questionnaire. First, we requested the participants to report their level of familiarity with the restaurant before responding to questions about the constructs of interest. This is particularly significant given that the study focuses on a single restaurant and participants' previous experience with the establishment could potentially influence their perceptions of the menu items, products, ingredients and other relevant factors, consequently affecting their responses. Participants were particularly asked if this was their first time visiting the restaurant. First-time visitors may be less influenced by prior expectations tied to the restaurant's service reputation or manual ordering habits, allowing for a more unbiased assessment of the chatbot's PH and authenticity. On the other hand, frequent visitors may rely on past interactions, which could shape their responsiveness or resistance to digital assistance, thereby affecting their evaluation of the chatbot. Second, a question regarding computer literacy was included to evaluate participants' proficiency and familiarity with computer technology. Computer literacy was included as a control variable because it influences how comfortably users interact with AI tools. Unlike low computer literacy, participants with strong skills tend to navigate chatbots more easily. They can respond to prompts quickly, follow the conversation tone and understand the chatbot's features. This often leads to smoother interactions and greater acceptance of technology.

The collected data were subjected to cleaning and analyzed through the Statistical Product and Service Solutions software (Hejase and Hejase, 2013) (IBM SPSS, Version 25.0). Additionally, univariate analyses were conducted to detect and eliminate any outliers and to generate descriptive statistics of the sample. Additionally, exploratory factor analysis and reliability analysis (Cronbach's alpha) were carried out to assess the dimensionality and internal consistency of the measurement scales. Next, an analysis of moment structures on IBM AMOS allowed us to calculate standard measures for structural equation modeling: Root Mean Square Error of Approximation (RMSEA), Goodness of Fit index (GFI), Comparative Fit Index (CFI), Normed fit index (NFI), root mean square residual (RMR), CMIN/DF (the minimum discrepancy) and Adjusted Goodness of Fit Index (AGFI). This analysis allowed testing the measurement model and the conceptual model.

To examine the measures model's dimensionality and reliability, an exploratory factor analyses and reliability analyses in SPSS were carried out. For each construct, factor analysis with Varimax rotations was conducted to determine its dimensionality. These results suggest that conducting factor analysis is applicable. Furthermore, the Bartlett’s test of Sphericity yielded statistically significant results, indicating that the elements in the correlation matrix were correlated (sig = 0.000 < 0.05).

Additionally, Cronbach's alpha values were calculated to evaluate the internal consistency of the questionnaire items for each construct. The reliability coefficients, shown in Table 2, were found between 0.756 and 0.927, which are generally regarded as good to excellent (Hejase and Hejase, 2013).

The relationship between the variables of the structural equation model was tested using IBM AMOS. Based on the fit indices, the model appears to be acceptable, as evidenced by a CMIN/df value of >3, CFI (West et al., 2012), NFI, AGFI (Tabachnick and Fidell, 2007) and GFI (Hu and Bentler, 1998) values of >0.9. Additionally, the RMR value is < 0.08 (Diamantopoulos and Siguaw, 2000) and the RMSEA value is between 0.05 and 0.08 (MacCallum et al., 1996).

A subsequent confirmatory analysis (Table 3) showed an acceptable fit of the model in accordance with Byrne (2013). This was achieved through a purification process by reducing the PH scale to 3 items, the PA scale to 5 items and the PI scale to 3 items.

As illustrated in Table 3, our model conceptualizes experiences of chatbot satisfaction as arising from the attitude toward the chatbot, which in turn arises from PI. These experiences are facilitated by variables such as chatbot PH and PA that arise from CE.

The test of the structural equation model yielded satisfactory model fit indices (GFI = 0.89, RMSEA = 0.036, CFI = 0.975, NFI = 0.904, RMR = 0.35, CMIN/DF = 1.309, AGFI = 0.869). All the hypotheses were supported except H₄ and H₈ (Table 3). These results show that many relationships of the structural model are validated.

This study offers several theoretical contributions to the literature on AI-mediated service encounters. First, by combining CE and PA as joint antecedents, we move beyond traditional models that focus on either technical performance (e.g. TAM; Davis, 1989) or social cues alone. Findings show that efficiency not only speeds up transactions but also serves as a visible signal of sincerity, thereby enhancing authenticity.

Second, we validate a clear, end-to-end sequence, from foundational inputs (efficiency, authenticity and hospitality) through PI and consumer attitude, to reusage intention, within the Stimulus–Organism–Response framework. While prior studies have examined isolated links such as authenticity and satisfaction, this research is among the first to empirically test the entire chain in a food-ordering context. This holistic approach provides a robust template for future research on user adoption of AI agents in hospitality and beyond.

