Encourage Online Consumers to Embrace Voice Shopping: Roles of Tasks, Technology, and Individual Characteristics

Abstract

Voice shopping brings consumers convenience and retailers a new channel to reach buyers, which is an important component of online shopping. However, it has received a tepid response recently. Two issues must first be addressed to promote voice shopping: why consumers have a low acceptance of voice shopping and how to motivate their use. Since technology is intended to serve practical purposes, it is necessary to achieve a fit among voice shopping, online shopping tasks, and consumers. Accordingly, this study elaborates on fit and integrates the task-technology fit (TTF) theory (for objective fit) and the technology acceptance model (TAM, for subjective fit) to build a research model in which task, technology, and individual characteristics affect fit that will stimulate voice shopping intention. Using Structural Equation Modeling to analyze the data collected from 425 consumers who do not accept voice shopping, the results show that both objective fit and subjective fit can determine voice shopping intention; however, some critical task, technology, and individual characteristics cannot affect either aspect of fit, indicating that consumers still do not believe voice is workable in online shopping. These findings make suggestions for a purposeful upgrade of the voice shopping experience, which helps promote voice shopping and ultimately contributes to the prosperity of online shopping. This study also offers insights into what constitutes fit and its roles in the integrated model of the TTF theory and TAM.

1. Introduction

Since the computer was introduced to the public, typing has been a common way to interact with devices. For a long time, people have become accustomed to opening a website and typing a few words when they need information. Recently, the rapid development of voice technology has changed this situation—it serves users via their most familiar way of interaction, i.e., speaking in a natural language. The voice assistant, one of the most typical applications, such as Siri, Google Assistant, and Alexa, stands out. With Speech Recognition, Natural Language Processing, and Text-to-Speech functions, they can elaborate on human speech, understand their intent, and respond in a human-like voice. Increasingly, people are showing enthusiasm for speaking to check the weather, send messages, and control home appliances. By 2026, over 150 million voice assistants are projected to be found in the US alone [1].

While being popular among users, voice assistants also garner attention in business. Retail giants like Amazon and Alibaba have incorporated them into shopping functions, which enables consumers to make purchases through voice commands. This new trend—voice shopping—supports an on-the-go shopping experience and creates another touchpoint for businesses to reach buyers [2], showing a disruptive potential in e-commerce. Studies have found that smart speakers can encourage users to make more purchases [3,4] and be more loyal to sellers [5]. The voice shopping market share is predicted to rise by USD 80.21 billion from 2024 to 2029 at an increasing CAGR of 22.7% [6]. Apparently, voice shopping is becoming an important component of the online shopping experience. However, it is worth noting that over the past few years, the adoption of voice shopping has been slower than expected and even stagnated [7,8]. About 60% of online buyers made daily or weekly purchases via smart home voice assistants in 2021, but a 2024 study reported that only 24.5% of consumers conducted voice shopping every day or every week [9,10]. Although professional service providers, such as Amazon and Apple, have been racing to an overhaul to improve voice shopping [11], this “most overlooked innovation” in e-commerce still faces challenges in acceptance [12], which raises the following two research questions:

RQ1 

: Why do many consumers not use voice shopping currently?

RQ2 

: How can the voice shopping experience be improved to encourage more usage?

Despite growing interest in voice assistants, little is known about consumer behaviors in the voice shopping context. Most existing studies focus on the determinants of consumers’ purchasing behavior or final decision (e.g., [2,3,4,7,13,14,15,16,17]), contributing to the improvement of voice shopping. However, they were mostly conducted under the assumption that consumers have already accepted or regularly used voice shopping, without considering the situation where consumers have not accepted it. Suggestions from the above works cannot directly apply to voice shopping promotion. Scholars took notice and started discussing the acceptance of voice shopping. They have investigated technology characteristics, such as privacy concerns, message interactivity, and human–AI interaction fluency [7,18,19,20,21,22,23,24,25,26,27], and individual characteristics, such as innovativeness and personal attitudes [8,28]. However, to date, many technology characteristics of interest pertain to the general features of voice assistants rather than attributes specific to shopping purposes. Moreover, only a little research examines the influence of individual characteristics and considers them with technology characteristics simultaneously. In general, technology has practical purposes [29], whose acceptance in the long term is not only an issue of accepting technology per se but an issue of accepting its specific functions for certain tasks [30,31]. Voice shopping requires consumers to actively interact with the voice assistant for conducting shopping tasks, indicating that the roles of task-related, technology-related, and consumer-related characteristics should not be ignored. Therefore, to fill the research gap, this study aims to investigate the acceptance of voice shopping by simultaneously considering task, technology, and individual characteristics specific to the voice shopping context.

Fit, the concept that aligns the interrelationships among technologies, tasks, and individuals [32], can be understood as the extent to which technology features meet the requirements of the task and abilities of an individual. Studies have emphasized that AI applications, such as voice assistants, should balance the human–computer roles they play as a consumer touchpoint when serving consumers (e.g., [7,8,33]), implying that the fit among voice-assistant-supported shopping technology, online shopping tasks, and consumers is crucial. Therefore, this study leverages the concept of fit to understand and encourage the acceptance of voice shopping. In the fit area, the task-technology fit (TTF) theory and technology acceptance model (TAM) are the two best-known theories, but they have different focuses. The TTF theory posits that matching technology capabilities and task demands contributes to technology use. It is task-focused without much consideration for the users’ beliefs and attitudes [34], which lays particular stress on objective fit. The TAM proposes that users’ perceptions of the technology are essential, which can potentially present the subjective fit between users and technology [32,34,35]. To maximize their advantages, this study integrates the TTF theory (for objective fit) and the TAM (for subjective fit) to develop a research framework in which task, technology, and individual characteristics affect the two aspects of fit that determine the voice shopping intention.

This work expects to make two major contributions. First, leveraging fit to construct the research model for voice shopping acceptance will not only enhance the theoretical basis for integrating the TTF theory and TAM but also enrich the concept of fit. This work introduces fit among task, technology, and consumers into the voice shopping domain and applies it to strengthen the theoretical foundation of integrating the TTF theory and TAM. It further examines the validity of the integrated model by comparing the data analysis results of the integrated model with two individual models, which is expected to offer a robust basis for future research on technology usage in voice shopping or other AI application settings. Moreover, in the process of clarifying the influences of fit on voice shopping acceptance, this study breaks fit into objective fit and subjective fit, which provides insights into the concept of fit, as divergence still exists regarding what constitutes it. Second, focusing on the influences of important task, technology, and individual characteristics specific to the voice shopping context can yield precise and actionable suggestions for service providers to encourage technology acceptance. This study aims to address two research questions: the reasons many consumers do not use voice shopping currently and solutions to improve the voice shopping experience for wider adoption. The results can provide a practical basis for service providers and retailers to upgrade and promote voice shopping, ultimately fueling the growth of online shopping.

2. Conceptual Framework Development

2.1. Theoretical Foundation for Fit

The concept of fit has continued to receive attention in technology acceptance studies. Technology can be broadly understood as “a designed, material means to an end” [36]. That means technology has its practical purposes [29]. Accordingly, scholars considered that technology acceptance in the long term is not only an issue of accepting technology per se but also, of primary importance, an issue of accepting its specific functions [30,31], emphasizing the significance of fit.

