From Perception to Purchase: How AI Literacy Shapes Consumer Decisions in AI-Generated Sponsored Vlogs Across Products and Services

Abstract

This study investigates the perception-to-purchase journey by examining how consumer artificial intelligence (AI) literacy influences the effectiveness of AI-generated sponsored vlogs (AISVs), an emerging digital marketing format. Using survey data from 413 consumers and structural equation modeling, we develop and test the AI Literacy Perception–Decision Model (AILPDM). Results show that AI literacy affects information adoption through three pathways: emotional value, information usefulness, and source credibility. Separate SEM analyses further suggest that the direct effect of AI literacy on purchase intention was observed in experiential service AISVs, whereas in tangible product AISVs the effect operated mainly through information adoption. The AILPDM framework advances marketing theory by tracing a decision pathway from AI literacy, through perceived value and information adoption, to purchase intention, thereby demonstrating how technological competence evolves from a cost barrier into a cognitive resource that shifts source credibility evaluation from peripheral to central processing. For practitioners, the findings suggest differentiated strategies: Marketers of experiential services should emphasize anthropomorphic elements, whereas marketers of tangible products should prioritize technological transparency to foster consumer trust.

1. Introduction

With the rise of generative artificial intelligence (AI), automated video creation has entered a new era [1]. In February 2024, OpenAI released Sora, the first text-to-video generation model, marking a milestone in video production [2]. Since then, the technology has advanced rapidly. In February 2025, the University of Hong Kong and TikTok launched Goku Plus for video advertising [3], and in May 2025, Google Veo 3 achieved synchronized audio–video generation [4]. These large-scale video generation models represent a breakthrough for AI advertising, with vlogs—one of the most popular forms of social media content—most directly affected.

Vlogs, or video blogs, can be conceptualized as an audio-visual form of electronic word-of-mouth (eWOM) in which individuals produce and disseminate self-created videos [5]. Vloggers share their daily lives, experiences, opinions, and expertise with audiences [6], covering themes such as sports, education, forensics, entertainment, travel, advertising, social media, and gaming [7]. vloggers range from non-professional individuals and independent producers to professional influencers and creators supported by multi-channel networks [8]. With increasing commercialization, sponsored vlogs have become an important vehicle for brand marketing [9]. Recently, AI has begun transforming vlog production, as video generation models reshape creation processes [6], giving rise to AI-generated sponsored vlogs (AISVs). AI offers clear practical benefits to this medium: it shortens production cycles, reduces costs nearly 100-fold compared to traditional advertising [10], frees marketing teams to focus on strategy, and decreases reliance on human resources, thereby driving innovation in marketing and business models [11].

However, the rapid spread of generative AI in marketing also presents challenges, especially for AISVs. First, AI-generated content (AIGC) can be misused for deepfakes and misinformation, undermining perceptions of content usefulness [12]. Second, AI novelty may trigger reliability questioning [13] and compromise consumer privacy through big data analysis [14]. Finally, whether AISVs can effectively convey emotional value—crucial to young consumers’ decision-making [15]—remains uncertain. These issues have led scholars and practitioners to call for closer scrutiny of AI’s ethical and societal risks [16] and for the establishment of standards and guidelines to ensure its responsible use [11].

With over 75% of consumers now using AI-enabled services or devices [17], AI-driven transformation has created fundamental differentiation among consumer groups, just as the internet era distinguished digital immigrants from digital natives [18]. At present, most consumers can be regarded as AI immigrants—individuals whose consumption habits were originally formed before the widespread adoption of AI and have since been adjusted through gradual exposure to it. However, the population exhibits substantial heterogeneity in AI literacy [19]—the knowledge, skills, and attitudes that enable users to understand, evaluate, and interact with AI [20]. This disparity has generated an AI divide [21], reflected in differing levels of adaptation and benefit across consumer segments. In the coming decade, the rise of AI natives is expected—consumers socialized in AI-saturated environments who treat AI as an inherent element of decision-making and value formation. Within social e-commerce, such heterogeneity in AI literacy intensifies regulatory challenges and underscores the need for adaptive marketing governance [19].

Traditional regulations assume consumer information asymmetry, with consumers typically at a disadvantage [22] (pp. 1–10). In AISV contexts, however, high-literacy consumers—both immigrants and natives—can effectively identify and evaluate AIGC [23], while low-literacy consumers remain disadvantaged. This divergence makes uniform protection standards inadequate; restrictions may overregulate AI natives while failing to safeguard traditional consumers [24]. According to Coffin’s [25] ontological, technical, and ethical (OTE) framework, resolving this dilemma requires clarifying AISVs’ characteristics and consumers’ differentiated responses before establishing robust ethical regulations. Despite this need, research on AISV consumer responses remains structurally imbalanced. Studies focus heavily on the acceptance of AI-powered tools, yet it remains insufficiently understood how the integration of AI technologies into advertising shapes consumers’ attitudes and behaviors [26]. Particularly, attitudes toward AI-generated advertising [27] are insufficiently studied, leaving gaps in understanding AI’s influence on marketing [1]. Current AI literacy research emphasizes conceptualization, assessment, and cultivation, with most studies limited to students or professionals in specialized fields (e.g., doctors, developers, teachers) [23]. Applied research on consumer AI literacy remains scarce.