Third, by focusing on a Lebanese restaurant that offers both local and international dishes, the study highlights how cultural diversity and menu complexity influence chatbot acceptance. This case underscores the importance of PH in settings where customers expect both traditional warmth and global standards. Finally, these contributions refine theoretical models of human–chatbot interaction and lay the groundwork for future research on multi-dimensional drivers of AI adoption in service industries. Moreover, by demonstrating how these constructs support SDG 11 through enhanced urban service delivery and SDG 12 via responsible consumption behaviors, the research offers a multidimensional framework that bridges consumer experience with sustainable development. These insights encourage future studies to explore AI's role not only in shaping user experience but also in advancing ethical and scalable service models that align with global sustainability agendas.

This study also offers several implications for practitioners. In general, results indicate that CE, PA and PH jointly shape user impressions and attitudes, ultimately driving reuse intention. Restaurants can leverage these cues to design and deploy chatbots that better meet customer expectations in food-ordering settings. First, enhance CE to build trust and smooth transactions through investing in optimized natural language processing so that the chatbot processes orders quickly and accurately. For instance, use scalable cloud services to reduce response delays and continuously chatbot language model on real customer queries to minimize misunderstandings. Second, embed authenticity cues into conversational scripts. Design dialogue flows that explain recommendation logic and acknowledge mistakes with brief apologies when errors occur. Transparent and human-like touches convey sincerity and help customers feel the chatbot genuinely cares about their needs. Third, implement continuous performance monitoring and iterative A/B testing of your chatbot's conversational flows. Integrating analytics tools that track key metrics such as response time, completion rates, drop-off points and customer feedback scores helps in identifying friction points in real time.

In addition to improving experience, chatbot deployment in food services presents a strategic opportunity to support sustainability objectives. Restaurants can leverage chatbot-driven personalization and automation to enhance service delivery capacity in urban environments (SDG 11) and reduce resource waste through accurate order processing and tailored recommendations (SDG 12). These findings suggest that chatbot design should not only prioritize user experience but also embed sustainability cues such as eco-friendly menu suggestions or low-impact delivery options to reinforce responsible consumption and urban resilience.

Despite the valuable findings, this research has several limitations. The study relies on cross-sectional, survey-based data from a single restaurant, which may limit the generalizability of the findings across different service contexts and over time. Real-time behavioral data, such as actual chatbot usage logs or error rates, were not captured, which could provide deeper insights into how users interact with and react to chatbots. Future research should address these gaps by implementing continuous performance monitoring and iterative A/B testing of chatbot conversational flows. Also, proven questionnaire questions have been used; researchers still need to make sure the variable items really fit the food-ordering chatbot setting. Future studies could use customer interviews or guided feedback sessions to tweak the wording and capture exactly how users feel about digital hospitality and authenticity. Furthermore, although we measured technology readiness in our survey, we didn't explore how this and other personal factors, such as reluctance to try new foods, affect users' reactions to chatbot cues. Future studies should look at these differences to create more tailored chatbot experiences for different customer segments. Lastly, while the study offers valuable insights into chatbot delivery service experiences, it does not empirically reflect the sustainability outcomes linked to SDG 11 and SDG 12. Although the conceptual model integrates urban accessibility and responsible consumption as theoretical anchors, future research should incorporate direct sustainability metrics such as reductions in resource waste, delivery efficiency or digital inclusion in urban contexts to validate these contributions.

This study has provided valuable insights into the influence of chatbot qualities on user experience in the food ordering service. The findings highlight that CE plays a fundamental role in enhancing PH and authenticity, which in turn positively influence customer satisfaction and acceptance. Specifically, the research confirms that while chatbots are seen as effective tools for improving the food ordering process, their PA does not directly affect the PI of the user, indicating that customers appreciate the functional aspects more than the human-like qualities of chatbots. Additionally, even though a positive attitude toward chatbots increases satisfaction, it does not necessarily lead to acceptance, suggesting that users may view chatbots as helpful but are not yet fully ready to embrace them as a replacement for human interaction.

These results suggest that restaurants and food delivery services can enhance customer experience by focusing on improving CE and hospitality. However, businesses should also recognize the limitations of chatbots in terms of their PA and acceptance. Future research should explore these dynamics further, considering factors like user demographics, chatbot design sophistication and broader applications beyond food ordering to better understand and improve chatbot integration in various customer service contexts. Moreover, integrating sustainability messaging via chatbots can help promote SDG 11 by encouraging eco-friendly dining choices.

Beyond enhancing customer experience, the findings underscore the potential of chatbots to support sustainability goals. By improving service delivery capacity and promoting ethical consumption behaviors, chatbot deployment in food services contributes meaningfully to SDG 11 and SDG 12. These results suggest that AI-driven service tools are not only operational assets but also strategic enablers of urban resilience and responsible consumer engagement.

Future research should build on this work by examining how different customer segments, such as by age or tech-savviness, respond to chatbot traits, testing more sophisticated dialogue designs and exploring chatbot use in other service industries like retail or banking. Additionally, researchers should incorporate measurable sustainability indicators and explore contextual variables that influence chatbot-delivery service across diverse urban environments.

The supplementary material for this article can be found online.