Initially, fit was introduced to explore the interaction between task and technology [37,38]. It refers to the degree to which technology functionality satisfies the task requirements (e.g., [39]). Goodhue and Thompson’s TTF theory precisely explains this interaction [40]. The theory posits that, regardless of people’s attitude toward the technology, they will not accept it if a mismatch exists between the technology’s functionality and task demands, indicating that task characteristics and technology characteristics are important. In this theory, the degree of fit between tasks and technology is considered the determinant of the user’s performance and technology utilization (see Figure 1). As the significant role of the original task-technology fit has been proven, scholars have gradually expanded its scope to include task-individual fit and individual-technology fit (e.g., [32,35,41]). Therefore, the concept of fit was extended to the alignment of interrelationships between tasks, technology, and individuals [32,35,42]. In recent years, new terms such as environmental-technology fit and experience-technology fit have also been proposed [43,44]. However, very few studies can test all dimensions of task-technology fit due to the model complexity. Consequently, even if fit has been applied to explain the acceptance and usage of various technology tools, divergence still exists about what constitutes it [37,42].

Another approach to operationalizing fit offers insights into a better understanding of this concept—some scholars consider that the support fit is also important. Aspects such as ease of use and accessibility are often used to describe the fit [47]. Such an operation was also recognized by Goodhue et al. [41]. They introduced user evaluations of fit (e.g., ease of use) to make up for the potential shortage of task-technology fit. As a result, studies treat task-technology fit as actual and rational, but user evaluations of fit are sensitive and depend on the users [34,35,41]. Among the studies, the most typical approach is to combine the TAM and the original task-technology fit (see Figure 1), as the TTF theory is primarily task-focused, whereas the TAM focuses on users’ beliefs and attitudes toward the technology [34]. Figure 1 shows that perceived usefulness (PU) and perceived ease of use (PEOU) are the two crucial factors of the TAM. Both of them reflect an individual’s subjective evaluation of the technology for being applied to specific tasks. PU refers to the probability that a prospective user believes using the technology will help them enhance their performance, and PEOU means the extent to which a prospective user considers the technology to be effortless [45]. The results have proven that integrating the TTF theory and TAM can assist in better understanding why people use certain technologies for specific tasks (e.g., [34,48,49]).

The above indicates that fit has two main aspects: objective fit and subjective fit. The objective fit shows the extent to which technology is suitable for performing a portfolio of tasks; the subjective fit is the degree to which a user believes that technology can help them conduct the task.

Like other technologies, voice shopping also has its practical purposes—it should support online shopping tasks. In addition, when interacting with voice assistants for shopping, users can also evaluate them [22], emphasizing the importance of subjective evaluation. Thus, when applying the fit concept to understand the acceptance of voice shopping, the researchers chose to integrate the objective fit and subjective fit. The objective fit refers to the fit between online shopping tasks and voice shopping technology, which the task-technology fit in the TTF theory can measure. The subjective fit refers to the degree to which a user believes voice shopping can help them conduct online purchase activities, as reflected by perceived usefulness and perceived ease of use in the TAM in this study. According to Davis et al. [45], PU is further defined as the degree to which a person believes that using voice shopping would increase his or her performance in online shopping and PEOU as the degree to which a person expects that using voice shopping would be free of effort. Based on the TTF theory and TAM, it is reasonable to propose that the higher the task-technology fit, PEOU, and PU, the more likely individuals are to accept voice shopping. Thus

H1. 

Task-technology fit positively influences the intention to use voice shopping.

H2. 

PEOU positively influences the intention to use voice shopping.

H3. 

PU positively influences the intention to use voice shopping.

As mentioned, studies have demonstrated that the combined model of the TTF theory and TAM has a greater explanatory power for technology use than the two individual models. Some researchers also directly applied the integrated model to explain technology acceptance and discussed the relationship among task-technology fit, PEOU, and PU [50,51,52,53]. Task-technology fit is often expected to predict a user’s perception of technology utilization. For example, Goodhue and Thompson [40] claimed that task-technology fit could determine the evaluation of technology’s advantages, like usefulness. Park et al. [53] hypothesized that task-technology fit could positively influence an individual’s expectancy of performance (i.e., PU) and effort (i.e., PEOU). Similarly, if the voice shopping method fits well with one’s online shopping needs or tasks, they are more likely to perceive the technology as easy to use and useful. In addition, the TAM claims that PEOU can affect the acceptance intention directly or through PU [45] because technology that is easy to use can increase an individual’s behavior performance [54]. Based on the above, it is reasonable to infer the relationship as follows:

H4. 

Task-technology fit positively influences the PEOU of voice shopping.

H5. 

Task-technology fit positively influences the PU of voice shopping.

H6. 

PEOU positively influences the PU of voice shopping.

2.2. Voice Shopping and Related Task, Technology, and Individual Characteristics

According to the definition of fit, it is closely related to task, technology, and individual characteristics. However, when integrating the TTF theory and TAM, research usually discusses technology characteristics from an overall perspective, i.e., measuring it by one construct. In practice, however, technology often has multiple features affecting user acceptance. Analyzing their effects can help service providers promote technology usage more effectively. In addition, individual characteristics have not received much attention in the integrated model. For voice shopping, individual initiatives could be crucial, as users should speak with their devices actively. Thus, this study proceeds to clarify the important task, technology, and individual characteristics in the voice shopping context.

2.2.1. Studies Related to Voice Shopping Acceptance

Voice shopping is one of the novel trends in retailing. It enables individuals to conduct online shopping simply by speaking to voice assistants embedded in smartphones or smart speakers. This new shopping method is promising for improving the online shopping experience, thereby benefiting both retailers and consumers. However, compared with other voice-activated functions, such as texting messages and asking for the weather, voice shopping is still in its initial stage [8].

As mentioned, in the voice shopping setting, antecedents of a consumer’s buying behavior or decision have attracted much attention [2,3,4,7,13,14,15,16,17]. Research on the determinants of voice shopping acceptance is still relatively scarce [19]. Early studies often asked consumers about their perceptions of voice-activated devices and then categorized reasons into benefits and costs [22,23,24,25,26]. To date, the technology characteristics of voice assistants and individual characteristics of consumers have become the focus of published literature (see

Table 1. A summary of factors in several representative voice shopping studies.

References	Characteristics		Final Outcome Variables	Methodology
	Technology Characteristics	Individual Characteristics
[18]	Privacy concerns	Extroversion, agreeableness, conscientiousness, emotional instability, intellect, and prior experience	Customer experience performance	Regression analysis
[55]	Perceived risk, privacy concern, and technology learnability	N/A	Voice shopping behavior	Regression analysis
[20]	Privacy concerns	Technology anxiety	Behavioral intention	Regression analysis
[7]	Biased offers	N/A	N/A	Interview
[56]	Anthropomorphism	N/A	Intention to adopt voice shopping	Regression analysis
[21]	Message interactivity	Machine heuristic	Intention to use	Regression analysis
[19]	Anthropomorphism, i.e., human-like voice, social presence, and friendliness	N/A	Word of mouth and voice shopping intention	Regression analysis
[57]	Enjoyment, performance expectancy, and perceived safety	N/A	Voice shopping intention	Regression analysis
[27]	Human–AI interaction fluency	N/A	Voice shopping intention	Regression analysis
[58]	Perceived competence and perceived warmth	N/A	Voice shopping intention and value co-creation intention	Regression analysis
[28]	N/A	Functional, hedonic, social, and cognitive motivated consumers innovativeness	eWOM and purchase intention	Regression analysis
[8]	N/A	Cognitive beliefs, affective feelings, and conative behavioral intentions	N/A	Fuzzy Analytical Hierarchy Process
[59]	Positive and negative characteristics	N/A	Customer engagement and enhanced usage	Interview
[60]	Anthropomorphism, privacy concerns, and functional intelligence	N/A	Behavioral intention	Regression analysis

Note: Only technology characteristics, individual characteristics, and final outcome variables are included for a clear view.