Crucially, preliminary evidence suggests that differences in AI literacy [28], leading to an AI divide [21], may significantly shape AI advertising effectiveness. Familiarity with AI alters consumer response patterns [16], with varying digital literacy levels producing distinct attitudes toward AI marketing [29]. Yet these fragmented findings have not been integrated into a systematic framework. Particularly, theoretical models lack explanations of how AI literacy influences cognitive assessment, emotional response, and source evaluation, as well as how these effects vary across AISV types.

Research Objectives and Questions

This study primarily aims to develop and empirically test an integrated theoretical framework explaining how AI literacy shapes consumer responses to AISVs and subsequent purchase intentions, while also examining how these relationships vary across product contexts.

To address this problem, this study investigates three research questions (RQs):

RQ1: How does consumer AI literacy influence perceptions and information adoption from AISVs?

This question aligns with the technology problem dimension of the OTE framework, with emphasis on differential consumer response patterns. Prior studies recognize the need to examine how AI literacy shapes cognitive responses [30] and acceptance of AI-driven advertising [31]. Nonetheless, no framework systematically integrates AI literacy with AISV response mechanisms across cognition, emotion, and source evaluation, highlighting a notable theoretical gap in AI marketing research.

RQ2: What are the direct and indirect effects of AI literacy on purchase intention in AISV contexts?

This question extends inquiry from cognitive influence to purchase decision formation. While existing research addresses AI marketing applications, it has not fully explained their long-term behavioral impact [32]. A systematic understanding of how AI literacy shapes purchase intention through multiple mediating pathways is lacking. This study evaluates these pathways, constructing an integrated framework that traces the influence of technological literacy through purchase decisions, thereby offering guidance for multi-path marketing strategies and advancing understanding of decision processes in AI video marketing.

RQ3: How do these relationships differ across AISV types (products and services)?

According to Kumar and Singh [7], AISVs can be categorized by video-generation technique into actorless and actor-present forms, reflecting two principal modes of generative video production. Building on this foundation, the present study further differentiates between product-oriented and service-oriented AISVs to capture variation in the nature of the advertised offerings. In line with prior service-marketing literature, products refer to physical goods that can be owned, stored, and evaluated primarily based on functional and quality attributes, whereas services are largely intangible, produced and consumed simultaneously, and valued predominantly through consumers’ lived experiences and affective engagement rather than physical ownership [33]. This distinction is theoretically important because consumers process information about products through attribute-based evaluation, whereas they approach services through experiential or affective processing [34]. Prior studies confirm that such differences shape consumer preferences: consumers of search products prefer collaborative filtering recommendations, while experiential product consumers favor content-based filtering [35]. Similarly, virtual influencers are more persuasive for technological than for non-technological products [36]. Yet no analytical framework integrates AI video-generation modes, the product–service distinction, and consumer AI literacy into a unified model. This study addresses this gap by comparing actorless tangible product-oriented AISVs (e.g., clothing, food, household items) with avatar-mediated, experiential service-oriented AISVs (e.g., dance, sports, fitness). It explores structural differences in AI literacy pathways across these two categories of AISV and provides a theoretical basis for targeted AISV marketing strategies.

This research contributes to AISV literature by introducing the AI Literacy Perception–Decision Model (AILPDM), which reconceptualizes consumer technological competence as a dynamic antecedent that transforms from a processing barrier into a cognitive resource during AISV evaluation, thereby extending VAM by moving beyond the notion of technology as a pure adoption cost and highlighting its dual role as both a transitional burden and an enduring resource shaping value perceptions. Drawing on the Elaboration Likelihood Model, the study demonstrates that AI-literate consumers process technological transparency as a central rather than a peripheral cue, thus refining ELM by repositioning source credibility from a peripheral signal toward substantive evidence of honesty and expertise in AI-mediated persuasion pathways. The investigation reveals systematic differences in AI literacy mechanisms between actorless tangible products and avatar-mediated experiential services, establishing AILPDM’s contextual boundaries while validating AI literacy measurement scales in commercial contexts. Collectively, these advances enhance theoretical understanding of the drivers of AISV effectiveness and provide practitioners with frameworks for developing targeted implementation strategies across product and service categories.

The remainder of this paper is organized as follows: Section 2 reviews the literature on AISV and AI literacy to establish theoretical foundations and develop hypotheses. Section 3 outlines the methodology, and Section 4 presents the empirical results. Section 5 discusses the findings and their theoretical implications (Section 6) and managerial implications (Section 7). Section 8 addresses limitations and future research directions, and Section 9 concludes.