).

Table 1 reveals the following: (1) the existing studies paid more attention to the technology characteristics of voice assistants. Different from initial research, the current studies mainly examine either the positive factors [19,21,27] or the negative ones [7,18,20], such as interaction advantages and privacy risks. Most of them belong to the general characteristics of voice assistants rather than attributes specific to shopping purposes. However, technology often has practical purposes [29]. Its acceptance is not merely an issue of accepting technology per se but an issue of accepting its specific functions [30,31]. Thus, it is necessary to delve into the characteristics of voice technology for shopping purposes. (2) The influences of personal factors need further investigation. Only a few scholars have discussed the role of personal factors in voice shopping acceptance [8,28]. Moreover, little research considers the impact of both technology and individual characteristics in this setting simultaneously. Voice shopping requires consumers to use their voice to actively interact with the voice assistant for conducting shopping tasks, indicating the significant roles of shopping tasks, technology, and consumer-related factors. Therefore, this study focuses on the task, technology, and individual characteristics specific to the voice shopping context and examines how these factors affect consumers’ acceptance of voice shopping.

2.2.2. Task Characteristics

Tasks refer to the actions conducted by individuals in turning inputs into outputs to satisfy their needs [40]. It is a piece of work that people perform through a series of actions to achieve a goal [61]. Thus, the task characteristics can be regarded as requirements, perceived needs, or dimensions related to those that a user might apply technology to perform [49,61,62]. This study defines the task as the activities in which individuals engage during online shopping. Accordingly, the task characteristics are an individual’s online shopping requirements. With the development of online shopping, recent studies have found that online shoppers increasingly face trends such as choice and information overload [63,64] and multiple tasks [65,66]. As a result, they inevitably expect an adaptable, immediately accessible, and custom “do it for me” shopping experience [18,66].

The TTF theory hypothesizes that task characteristics influence the task-technology fit because the corresponding requirements encourage users to be more dependent on the technology [40]. Several studies have provided evidence for this relationship [52,61,67]. Regarding PU and PEOU, some work proposes that task characteristics can influence PU but not PEOU. As Lee et al. [21] claimed, examining the direct relationship between task characteristics and PEOU may not be plausible because PEOU is contingent on whether the target system is easy to use or not. Similar to voice shopping, it can satisfy users’ needs when it is inconvenient to type for shopping. That helps users perform online shopping tasks anytime and anywhere, making them perceive that voice shopping is useful. However, it cannot change the operation process of voice shopping. Thus, this study posits the following:

H7a. 

Task characteristics positively influence task-technology fit.

H7b. 

Task characteristics positively influence the PU of voice shopping.

2.2.3. Technology Characteristics

Technology characteristics refer to the attributes of tools individuals use in performing tasks [68]. According to the TAM and TTF theory, technology attributes can influence its acceptance through task-technology fit, PEOU, and PU [34,52,69]. In practice, technology has both positive and negative attributes. This work focuses on the positive aspects of voice assistants for shopping. There are two main reasons. First, recent studies have disclosed that positive attributes or feelings toward voice shopping are taking precedence over the negative ones (e.g., [8,19,70]). Second, the primary purpose is to uplift the voice shopping experience through possible technologies or strategies. Therefore, this study continues to extract the positive features of voice shopping compared with other popular shopping methods.

Voice technology has tremendously changed how people interact with information sources, altering the way they purchase products. Before the Internet era, people were accustomed to searching for products and buying them in an offline context. They usually walked to the physical store, checked products with a sales clerk in most cases, and paid. It may provide more accurate product information, but this shopping method is usually inconvenient and offers limited choice. With the popularity of the Internet, people can navigate the sea of information. They start to search for products on the computer by typing on the keyboard. In this period, people have more choices, but they still need to sit in front of computers. As mobile technology has developed, users have gradually obtained timely access to product information, but typing on mobile phones cannot separate them from their devices physically, and they become more used to being “silent” with their phones. With the arrival of voice technology, consumers have found speech-based communication to be more natural and effective [24,71]. Voice shopping enables users to obtain information by simply asking their voice assistants in spoken language anytime and anywhere [72,73]. Based on the above, this study identifies three prominent characteristics of voice shopping: convenience (one-step search and possible purchase), accessibility (the ability to enable consumers to obtain shopping information anywhere and anytime), and context awareness (the ability to understand what a user really wants in spoken language, including meaning and intent). The following will elaborate on them and their possible influence on two aspects of fit.

• Convenience

Despite much attention to convenience in various areas, studies have not reached a consensus on its definition and dimensions [74,75]. Since time and effort utilities have mainly been discussed [74], this study defines convenience as the extent to which voice shopping could provide consumers with time and effort utility. It is the most critical characteristic of voice shopping [19].

The convenience of information technology tools can impact task-technology fit and users’ perception of ease of use and usefulness [69,76]. Similarly, voice shopping enables users to perform shopping tasks by speaking to their voice assistants without steps like clicking or filling out forms [73,77]. That means voice technology automatically realizes the one-step search for shopping information. Additionally, spoken language is the most natural approach for humans [24]. Compared with typing on a computer or smartphone, voice shopping is often easier to conduct [18] because it does not require additional education, effort, and skill related to typing [77,78]. As a result, voice shopping can save users time and effort [16], leading to perceived ease of use and usefulness. Related advantages can also match the shopping needs of current consumers who seek to quickly and easily find what they want in the era of choice and information overload [64]. Thus, this study hypothesizes that

H8a-1. 

Convenience positively influences task-technology fit.

H8a-2. 

Convenience positively influences the PEOU of voice shopping.

H8a-3. 

Convenience positively influences the PU of voice shopping.

• Accessibility

Accessibility can be briefly understood as the ability to access a resource, such as an entity, information, or service. For voice shopping, this study considers accessibility as the ability to allow users to reach shopping information and services anytime, anywhere, and in any situation [18]. As mentioned before, previous shopping methods, including offline shopping, online shopping on computers, and mobile shopping, have limitations in time and/or space. On the contrary, voice technology can physically separate users from their devices. Voice control and the innovation of far-field voice capture technology can enable users to initiate the shopping journey without many limitations. That means users can remain reachable and engaged in shopping activities at all times.

Despite limited direct evidence for the effect of accessibility on task-technology fit, PEOU, and PU (e.g., [79,80]), factors with a similar concept, such as reachability [69] and portability [81], have shown positive influences on these three constructs. By prioritizing accessibility, voice shopping can meet the increasing demand for reaching and interacting with shopping systems across time, space, and contexts, especially among those who often have to or prefer to handle multiple tasks [66,82]. In addition, it ensures equal access to online shopping for diverse consumer groups [83]. Thus, this study proposes the following:

H8b-1. 

Accessibility positively influences task-technology fit.

H8b-2. 

Accessibility positively influences the PEOU of voice shopping.

H8b-3. 

Accessibility positively influences the PU of voice shopping.

• Context awareness

Context means the information that can be applied to characterize an entity’s situation [84]. A system is context-aware when it can provide information and services, automatically execute these services, and tag information for later retrieval based on context [85]. A context-aware information technology device or application would improve its ease of use, usefulness, and efficiency rating (e.g., [85,86]).

Voice assistants leverage conversational AI to communicate with users. AI refers to the study of agents that receive percepts from the environment and perform actions [87]. Thus, high context awareness can be a distinctive feature and the future of voice assistants for shopping [88]. With this sophisticated feature, voice assistants can grasp what a user says, including his or her location, mood, and even intent, and respond to commands based on different situations. Such a deep understanding of conversation could guarantee a high-quality shopping process, which enables a personalized experience for consumers [18,89]. Accordingly, it is reasonable to assume that context awareness can influence the task-technology fit, PEOU, and PU.

H8c-1. 

Context awareness positively influences task-technology fit.

H8c-2. 

Context awareness positively influences the PEOU of voice shopping.

H8c-3. 

Context awareness positively influences the PU of voice shopping.

2.2.4. Individual Characteristics

Individual characteristics are essential variables in studies on acceptance behavior. Factors like personal innovativeness, self-efficacy, and experience have received considerable attention (e.g., [50,80,90,91,92]). Voice technology shifts the way people engage in online shopping—they need to open their mouths [71]. Therefore, it is reasonable to assume that individual initiative can affect their acceptance of this new shopping method. That could also explain why studies in this area start from the motivation or attitude of users. Among the characteristics above, personal innovativeness is closely related to one’s initiative [92], a trait that also attracts interest in the adoption of voice assistants [90]. In addition, an increasing number of people use human metaphors to describe their voice assistants. Human-like cues give them a sense of social presence through conversation [71], rendering the human–voice assistant interaction a social interaction [93]. In particular, when it comes to shopping, users are often expected to initiate the interaction and express their needs. However, people still feel creepy and socially awkward about voice shopping, revealing that they are not fully comfortable asking voice assistants to perform this task [8]. That is, people are still embarrassed about talking to a machine when they are on their own, not to mention when they are in public. According to the American Psychological Association, feeling creepy or awkward during social interaction could be more due to the trait of shyness. This personality trait can easily cause people to talk less [94] and limit their initiatives [95], which deviates from the basic requirements of voice shopping. As a result, this study focuses on two individual characteristics, personal innovativeness in information technology and shyness.

• Shyness

Shyness refers to an individual’s tendency to avoid social interactions and fail to participate appropriately in social situations [96]. It reflects the actual frequency of an individual talking, which represents a behavioral pattern [94]. Scholars imagined that compared to face-to-face, shy people might prefer communicating in online settings [97,98]. Compared with less shy individuals, however, shy people are more inclined to avoid social behaviors like interaction [99,100,101], are more afraid of negative evaluation, and are more likely to poorly self-evaluate and show deficiencies in self-presentational skills [97,102,103]. As a result, shy people tend to talk less [94]. However, voice shopping requires users to actively speak with devices. Due to the fear of talking in others’ presence and/or being negatively evaluated [96,97], shy users may consider typing to purchase more suitable and easier. Thus, this work proposes the following:

H9a-1. 

Shyness negatively influences task-technology fit.

H9a-2. 

Shyness negatively influences the PEOU of voice shopping.

H9a-3. 

Shyness negatively influences the PU of voice shopping.

• Personal innovativeness in information technology (PIIT)

Personal innovativeness, the degree to which people accept an innovation before others in a social system [104], is an essential determinant of acceptance behaviors [90]. An individual can be regarded as innovative if he or she is early to adopt an innovation [105]. With the development of information technology, researchers specified personal innovativeness in information technology and defined it as an individual’s willingness to experience new information technology (e.g., [105]). PIIT is usually applied in technology acceptance research (e.g., [50,90,92]). Innovative people are open to change and expect to realize an innovation’s potential [91] because they tend to be curious about new things, active in searching for innovation-related information, willing to take risks, and communicative [69,106]. Thus, they are more likely to feel that it is not hard to learn about a new thing [107]. A positive relationship between PIIT and PEOU can exist for these users; however, PIIT may not directly influence the utilities of voice shopping [107], especially the objective fit between tasks and technology. Thus, the study proposes that

H9b. 

PIIT positively influences the PEOU of voice shopping.

Based on the above, this study finally constructs the research model in Figure 2. To reveal the possible influences of task, technology, and individual characteristics on voice shopping acceptance, this model introduces and clarifies the concept of fit. In this process, the TTF theory and TAM are integrated to explain the two aspects of fit. That is, task-technology fit from the TTF theory can indicate the objective fit that the available functionality of voice shopping technology is suitable for online shopping tasks; perceived ease of use and perceived usefulness present the subjective fit that consumers believe the voice shopping technology can help them purchase online. Meanwhile, compared to Figure 1, the research model forges links between the two aspects of fit. It proposes that the objective fit can trigger subjective fit. In addition, the research model also makes theoretical improvements in two aspects. First, the current work classifies the technology characteristics of voice shopping into convenience, accessibility, and context awareness rather than a single construct. Second, the model adds individual characteristics that may influence technology acceptance. As mentioned, the two kinds of characteristics have received attention in both the TTF theory and TAM. However, their contributions to the integrated model remain unclear. Understanding their impact on subjective and objective fit in the voice shopping setting enables more precise marketing implications. In summary, the current research model can amplify the benefits of integrating the TTF theory and TAM.

3. Research Methods

3.1. Data Collection

This work aims to understand why consumers do not accept voice shopping and how to promote voice shopping. Therefore, the consumers who do not accept voice shopping are the research target, including those who have once experienced voice shopping but no longer use it and those who have not experienced voice shopping. To test the research model, the researchers conducted an anonymous questionnaire survey among them. For consumers who have tried voice shopping, the selection criterion was set as “has once experienced voice shopping, but has not used it in the past six months”. Since there is no need to explain too much about voice shopping, a brief explanatory text was attached to the front of the questionnaire. The potential respondents were requested to read the following: “Speech recognition technology assists us in speaking with devices. A representative example is talking with voice assistants, like Apple’s Siri, Xiaomi’s Xiaoai, and Alibaba’s Tmall Genie. Instead of typing on the keyboard or screen, we can directly ask our voice assistants to search for the weather, phone number, and time. What is more, voice assistants can help us make purchases online. Please recollect the experience when asking voice assistants to open a shopping app, check product information, and/or pay for the product, and answer the following questions according to your real situation.” The questionnaire was posted on the online survey platform Wenjuanxing, a professional survey website. It was distributed through social networking sites and online shopping enthusiasts’ chat groups. For the consumers who had never experienced voice shopping, it was necessary to give them an explanation so they could understand the voice-shopping-related items in the questionnaire. Therefore, the researchers chose to distribute the questionnaire offline. The busiest shopping mall was the main location where the survey was conducted. After ensuring that the voluntary respondent had not experienced voice shopping, the researchers showed him or her a 30 s video about voice shopping in a neutral manner and encouraged him or her to try it. After that, they were requested to answer the questionnaire. Invalid responses with short response times and responses that were the same answer to all questions were removed. There were 425 valid responses, including 204 responses from consumers who have once experienced voice shopping and 221 responses from those who have not. To ensure that the data from the two groups could be combined, this study conducted a chi-square test for homogeneity. The results showed that the two groups have the same underlying characteristics in gender (Pearson chi-square = 0, Sig. = 0.229), age (Pearson chi-square = 6.582, Sig. = 0.160), occupation (Pearson chi-square = 2.414, Sig. = 0.660), and education (Pearson chi-square = 3.210, Sig. = 0.523). Finally,

Table 2. Demographic statistics of respondents.