2. Literature Review, Research Hypotheses, and Conceptual Model

2.1. AISVs

Defining AISVs requires an understanding of their fundamental components. Sponsored vlogs integrate advertising into video blogs [37], while the term “AI-generated” refers to content produced through generative AI techniques. This field encompasses algorithms that learn underlying patterns and structures from training data to create new data instances resembling a given dataset [9]. It is widely applied in image and video synthesis [11]. Advanced models such as Video-GPT generate video content from text descriptions [6], creating new opportunities for vlog production [7]. The diffusion of generative video technologies is already evident in platform adoption patterns. Synthesia, for instance, is reported to have attracted over one million users worldwide [38]. Runway (runwayml.com), frequently cited as a leading creative tool, reached an estimated 8.43 million global visits in August 2025, reflecting a 6.78% increase from the previous month according to third-party traffic data [39]. In parallel, ByteDance’s Dreamina AI, launched in 2024 and closely tied to TikTok’s ecosystem, was estimated to reach the scale of 5–10 million monthly active users by the same period, with a month-on-month growth rate of 10.8% [40]. These adoption trends underscore the growing integration of AI-generated content into social platforms, creating new formats of advertising and sponsorship [41].

Against this backdrop, we define AISVs as videos created by AI algorithms according to vlogger instructions, using sponsor-provided product information and materials, embedded with purchase links, and published on social media in exchange for sponsorship compensation. This definition highlights three core elements: the AI-driven creation process, vlogger’s guiding role, and commercial sponsorship nature. Beyond these structural components, AISVs can also be situated within advertising’s executional traditions. As an advertising format, AISVs also inherit executional elements long established in traditional television advertising—such as verbal and visual components [42], music [43], voice-overs and presenters [44], product display [45], humor and emotional appeals [46]. By incorporating these traditional elements into AI-generated content, AISVs extend classical advertising practices into a novel, algorithm-driven context.

According to Kumar and Singh’s [7] framework, AISVs can be divided into two main categories. The first is actorless videos, produced without human participation and appearing either in static form (e.g., product stills, landscape views) or dynamic form (e.g., natural scenes with moving elements). The second is actor-present videos, which feature real or virtual actors. These may include partially dynamic content, where the actor moves while the background remains static or semi-dynamic, and fully dynamic content, where both the actor and the environment change simultaneously. This classification is not merely theoretical; recent advances already demonstrate both forms. For example, the latest Goku+ foundation models can generate actorless videos from product images, create hyper-realistic marketing avatars from text, and even produce realistic product–human interaction videos [3], demonstrating the practical realization of both actorless and actor-present forms.

As a form of AI advertising, AISVs also inherit four key characteristics identified by Wu et al. [47]: data-driven, tool-enabled, process-synchronized, and efficiently executed. These features provide distinct advantages—precise audience targeting, creative automation, enhanced efficiency, and real-time content optimization—supporting their emergence as a powerful marketing tool.

2.2. AI Literacy

AI literacy is emerging as a key capability in the digital economy. Long and Magerko [20] describe it as “a set of competencies required to evaluate, interact with, and use AI systems,” while Almatrafi et al. [23] expand this into six dimensions: Recognize, Know and Understand, Use and Apply, Evaluate, Create, and Navigate Ethically. This framework provides the theoretical basis for examining how consumer AI literacy shapes responses to AISVs. Recent reviews have observed that AI literacy is predominantly measured through self-assessment instruments, while knowledge-based approaches are comparatively rare [48]. This pattern underscores the prevalence of perceived literacy as the main operational form adopted in contemporary studies.

As AI penetrates marketing, literacy differences have become a critical factor in consumer information processing and decision-making. Consumers lacking digital knowledge not only face barriers to online advertising services but may also overlook concerns about algorithmic predictions [25]. Such disparities influence how consumers perceive, interpret, and adopt AISV content. Recent scholarship calls for extending digital literacy to include AI cognition [31]. While prior research highlights the importance of AI literacy for marketers [49] and corporate executives [50], studies focusing on its impact on consumer responses to AI advertising remain limited. This gap is particularly significant as AI literacy diffusion coincides with the rapid commercialization of AIGC applications.

Consumer expertise levels may further shape perceptual frameworks for AISVs. Greiner and Lemoine [28] find that non-expert users display polarized trust in AI systems, whereas Ratta et al. [10] show that professionals often regard AI advertising as more effective than human-created content in driving engagement and purchases. These findings raise a central question: how do consumer AI literacy levels influence perceptions, information adoption, and purchase intention in AISV contexts? This study addresses this gap by systematically analyzing the pathways through which AI literacy, as a key independent variable, affects consumer decision-making.

2.3. Theory and Models

This research integrates the Value-Based Adoption Model (VAM) and Elaboration Likelihood Model (ELM) to construct a framework for analyzing consumer responses to AISV marketing. As advertising operates as a brand-centered persuasive communication tool [25], AISVs function both as applications of AI technology and as vehicles for advertising delivery, creating a hybrid context in which technological acceptance and advertising persuasion intersect, necessitating multi-dimensional theoretical perspectives.

VAM was developed to explain user acceptance of emerging technologies. It defines behavioral intention as a cognitive appraisal of perceived value derived from a trade-off between benefits and sacrifices—encompassing both monetary and non-monetary factors [51]. Over time, the model has been extended to both traditional and social-media marketing contexts [52], where users evaluate advertising and technological elements simultaneously through their value perceptions. The model has also been validated in AI-based wearables [53] and augmented-reality (AR) shopping contexts [54], demonstrating its adaptability across digital media environments. VAM provides the theoretical backbone for this research by positioning perceived value—rather than functional utility alone—at the center of adoption decisions and by integrating both cognitive and affective dimensions. Within this model, consumers assess AISVs by weighing functional value against experiential enjoyment. VAM also explains how technological complexity lowers perceived value by increasing cognitive demands, offering insight into how AI literacy shapes evaluations.

ELM complements VAM by addressing persuasion processes through dual-path information processing [55]. In traditional online shopping research, ELM has long explained how high-involvement audiences attend to argument quality via the central route, whereas low-involvement audiences respond to peripheral cues such as endorser credibility, brand familiarity, or visual appeal [56]. Accumulated evidence underscores ELM’s robustness across persuasive domains, from online advertising settings such as eWOM’s effects on purchase intentions for private-label products [56] and cosmetics [57] to technology-mediated contexts including GPT adoption [58] and augmented-reality (AR) advertising [59].

The two models align naturally: content usefulness in VAM corresponds to information quality assessment in ELM’s central route, while the peripheral route highlights the importance of source characteristics (e.g., vlogger credibility) when cognitive resources are constrained. In AISV contexts, this integration explains why consumers with different AI literacy levels rely on distinct evaluative pathways. Building on this integration, this study proposes the AILPDM, which identifies key influence pathways across three dimensions: emotional, cognitive, and trust. AILPDM fills a critical gap by linking consumer AI literacy with multi-dimensional information processing, providing a systematic framework to explain the antecedent conditions of AISV marketing effectiveness.

2.4. Hypothesis Development

2.4.1. Emotional Pathway of AI Literacy: Shaping Emotional Value and Its Subsequent Effects

Emotional value, a key dimension of consumer perceived value, refers to the pleasure provided by a product or service [60] and the utility derived from the affective states it generates [61]. It is primarily reflected through two constructs: enjoyment and emotional attachment [62].

Consumer AI literacy directly shapes the ability to extract emotional value from AISVs, and high AI literacy enhances technological pleasure [63]. In AISV contexts, high-literacy consumers better understand technological features, perceive personalization more fully, and derive greater satisfaction from AI-driven recommendations and customized content [64]. They also respond more strongly to AI-induced novelty [31], and their broader hedonic motivations and technological experience further enrich emotional responses [65].

Although vlog advertising is often function-oriented, emotional value significantly influences consumer decision-making, even for durable goods [61]. Positive evaluations of AI performance increase user emotional responses, which in turn strengthen acceptance [65]. Accordingly, we propose the following:

H1: 

Consumer AI literacy directly influences AISVs’ emotional value.

Within the VAM, users’ perceived value of a technology positively influences their adoption intention [51]. In the context of the experience economy, users seek not only functional value but also the pleasurable experiences brought by emotional value [62]. Once emotional value is established, it affects behavioral outcomes. For example, entertainment satisfaction among digital natives enhances attitude formation [18], and in AISV contexts, pleasurable experiences extend viewing, deepen engagement, and foster positive information processing. Moreover, prior work shows that emotional engagement from online advertisements can positively influence attitudes [10], while satisfied consumers are more likely to share and rely on AISV content [28]. Further evidence from AI influencer research demonstrates that emotional appeals promote consumer willingness to follow even more effectively than rational appeals [66]. Extending this line, studies on AI advertising indicate that emotional trust, beyond cognitive judgments, is also a strong predictor of behavioral intentions in uncertain contexts [26]. Collectively, these studies suggest that emotional value plays a critical role in AISV adoption. Thus, we propose the following:

H2: 

AISVs’ emotional value directly influences consumer adoption of AISVs.

Emotional value is a key driver of consumer purchase intention. In online advertising, emotional engagement generated by advertisements has a positive effect on purchase intention [10]. Extending this perspective to AI-driven contexts, studies in conversational AI reveal that emotional trust and attachment enhance users’ willingness to continue usage [62], while users’ subjective emotional cognition of AI technology or products positively shape their behavioral intentions [21]. Collectively, these findings indicate that emotional bonds with AI can be directly converted into behavioral intentions. Positive emotional experiences can also transfer to sponsored products through associative learning, strengthening purchase willingness and word-of-mouth [60]. Beyond product evaluations, evidence also suggests that consumers with positive emotions show stronger AI acceptance [65]. Given that AI interactions often lag behind human interactions in generating emotional connection [67], successful emotional value creation in AISVs becomes critical. This argument aligns with established findings in marketing research, where the primary function of emotional appeals in campaigns is to attract audiences, evoke desired responses, and ultimately encourage the purchase of branded products or services [66]. Therefore, we propose the following:

H3: 

AISVs’ emotional value directly influences purchase intention.