Categories		Freq.			%
		Experience	No Experience	Total
Gender	Male	93	88	181	42.6
	Female	111	133	244	57.4
Age	≤17	3	1	4	0.9
	18–24	52	59	111	26.1
	25–30	82	95	177	41.7
	31–40	20	32	52	12.2
	≥41	47	34	81	19.1
Occupation	Student	54	57	111	26.1
	Business employee	56	72	128	30.1
	Institution staff	52	53	105	24.7
	Freelancer	30	24	54	12.7
	Others	12	15	27	6.4
Education	≤Junior middle school	10	4	14	3.3
	(Vocational) High school	24	27	51	12
	College graduate	25	29	54	12.7
	Undergraduate	104	116	220	51.8
	≥Master	41	45	86	20.2
Frequency of using voice search	Never	0	4	4	0.9
	Sometimes	172	189	361	85
	Every week	17	21	38	8.9
	Everyday	15	7	22	5.2
Number of respondents		204	221	425

summarizes the demographic statistics. The demographic characteristics of the sample (e.g., age, gender, and frequency of using voice search) are similar to those of the population of interest. In addition, the number of responses exceeds general benchmarks for Structural Equation Modeling (e.g., [108]), indicating the sample is adequate and suitable for the research setting.

3.2. Instruments and Measures

The measurement items were selected from previous research and revised to fit the current voice shopping context. All items were measured with a five-point Likert scale. Items and sources are shown in Appendix

Table A1. Items used in this study.

Factor	Items	Wording	Source
Task characteristics	TC1	I want to avoid some messy operations when typing for shopping.	Self-developed
	TC2	I want to find the product more quickly.
	TC3	I want to check shopping information when I want.
	TC4	I want to find the product without typing by myself.
Convenience	CO1	Voice shopping frees our hands.	[120]
	CO2	Voice shopping helps purchase online more quickly.
	CO3	Voice search helps purchase online more easily (e.g., reducing mistyping)
Accessibility	AC1	When it is needed, voice shopping can be easily activated and run.	[69]
	AC2	Voice shopping helps to purchase online anywhere (e.g., when devices like my phone are not in my hands).
	AC3	Voice shopping helps to purchase online anytime.
Context awareness	CA1	When conducting voice shopping, the device can understand what I want with simple formal words.	Self-developed
	CA2	When conducting voice shopping, the device can understand what I want when I speak naturally.
	CA3	When conducting voice shopping, the device can recognize what I said literally.
	CA4	When conducting voice shopping, the device can continuously understand what I mean in the context.
Shyness	SH1	I feel tense when I’m with people I don’t know well.	[121]
	SH2	I do not find it difficult to ask other people for information.
	SH3	I do not feel uncomfortable at parties and other social functions.
	SH4	I do not find it hard to talk to strangers.
Personal innovativeness in information technology	PIIT1	If I heard about a new information technology, I would look for ways to experiment with it.	[92]
	PIIT2	Among my peers, I am usually the first to try out new information technology.
	PIIT3	In general, I am active in trying out new information technology.
	PIIT4	I like to experiment with new information technology.
Task-technology fit	TTF1	The voice shopping method is enough to help people purchase online.	[122]
	TTF2	The voice shopping method is appropriate to help people purchase online.
	TTF3	In general, voice shopping method can meet needs for purchasing online.
Perceived ease of use	PEOU1	I think the voice shopping method is easy to learn.	[52]
	PEOU2	I think it is easy to employ the voice shopping method in practice.
	PEOU3	It would be easy to conduct voice shopping expertly.
	PEOU4	Overall, I think it is easy to conduct voice shopping.
Perceived usefulness	PU1	I believe that the voice shopping method can help me easily find the products that interest me.	[34,52]
	PU2	I believe that voice shopping improves online purchase performance of mine.
	PU3	I believe that voice shopping enhances online purchase effectiveness of mine.
	PU4	Overall, I believe that voice shopping is useful for my online purchase.
Voice shopping intention	VSI1	Given the opportunity, I will use voice shopping.	[123]
	VSI2	I expect to use voice shopping in the near future.
	VSI3	I will frequently use voice shopping in the near future.	[76]
	VSI4	I will recommend others to use voice shopping.

(Appendix A). To increase the validity and reliability, the researchers invited three professors specializing in information technology to check and revise the questionnaire. A pilot test was conducted with 30 undergraduates who have experienced voice shopping to help slightly modify the instruments.

Finally, items for ten constructs were obtained. The construct of task characteristics was measured by four items (TC1-4). For technology characteristics, both convenience (CO1-3) and accessibility (AC1-3) were tested by three items, and context awareness was tested by four items (CA1-4). For individual characteristics, both shyness (SH1-4) and personal innovativeness in information technology (PIIT1-4) were measured by four items. While task-technology fit (TTF1-3) was explained by three items, perceived ease of use (PEOU1-4), perceived usefulness (PU1-4), and voice shopping intention (VSI1-4) were tested by four items. Appendix Table A1 (Appendix A) includes the items.

4. Results

4.1. Measurement Model

The CFA was conducted in AMOS 18.0 to validate the measurement model. According to the criteria of factor loadings and modification indices (MI) [52,109,110,111], the measurement model was refined. Ten items were dropped because their standardized factor loadings were lower than 0.6. To ensure adequate fit with the ratio of chi-square to d.f., this study further checked MI to examine if the overall model fit can be improved. The final model encompassed 26 items (see Appendix

Table A2. Confirmatory factory analysis.

Factor	Items	Factor Loadings			CR	AVE
		Total	Experience	No Experience	Experience/ No Experience	Experience/ No Experience
Task characteristics	TC2	0.719	0.741	0.663	0.653/0.741	0.486/0.593
	TC4	0.714	0.650	0.864
Convenience	CO1	0.817	0.838	0.797	0.878/0.842	0.783/0.728
	CO2	0.921	0.929	0.906
Accessibility	AC1	0.839	0.810	0.859	0.865/0.927	0.681/0.81
	AC2	0.847	0.785	0.901
	AC3	0.907	0.878	0.938
Context awareness	CA1	0.809	0.790	0.816	0.776/0.868	0.635/0.768
	CA2	0.863	0.803	0.933
Shyness	SH2	0.758	0.780	0.707	0.732/0.679	0.577/0.514
	SH4	0.724	0.739	0.727
Personal innovativeness in information technology	PIIT1	0.869	0.855	0.882	0.876/0.875	0.701/0.7
	PIIT2	0.825	0.827	0.826
	PIIT4	0.817	0.830	0.8
Task-technology fit	TTF1	0.912	0.887	0.929	0.865/0.907	0.762/0.83
	TTF2	0.875	0.859	0.893
Perceived ease of use	PEOU1	0.766	0.795	0.74	0.925/0.92	0.756/0.744
	PEOU2	0.893	0.873	0.908
	PEOU3	0.879	0.892	0.861
	PEOU4	0.922	0.914	0.929
Perceived usefulness	PU2	0.879	0.871	0.884	0.908/0.941	0.767/0.841
	PU3	0.917	0.884	0.941
	PU4	0.899	0.872	0.926
Voice shopping intention	VSI2	0.918	0.899	0.935	0.911/0.945	0.774/0.851
	VSI3	0.904	0.888	0.91
	VSI4	0.89	0.851	0.922

Note: CR and AVE values of total responses are not present in this table since they have been mentioned before.