2.4.2. Cognitive Pathway of AI Literacy: Assessment of Information Usefulness and Its Effects

Consumer AI literacy, representing the cognitive ability to think analytically and critically assess information beyond simple intuition, shapes how consumers evaluate AIGC [12]. Utilitarian value is the core dimension in consumer evaluation of AISVs. Lim et al. [18] define usefulness as the value of product information dissemination in online advertising. In AISV contexts, utility reflects the extent to which vlog product information assists consumer decision-making. Consumer AI literacy directly shapes perceptions of information usefulness. Lower AI literacy consumers tend to perceive AI as magical [19], while knowledgeable users form more reasonable expectations, and alignment between expectations and technological capabilities strengthens satisfaction and perceived usefulness [63]. High-literacy consumers better understand AISV content generation mechanisms, set realistic quality standards, and evaluate utilitarian value more accurately. Wu et al. [31] note that “machine heuristics” lead people to attribute objectivity and accuracy to AI, particularly in processing quantifiable tasks. High-literacy consumers are more adept at recognizing AI’s boundaries, identify misinformation [12], judging content usefulness rationally, and making trust decisions accordingly. When AISVs deliver accurate, relevant, and high-quality information, they create positive user experiences that encourage continued engagement [64]. High-literacy consumers are especially capable of filtering irrelevant content, thereby enhancing perceptions of usefulness. We thus propose the following:

H4: 

Consumer AI literacy directly influences perceived usefulness of AISVs.

In the context of e-commerce, information usefulness is widely recognized as a key determinant of information adoption [56]. Lim et al. [18] show that advertising usefulness positively affects attitudes and improves purchase decision efficiency. When consumers view AISVs as providing valuable information, they are more likely to adopt and apply it in decision-making. Argan et al. [17] further demonstrate that interactive behavior toward advertisements depends on value assessment; when information matches needs, perceived usefulness rises, encouraging action. Chen et al. [27] confirm that source credibility, content quality, and need matching collectively determine persuasive effectiveness.

Moreover, since AI is not perceived as a stakeholder, it is often regarded as an objective tool [16]. Emphasizing product functional attributes reduces resistance to AI advertising by leveraging the “machine effect,” wherein consumers perceive AI as more capable than humans in functional and objective assessments [68]. Perceiving AI advertising as useful increases consumers’ likelihood of integrating it into their decisions [41]. In addition, studies on conversational AI further show that perceived usefulness directly affects continued use, indicating that the usefulness consumers perceive from AI interactions translates into adoption [62]. These findings underscore the central role of information usefulness in AISV adoption. Therefore, we propose the following:

H5: 

AISVs’ usefulness directly influences consumer adoption of AISVs.

2.4.3. Trust Pathway of AI Literacy: Evaluating Source Credibility and Its Consequences

Source credibility is a central factor shaping consumer information processing and decision-making, classically defined as “positive characteristics of a communicator that influence the receiver’s acceptance of information,” typically measured through expertise and trustworthiness [69] (pp. 20–21). Credibility is regarded in psychology as both a personality trait and an interpersonal phenomenon, representing a relatively stable belief [70]. In online advertising, credibility reflects both the objective and subjective trustworthiness of information or entities [18]. With the rise of AISVs, vloggers simultaneously act as content creators and product sellers, requiring credibility assessment mechanisms to adapt to AI-driven contexts.

In AISV contexts, evaluating vlogger credibility requires consumers to possess specific AI-related cognitive abilities. AI literacy equips consumers with three core skills: technological discernment to distinguish between high- and low-quality AIGC, understanding AI’s boundaries to judge appropriate application of the technology, and professional judgment to evaluate vloggers’ proficiency in integrating AI with product information [23]. These abilities form the cognitive foundation for assessing vlogger credibility and directly influence trust formation in their dual roles.

Prior research has established that AI literacy has a positive effect on attitudes toward generative AI technologies [71]. In parallel, AR advertising studies show that information credibility significantly shapes consumers’ attitudes, which in turn influence purchase intentions [59]. More recent evidence indicates that consumers’ trust in AI must align with their actual competencies. Low-literacy users often fail to identify hallucinations in AIGC, thereby reducing perceived credibility in specific contexts [70]. In AISV contexts, these findings highlight that AI literacy enables consumers to critically evaluate AI-mediated content. Digital natives, who are adept at navigating AI-driven systems, exemplify the importance of AI literacy as an antecedent of credibility assessment in AISV contexts [29]. High-literacy consumers can thus make more accurate evaluations of vloggers’ credibility. Therefore, we propose the following:

H6: 

Consumer AI literacy directly influences perceived vlogger credibility.

Once credibility is established, it functions as a decisive factor in adoption, as evidenced by research showing that ChatGPT’s credibility is positively associated with its adoption [58]. Prior studies further demonstrate that perceptions of advertising credibility affect information processing and decision-making [18]. When consumers view information sources as credible, they are more likely to trust recommendations and form purchase intentions [72]. In AIGC contexts, maintaining the credibility of generated content secures user recognition and strengthens their trust [70], and trust has been proven to be a key factor in consumers’ decisions to purchase AI-driven products [26]. Credibility significantly influences viewing behavior and purchase decisions [41]. Source credibility affects adoption through two pathways: enhancing persuasiveness, which increases receptivity to claims, and reducing uncertainty, which lowers decision risk [73]. In AISVs, where AI application itself may raise trust concerns [16], credible vloggers can mitigate skepticism. When vloggers are considered both technically proficient in AI and trustworthy, consumers are more willing to adopt their recommendations [74]. Therefore, we propose the following:

H7: 

Vlogger credibility directly influences consumer adoption of AISVs.

Existing research shows that users’ technological literacy significantly affects their willingness to adopt AI-based applications, such as chatbots [75]. Evidence from organizational contexts further indicates that firms with advanced AI literacy demonstrate stronger absorptive capacity, allowing them to adopt complex knowledge, including green technologies, sustainability methods, and regulatory requirements [76]. Taken together, AI literacy plays a critical role in shaping how different actors process and adopt new information. Extending this logic to consumer contexts, in AISV marketing environments, consumer AI literacy functions as a fundamental determinant of how consumers process, evaluate, and adopt information [19]. AI-driven marketing affects consumer behavior, while AI literacy determines how consumers interpret and respond to content [36]. Higher digital literacy enables consumers to rationally evaluate AI marketing strategies, recognize their value and limitations, and make more informed decisions [29]. Beyond stronger evaluative capabilities, consumers with higher levels of AI literacy can employ more sophisticated interaction strategies to influence AI systems. Compared with low-literacy consumers, they tend to rely on advanced feedback signals (e.g., LIKE and FOLLOW) rather than merely depending on CLICK or DISLIKE to adjust AI outputs. Consumers who are confident in their ability to manipulate AI not only achieve superior outcomes but also exhibit lower resistance toward AI-generated content [77]. Such strategic interactions gradually guide AI to produce outputs that better align with rational consumer preferences. This alignment further enhancing consumers’ willingness to adopt the information. Therefore, we propose the following:

H8: 

Consumer AI literacy directly influences information adoption from AISVs.

Consumer AI literacy, developed from digital literacy frameworks, encompasses core competencies that enable consumers to effectively navigate AI-mediated commercial environments [23]. In AISV environments, it functions as a cognitive resource guiding how consumers process algorithmic information and form intentions. Empirical evidence already shows that digital natives’ favorable attitudes toward online advertising enhance purchase intentions [18], and that higher technological literacy increases perceived usefulness of AI tools [75]. Similar patterns are consistently observed in digital commerce: higher digital literacy strengthens online purchase intention [78], predicts female consumers’ shopping behavior [79], and directly impacts consumers’ willingness to buy [80]. Extending these findings, AI literacy is expected to provide greater confidence and control over AI-mediated shopping, thereby fostering stronger purchase intentions. Thus, we propose the following:

H9: 

Consumer AI literacy directly influences purchase intention.

Prior research establishes that consumers’ adoption of social media eWOM positively shapes purchase intentions [57]. Information adoption functions as a central cognitive mechanism in online shopping, directly linking content engagement to purchasing outcomes [56]. Extending this logic to AISVs, adoption reflects the transition from perceiving content as valuable and credible to forming the cognitive basis for purchase decisions [18]. Empirical evidence reinforces this pathway: once integrated, AI-enhanced marketing information transforms into a cognitive foundation for consumer behavior [29]; willingness to follow AI influencers has been shown to encourage purchase intentions [66]; and consumer acceptance of AI-based services similarly fosters purchase intentions [81]. Collectively, these findings indicate that information adoption within AISV contexts signals the overcoming of cognitive barriers and the establishment of trust, thereby enabling purchase behavior. Therefore, we propose the following:

H10: 

Information adoption from AISVs directly influences purchase intention.

Following the development of the research hypotheses,

Table 1. Operational definitions.

Constructs	Operational Definitions	References
Consumer AI literacy	The degree to which consumers possess the knowledge, skills, and critical understanding necessary to recognize, evaluate, and engage with AIGC, particularly AISVs.	[20,23]
AISV emotional value	The affective utility consumers experience during AISV viewing, characterized by positive emotional responses.	[60,61]
AISV information usefulness	The extent to which product or service-related information in AISVs enhances consumers’ purchase decisions.	[18,56]
Source credibility	The perceived ability and motivation of the vlogger in an AISV to produce accurate and truthful information.	[70,82]
AISV information adoption	The extent to which consumers accept and utilize the information from AISVs to support their purchase decisions.	[83]
Purchase intention	Consumers’ willingness to purchase products or services after watching AISVs.	[84]

Note. Compiled by the author.

summarizes the operational definitions of the variables employed in this study.