(Appendix A)). The values, including the ratio of chi-square (767.860) to d.f. (254) at 3.023, GFI at 0.879, CFI at 0.941, IFI at 0.942, TLI at 0.925, RMR at 0.028, and RMSEA at 0.069 indicate a good fit statistics for the measurement model.

According to the criteria of convergent validity, reliability, and discriminant validity [112,113,114], the validity and reliability of all measures are acceptable in this study. First, the Appendix shows that all factor loadings exceed 0.6 at a significant level, revealing strong convergent validity. Second, reliability was tested by composite reliability (CR) and average variance extracted (AVE). According to

Table 3. Construct reliability and discriminant validity.

	CR	AVE	TC	CO	AC	CA	SH	PIIT	TTF	PEOU	PU	VSI
TC	0.679	0.513	0.716
CO	0.862	0.758	0.528	0.871
AC	0.899	0.748	0.498	0.682	0.865
CA	0.823	0.7	0.461	0.547	0.566	0.837
SH	0.709	0.549	−0.383	−0.4	−0.362	−0.351	0.741
PIIT	0.875	0.701	0.394	0.438	0.546	0.425	−0.588	0.837
TTF	0.888	0.799	0.514	0.602	0.713	0.553	−0.506	0.524	0.894
PEOU	0.923	0.752	0.408	0.52	0.672	0.57	−0.532	0.489	0.729	0.867
PU	0.926	0.807	0.47	0.538	0.697	0.572	−0.385	0.479	0.693	0.733	0.898
VSI	0.931	0.817	0.404	0.626	0.706	0.589	−0.539	0.523	0.725	0.71	0.777	0.904

Note: TC (task characteristics), CO (convenience), AC (accessibility), CA (context awareness), SH (shyness), PIIT (personal innovativeness in information technology), TTF (task-technology fit), PEOU (perceived ease of use), PU (perceived usefulness), and VSI (voice shopping intention).

, the CRs of all constructs exceed the threshold of 0.6, and all AVEs are more than 0.5. The reliability can be accepted. Finally, the discriminant validity was measured by comparing correlations between constructs with the square root of AVE for each construct. Table 3 shows that the square root of AVE for each construct (on the diagonal) is greater than its correlation coefficients with other constructs. Thus, it is possible to argue that a reasonable extent of discriminant validity is examined.

4.2. Structural Model

AMOS 18.0 was used to assess the structural model. The fit indices (χ2/d.f. = 3.104, p < 0.001, GFI = 0.873, CFI = 0.937, IFI = 0.937, TLI = 0.922, RMR = 0.034, and RMSEA = 0.07) suggest a good model fit. Thus, this study could further examine the path coefficients of the structural model.

Table 4. Results of path coefficients (integrated model based on TTF theory and TAM).

Path				β	Result	R2
H7a	Task characteristics	→	Task-technology fit	0.075	Not supported	0.628
H8a-1	Convenience			0.09	Not supported
H8b-1	Accessibility			0.458 ***	Supported
H8c-1	Context awareness			0.132 *	Supported
H9a-1	Shyness			−0.248 ***	Supported
H8a-2	Convenience	→	Perceived ease of use	−0.08	Not supported	0.633
H8b-2	Accessibility			0.313 ***	Supported
H8c-2	Context awareness			0.175 **	Supported
H9a-2	Shyness			−0.239 ***	Supported
H9b	Personal innovativeness in information technology			−0.047	Not supported
H4	Task-technology fit			0.353 ***	Supported
H7b	Task characteristics	→	Perceived usefulness	0.069	Not supported	0.645
H8a-3	Convenience			−0.007	Not supported
H8b-3	Accessibility			0.254 ***	Supported
H8c-3	Context awareness			0.117 *	Supported
H9a-3	Shyness			0.036	Not supported
H5	Task-technology fit			0.163 *	Supported
H6	Perceived ease of use			0.37 ***	Supported
H1	Task-technology fit	→	Voice shopping intention	0.316 ***	Supported	0.695
H2	Perceived ease of use			0.159 **	Supported
H3	Perceived usefulness			0.445 ***	Supported

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

summarizes the final results.

For task, technology, and individual characteristics, the influences of accessibility (β = 0.458, p < 0.001), context awareness (β = 0.132, p < 0.05), and shyness (β = −0.248, p < 0.001) on task-technology fit, influences of accessibility (β = 0.313, p < 0.001), context awareness (β = 0.175, p < 0.01), and shyness (β = −0.239, p < 0.001) on PEOU, and influences of accessibility (β = 0.254, p < 0.001) and context awareness (β = 0.117, p < 0.05) on PU are significant; H8b-1, H8c-1, H9a-1, H8b-2, H8c-2, H9a-2, H8b-3, and H8c-3 are supported. The coefficients indicate that accessibility is worthy of much attention. While context awareness plays an important role in fit, the individual characteristic shyness exhibits a stronger negative impact, causing consumers to hesitate in voice shopping usage. The remaining paths are not significant, indicating that H7a, H7b, H8a-1, H8a-2, H8a-3, H9a-3, and H9b are rejected. Among them, there are two most surprising results. First, task characteristics cannot affect task-technology fit and perceived usefulness, indicating that consumers currently may not believe voice shopping is a good choice for conducting shopping tasks. Second, PIIT cannot significantly affect PEOU. The possible reason for this might lie in the operational complexity and defects of the early-stage voice shopping. Both of them can offer insights into reasons consumers do not use voice shopping and how to improve the corresponding experience, which will be further investigated in the Discussion section.

For the three constructs measuring objective and subjective fit, all of them—task-technology fit (β = 0.316, p < 0.001), PEOU (β = 0.159, p < 0.05), and PU (β = 0.445, p < 0.001)—positively influence the intention to use voice shopping. Thus, H1, H2, and H3 are supported. The above results reveal that the subjective fit between technology and individuals plays a more critical role in voice shopping acceptance, emphasizing the importance of a nuanced experience. In addition, the influences of task-technology fit on PEOU (β = 0.353, p < 0.001), task-technology fit on PU (β = 0.163, p < 0.05), and PEOU on PU (β = 0.37, p < 0.001) are proven, supporting H4, H5, and H6.

Moreover, the explained variances (R²) of task-technology fit, PEOU, PU, and voice shopping intention are 0.628, 0.633, 0.645, and 0.695, respectively. The integrated model can strongly explain user acceptance of voice shopping.

To verify whether uniting the TTF theory and TAM can offer a better understanding of voice shopping intention than the individual model alone, the researchers further summarized the results from the evaluation of the TTF theory and TAM, which are shown in

Table 5. Results of path coefficients (separating TTF theory and TAM).

Path				β	R2
TTF theory	Task characteristics	→	Task-technology fit	0.111	0.633
	Convenience			0.141 *
	Accessibility			0.3496 ***
	Context awareness			0.188 ***
	Task-technology fit	→	Voice shopping intention	0.772 ***	0.596
TAM	Convenience	→	Perceived ease of use	−0.029	0.594
	Accessibility			0.473 ***
	Context awareness			0.223 ***
	Shyness			−0.331 ***
	Personal innovativeness in information technology			−0.039
	Convenience	→	Perceived usefulness	0.036	0.636
	Accessibility			0.327 ***
	Context awareness			0.143 **
	Shyness			−0.008
	Perceived ease of use			0.406 ***
	Perceived ease of use	→	Voice shopping intention	0.313 ***	0.666
	Perceived usefulness			0.559 ***

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

. The analysis shows that for the TTF theory, the explained variance (R²) of voice shopping intention is 0.596; for TAM, the explained variance (R²) of voice shopping intention is 0.666. Thus, the integrated model provides a better result in the voice shopping setting.