3. Methodology

This study employed a quantitative research approach using a cross-sectional survey design targeting Chinese social media users who watch AISVs and make purchases via these platforms. The research utilized a structured questionnaire with adapted scales measuring AI literacy, purchase intention, and related constructs through five-point Likert scales (1 = strongly disagree, 5 = strongly agree). Convenience sampling was implemented via WeChat distribution, focusing on users with actual AISV viewing and purchasing experience. While this sampling approach has limitations in terms of statistical generalizability, the resulting sample characteristics closely matched the demographic profile of typical Chinese social media users engaged with AI-generated content, providing contextual representativeness for the research objectives. A minimum sample size of 385 was calculated based on a 95% confidence level and a 5% margin of error (Z = 1.96). The project was conducted under the auspices of Universiti Kebangsaan Malaysia and received ethical approval from its Research Ethics Committee (Ref. No. JEP-2024-382). Informed consent was obtained electronically from all respondents through acknowledgment of a consent statement presented on the first page of the questionnaire. Additionally, to help respondents clearly distinguish between tangible product AISVs and experience service AISVs, illustrative screenshots from representative TikTok vlogs were embedded in both the survey introduction and the relevant category-question sections.

The study’s effective sample comprised 37.53% male participants and 62.47% female participants, broadly consistent with TikTok e-commerce users (66% female) [15]. Most respondents were aged 18–39 (87.89%), with 18–24-year-olds accounting for 44.79%, suggesting that the survey primarily captured the perspectives of younger individuals. Education was predominantly undergraduate/bachelor’s level (73.61%), with 81.36% having college education or above. Occupations were mainly students (36.80%) and corporate employees (26.15%), and 75.79% of respondents watched vlogs at least once daily. These characteristics align with the core target audience for AISVs—young, highly educated, digitally engaged users—supporting the sample’s representativeness and the validity of the findings [21].

Research Instruments, Measures, and Variable Measurement

AISVs are a novel content format in which creators use AI tools to produce product-focused videos, with consumers primarily acting as passive viewers rather than active technology evaluators. Building on Carolus et al.’s [85] AI literacy framework, we adapted the construct for this context by retaining three dimensions—Use and Apply AI, Know and Understand AI, and Detect AI—while excluding AI Ethics. This decision followed a contextual relevance analysis, as ethics items mainly address macro-level societal issues and active technology evaluation, which are less applicable to passive AISV consumption. Consistent with prior studies [21,86,87], the adapted scale was administered as a self-assessment instrument in which participants evaluated their own knowledge, skills, and recognition of AI applications. This operationalization aligns with scale adaptation principles that recommend contextual refinement while preserving theoretical integrity, enhancing construct validity [88] (pp. 285–290).

Given the emerging nature of AISVs and the lack of mature measurement tools, scale items were selected and adapted from the literature (

Table 2. Measurement and questionnaire items and level of measurement.

Constructs	Measurement Items	Questionnaire Items	Variables’ Level of Measurement	References
AI Literacy (AIL)	AIL1	I can distinguish if I interact with an AI or a “real human”.	Interval	[85]
	AIL2	I can operate AI-generated video applications in everyday life.	Interval
	AIL3	I can tell if I am dealing with a vlog based on artificial intelligence.	Interval
	AIL4	I can assess what advantages and disadvantages the use of an artificial intelligence entails.	Interval
	AIL5	I know the most important concepts of the topic “artificial intelligence”.	Interval
	AIL6	I can think of new uses for AI.	Interval
		Based on your feelings after watching AISVs, please evaluate the following statements:		[65]
Emotion value (EM)	EM1	Very boring 1–2–3–4–5 Very relaxed	Interval
	EM2	Very depressed 1–2–3–4–5 Very satisfied	Interval
	EM3	Very hopeless 1–2–3–4–5 Full of hope	Interval
	EM4	Very annoyed 1–2–3–4–5 Very pleased	Interval
		If you choose to watch videos labeled “AI-generated” or “Potentially AI-generated,” please evaluate the following statements based on the AISVs you’ve watched		[89,90]
Information Usefulness (IU)	IU1	The product or service-related information in the AISV is valuable.	Interval
	IU2	The product or service-related information in the AISV is informative.	Interval
	IU3	The product or service-related information in the AISV is helpful.	Interval
	IU4	The product or service-related information in the AISV is useful.	Interval
Source credibility (SC)	SC1	I believe this vlogger is knowledgeable.	Interval	[82]
	SC2	I believe this vlogger is an expert.	Interval
	SC3	I believe this vlogger or is reliable.	Interval
	SC4	I believe this vlogger or is trustworthy.	Interval
Information Adoption (IA)	IA1	The information in this AISV enhanced my knowledge of the product or service.	Interval	[84,90]
	IA2	This AISV had a significant impact on me.	Interval
	IA3	I agree with the viewpoints presented in this AISV.	Interval
Purchase Intention (PI)		When the AISVs you watch and provides corresponding purchase links, please evaluate the following statements.		[84,91]
	PI1	It is very likely that I will buy the product or service.	Interval
	PI2	I will definitely try the product or service.	Interval
	PI3	If I am in need, I would buy the product or service.	Interval
	PI4	I will buy the product or service next time I need it.	Interval

Note. Compiled by the author.