The 425 respondents consist of consumers who have once experienced voice shopping (n = 204) and those who have not (n = 221). Studies on technology acceptance have found that individuals may pay attention to different factors over time with increasing experience [115]. To further test whether there is a difference in the path coefficients between the two groups, this study conducted a multi-group analysis and a t-test (see

Table 6. Path coefficient comparison for people with prior experience and no experience.

Path			Experience	No Experience	t-Test	Difference
			β	β
TC	→	TTF	−0.001	0.173 *	0.857	No
CO			0.275 *	−0.009	1.954	No
AC			0.335 *	0.475 ***	0.837	No
CA			0.196	0.098	0.636	No
SH			−0.198 *	−0.338 ***	1.062	No
CO	→	PEOU	−0.334 *	0.009	2.055 *	YES
AC			0.42 *	0.323 ***	0.468	No
CA			0.239 *	0.126	0.698	No
SH			−0.208 *	−0.297 **	0.493	No
PIIT			−0.086	−0.066	0.138	No
TTF			0.499 ***	0.264 **	1.256	No
TC	→	PU	0.058	0.154 *	0.546	No
CO			−0.006	0.006	0.089	No
AC			0.337 *	0.226 **	0.69	No
CA			0.065	0.126 *	0.111	No
SH			0.102	−0.071	1.329	No
TTF			0.112	0.107	0.032	No
PEOU			0.426 ***	0.347 ***	0.65	No
TTF	→	VSI	0.492 ***	0.255 ***	1.461	No
PEOU			0.015	0.194 **	1.23	No
PU			0.412 ***	0.461 ***	0.327	No

Note: TC (task characteristics), CO (convenience), AC (accessibility), CA (context awareness), SH (shyness), PIIT (personal innovativeness in information technology), TTF (task-technology fit), PEOU (perceived ease of use), PU (perceived usefulness), and VSI (voice shopping intention). * p < 0.05, ** p < 0.01, *** p < 0.001.

). Distinctions do exist between the two groups based on the path coefficients, but only the influence of convenience on PEOU showed a statistically significant difference.

5. Discussion and Implications

5.1. Discussion of Findings

To improve the online shopping experience, retailers continue to leverage emerging and promising technologies, including voice technology. However, a high percentage of consumers still rarely use voice shopping to date. This study applies the concept of fit among task, technology, and individual characteristics to figure out this phenomenon and help encourage the acceptance of voice shopping. It puts forward that the voice shopping tool should achieve both an objective fit with online shopping tasks and a subjective fit with users. Based on this, this study ultimately integrates the TTF theory and TAM to establish a research model for the acceptance of voice shopping. The data analysis results provide the following insights into the two research questions proposed in the introduction.

For RQ1: Why do many consumers not use voice shopping currently? The results offer clues from the following three aspects.

First, task characteristics cannot significantly influence task-technology fit and usefulness, suggesting that average consumers still do not believe that voice shopping can meet their shopping needs. Task characteristics are the requirements that people want to or have to perform in online shopping [61]. As summarized, an increasing number of consumers expect adaptable, immediately accessible, and customized shopping experiences. Voice shopping is expected to support corresponding requirements since it enables hands-free search, on-the-go purchases, and spoken language understanding. However, unlike many findings [52,67,116], results suggest that task characteristics do not lead to either the task-technology fit or the perceived usefulness. It indicates that people do not consider voice shopping a viable option for meeting their online shopping requirements.

Second, the convenience of voice shopping fails to persuade consumers. Generally, technology that brings convenience can positively influence its use (e.g., [69]). However, this study shows that convenience, one of the most important characteristics of voice shopping [19], has no effect on task-technology fit, PEOU, and PU. Convenience shows the degree to which voice shopping can provide time and effort utility, primarily by enabling purchases without typing. The conversational interface makes it easier to use, which is commonly recognized as the most important advantage of voice technology [117]. However, current consumers may have been used to typing for their online purchases. Compared with waiting for the speech-to-text process and potential interruptions caused by technological limitations, they easily feel it faster and more comfortable to open a shopping app and search with their hands. Furthermore, when conducting voice shopping, consumers often cannot directly view product pictures, as text and links are frequently used to display product information. The above causes less fit between online shopping tasks and voice shopping, as well as perceived ease of use and perceived usefulness. Therefore, service providers need to create a close match-up between shopping scenarios and voice shopping. To help users adapt to voice shopping, they may start with the products that require less browsing, such as daily necessities, food, and electronics. For other products, they can enrich search results, like adding pictures.

Third, important personal factors cannot play positive roles in encouraging user acceptance. A notable finding is that PIIT cannot influence PEOU, which is inconsistent with many studies (e.g., [50,61]). A possible reason is that consumers with highly innovative characteristics are generally open to voice shopping and are active in trying it. However, voice shopping is more complex than other voice commands, such as navigation. Moreover, early-stage voice shopping still needs further improvement. As a result, consumers must make efforts to have a better understanding of it, hindering PIIT from having a positive influence on ease of use. The result suggests that service providers should fully consider the feedback from early users. They can invite consumers to participate in voice shopping testing and optimization. Another remarkable finding is related to shyness. Despite no effect on PU, shyness negatively influences task-technology fit and PEOU, delaying user acceptance. Thus, helping people overcome embarrassment calls for extra attention. As a personality characteristic [94], shyness is hard to change directly and rapidly. Service providers may leverage external effects, such as others’ adoptions, to attract shy consumers. For example, they can attach unique tags like “coupons for voice shopping” to the retailer and “voice shopping user” to the consumers. In addition, not only shy individuals but many people can feel embarrassed when talking to those they are not familiar with. According to Rhee and Choi [17], the voice assistant’s social role as a friend can positively influence a consumer’s attitude toward the recommended product. Thus, service providers can emphasize the voice assistant’s role as a private friend. For example, users can be allowed to design personalities or shopping styles for their voice assistants.

For RQ2: How can the voice shopping experience be improved to encourage more usage? The rest of the results provide insights from the following three aspects.

First, service providers of voice assistants and retailers should establish a cooperation mechanism. The analysis reveals that task-technology fit, PEOU, and PU positively affect voice shopping intention, with PU exerting a greater effect than the other two factors. That means both objective fit between tasks and technology and subjective fit between individual and technology matter; the final acceptance of voice shopping lies more in the subjective perception of consumers. Therefore, service providers should create a more nuanced experience. Generally, consumers expect basics such as finding what they want quickly and buying it at a fair price. Voice assistants then must understand their users. To do so, service providers and retailers need to cooperate regarding the sharing of customer data. In addition, to date, many voice assistants cannot support a seamless and smooth shopping trip [17,118], especially those on mobile phones. Service providers of voice assistants and retailers should collaborate to overcome this limitation. In particular, when they are not the same entity, they need to carefully negotiate matters, such as on whose platform the products can be browsed.

Second, technical improvement requires attention. Significant relationships between technology characteristics and fit can offer clues for voice shopping upgrades. The results show that both accessibility and context awareness are catalysts for task-technology fit, PEOU, and PU. Accessibility refers to the ability that voice shopping allows consumers to reach shopping information and services anytime, anywhere, and in any situation. Thus, keeping users connected to their voice assistants is important. Service providers can enhance the ease of activation, expand the voice recognition range, and improve listening skills in more complex work environments, such as driving environments and noisy settings. Context awareness refers to the ability to understand what users say, including environmental factors, mood, and even intent. High context awareness often generates a personalized shopping process, which stimulates the use of voice shopping. Thus, voice assistants should learn to obtain more details from their users and enhance comprehension ability. To achieve this, service providers must develop the skills of voice assistants to handle longer sentences, continuous conversations, and emotion recognition.

Third, it is necessary to segment consumers based on the experience of voice shopping. Data analysis presents differences between online consumers who have once experienced voice shopping and those who have not. Service providers and retailers should give special attention to the convenience feature. It positively influences task-technology fit but hinders the PEOU of consumers who have once experienced voice shopping. This result highlights the importance of encouraging consumers to use voice shopping more frequently and helping them adapt to this new shopping method again. There is no statistically significant difference in other relationships between the two groups. However, it is still worth noting the positive influences of task characteristics on task-technology fit and PU, as well as the influence of PEOU on voice shopping intention among consumers who have not experienced voice shopping. The impact of task characteristics suggests that service providers need to attract those consumers by evoking their demand for an adaptable, immediately accessible, and customized shopping experience. The impact of PEOU stresses that service providers need to provide facilitating conditions to make voice shopping easy to operate. For example, a guided interaction, such as “You may speak like…”, could be more workable than a direct “Sorry, I do not understand.”

5.2. Theoretical and Practical Implications

5.2.1. Theoretical Implications

This work provides theoretical implications mainly from two aspects. First, it investigates the acceptance of voice shopping by considering task, technology, and individual characteristics simultaneously, which fills the research gap in this domain. As mentioned above, studies on the acceptance of voice shopping are still scarce. Although they emphasized the importance of technology and individual characteristics, most technology characteristics are general characteristics of voice assistants rather than those for shopping purposes, and little is known about the influences of individual characteristics. Under these circumstances, the current study elaborates on task, technology, and individual characteristics and investigates their roles in voice shopping acceptance. The results enrich the understanding of the factors influencing voice shopping acceptance. In particular, the surprising findings, such as the insignificant effects of task characteristics, convenience, and PIIT, challenge previous research. Future studies can explore whether there are hidden moderating variables. While carefully examining the impacts of technology characteristics and individual characteristics, this study also updates the integrated model of the TTF theory and TAM. Moreover, the data analysis result provides evidence for the robustness of the integrated model in the voice shopping setting.

Second, this study categorizes fit into objective and subjective fit and extends this concept into the voice shopping context. Fit is a critical basis for understanding technology acceptance, but divergence still exists about what constitutes fit [42]. The current study proposes that fit can be understood from two main aspects: objective fit between task and technology and subjective fit between individual and technology. To measure them, this work leverages task-technology fit, PEOU, and PU. It provides a fresh theoretical perspective not only for understanding fit but also for integrating the TTF theory and TAM to discover technology acceptance. The results indicate that one of the subjective fit aspects—PU—plays a more critical role in voice shopping acceptance. This further highlights the necessity of more nuanced investigations into consumers’ individual characteristics in this setting. In addition, fit so far has been mainly used in information technology, not digital technologies [37] or AI applications. This study brings it into the voice shopping area, paving the way for further investigation on adoption and continuous use behavior in emerging technology fields. For example, scholars can apply the fit concept and the research model to explore human–AI collaboration.

5.2.2. Practical Implications

The practical implications of this study can be understood from two streams. First, this study provides insights into why consumers still use voice shopping less. Data analysis infers three reasons: (a) average consumers still do not consider that their online shopping needs can be satisfied by voice shopping. (b) The convenience of voice shopping fails to persuade consumers. (c) Important personal initiatives do not play positive roles in user acceptance. Accordingly, service providers and retailers should make voice technology more suitable for online shopping tasks. In addition, they must make efforts to encourage consumers to try voice shopping and take advantage of the power of early adopters. When promoting emerging technologies or tools, marketers should also pay attention to such issues, as consumers do not always behave as expected in terms of technology use. It is important to figure out why they do not accept the technology or tool in the early stage.

Second, suggestions are given for increasing the acceptance of voice shopping. The results highlight that retailers and service providers should attach importance to the following three aspects: (a) the cooperation mechanism between the two parties. Updating the voice shopping experience is essential to persuade consumers that speaking is a suitable and useful method for online shopping. Thus, service providers and retailers must collaborate to understand their users better and overcome the current limitations of voice shopping. To achieve this, they must consider some issues seriously, such as data sharing and the platform for displaying information search results due to incompatibilities. (b) The next aspect is technical improvements. Voice shopping is still in the early stages, and its performance requires further improvement. Relationships between technology characteristics and task-technology fit, PEOU, and PU give clues. For example, to enhance the accessibility of voice assistants, service providers must expand their voice recognition range and improve listening skills in more complex work environments. (c) Last is customer segmentation based on the experience of voice shopping. This study reveals that there are differences between online consumers who have experienced voice shopping and those who have not. Service providers and retailers can draw those consumers via distinct strategies. For example, they should first evoke the demand for adaptable, immediately accessible, and customized shopping experiences among consumers who have not experienced voice shopping.

6. Future Research Directions

This study has leveraged fit among tasks, technology, and individuals to understand consumers’ intention to use voice shopping. It proposes a research model that integrates the TTF theory for objective fit and the TAM for subjective fit, where task, technology, and individual characteristics lead to the two aspects of fit, which ultimately influence voice shopping intention. Based on the results, this study explains why many consumers do not accept voice shopping and offers practical suggestions to improve the voice shopping experience to encourage more usage. Overall, the findings from this work enhance the understanding of voice shopping acceptance.

This study has certain limitations. First, as a relatively novel study on voice shopping acceptance, this study focuses more on the various determinants and their influences but pays limited attention to the mediating effects of subjective fit and objective fit. A future study can collect more survey data or real data from consumers in all walks of life to examine their influences, as well as verify and improve the current model. In addition, the technology and individual characteristics may need further investigation. For example, voice assistants exhibit human-like cues that might make consumers consider them as shopping companions. The current study only applies context awareness to represent this aspect, which could be enriched with other human-like characteristics. Furthermore, the results suggest that context awareness cannot significantly affect the task-technology fit and PU of consumers who have once experienced voice shopping, which may be due to the privacy risks that context awareness poses. Therefore, although studies have disclosed that positive attributes or feelings toward voice shopping are taking precedence over negative ones (e.g., [8,19,70]), important negative technology characteristics, like privacy invasion, may need further investigation.

Second, this study focuses on the task, technology, and individual characteristics without much consideration of possible environmental features. The environment-technology fit has not received much attention [44], which may provide a potential perspective for improving the current integrated model. For example, this study finds that shyness hinders voice shopping intention, indicating that the role of social influence deserves attention. Future work can introduce environmental factors as moderators.

Third, some demographic and contextual variables are worth noting. For example, (a) cultural differences might matter. Previous studies on technology acceptance also took cultural differences into account. Taking Pavlou and Chai [119] as an example, China is high on collectivism and power distance, while the U.S. is relatively low on collectivism and power distance, which may vary determinants of their voice shopping acceptance. (b) The product category and shopping scenarios can also be important. Different perspectives have been applied to discuss categories like fashion [28] and grocery [20]. The researchers aim to investigate whether there is heterogeneity in determinants across various product categories and shopping scenarios.