) to match the research variables. Participants rated items on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). Content validity was confirmed through expert review, and a pilot test (n = 128) demonstrated reliability (Cronbach’s α [CA] > 0.8647) and validity (average variance extracted [AVE] > 0.603; composite reliability [CR] and CA > 0.7), meeting widely accepted academic standards and providing a robust foundation for formal testing.

The formal survey conducted in May 2025 yielded 413 valid responses from 666 questionnaires (62.0% response rate) after excluding incomplete responses, underage participants, and those unexposed to either type of AISV. Social-media users are typically immersed in diverse algorithmically recommended content, making it implausible to assume exposure to a single AISV category. To capture this behavioral reality, the questionnaire included a multiple-choice item asking, “Which types of AISVs have you watched?” (Option 1: tangible products; Option 2: experiential services). Among the valid responses, 369 participants reported viewing tangible product AISVs and 333 reported exposure to experiential service AISVs. As expected, substantial overlap emerged between the two subsets, reflecting users’ hybrid exposure patterns and leaving only a minor proportion of tangible-only or experiential-only viewers. Consequently, group membership in subsequent analyses was not mutually exclusive but represented two interrelated exposure contexts.

To examine how AI literacy operates across these contexts, separate structural equation models (SEMs) were estimated for tangible product and experiential service AISVs using overlapping subsets of participants. The same measurement and structural configurations were applied to both models to ensure conceptual equivalence. Because the subsets were not mutually exclusive, the results are interpreted as type-specific pathways rather than formal group-level comparisons. Parameter estimates (standardized path coefficients and significance levels) were compared descriptively to identify pattern differences in direct and indirect effects while accounting for measurement error within each model. This analytic strategy isolates the mechanisms through which AI literacy influences information adoption and purchase intention across distinct, yet behaviorally intersecting, AISV contexts.

4. Results

Descriptive statistics for all constructs based on the valid sample (n = 413) indicated mean values ranging from 3.340 to 3.701 (SD = 0.957–1.187), suggesting moderate mean levels with acceptable dispersion. Across the sample, 369 participants (89.3%) reported exposure to tangible product AISVs and 333 (80.6%) to experiential service AISVs, including 289 (70.0%) who had encountered both. The remaining subsets—tangible-only (n = 80, 19.4%) and experiential-only (n = 44, 10.7%)—were comparatively smaller, consistent with behavioral overlap commonly observed in algorithmically curated content. Demographic indicators exhibited minimal variation between the tangible (T) and experiential (E) exposure subsets, further supporting the interpretive comparability of the two type-specific SEMs reported below. All variables exhibited acceptable normality, with skewness between −0.593 and −0. 203 and kurtosis between 2.147 and 2.912, within recommended thresholds (|3.0| and |10.0|, respectively) [92]. The measurement model demonstrated strong reliability (

Table 3. Reliability and validity statistics.

Constructs	Cronbach’s α	KMO	CR	Factor Loading
AIL	0.921	0.913	0.918	0.763–0.841
EM	0.928	0.897	0.932	0.865–0.895
IU	0.908	0.851	0.916	0.841–0.868
SC	0.925	0.839	0.921	0.839–0.887
IA	0.889	0.843	0.872	0.815–0.851
PI	0.921	0.844	0.917	0.833–0.879

Note. KMO = Kaiser–Meyer–Olkin; CR = composite reliability; AVE = average variance extracted; AIL = artificial intelligence literacy; EM = emotion value; IU = information usefulness; SC = source credibility; IA = information adoption; PI = purchase intention.

), with CA and CR values of 0.872–0.925 (>0.7). Kaiser-Meyer-Olkin measures (0.839–0.913) indicated excellent sampling adequacy. Convergent validity was supported by high factor loadings (0.763–0.895, all > 0.7) and AVE values (0.632–0.773, all > 0.5). Discriminant validity met the Fornell–Larcker criteria (√AVE > inter-construct correlations) (Appendix A) and was further supported by HTMT values below 0.85, confirming the measurement model’s suitability for hypothesis testing across both type-specific contexts [92]. Collinearity diagnostics (VIF = 1.75–2.12; Tolerance = 0.472–0.571) indicated no multicollinearity, confirming adequate discriminant validity among the constructs.

Before hypothesis testing, the fit of the measurement models was evaluated separately for each AISV type-specific model. As

Table 4. Model fit.

Fit Index	Tangible Product	Experiential Service
χ2_ms (df = 312)	881.437	924.563
p > χ2	0.000	0.000
χ2_bs (df = 351)	8103.275	7842.190
p > χ2	0.000	0.000
RMSEA	0.069	0.075
90% CI	[0.062, 0.076]	[0.068, 0.082]
p close	0.019	0.003
AIC	1011.437	1054.563
BIC	1258.966	1308.828
CFI	0.932	0.923
TLI	0.917	0.908
SRMR	0.051	0.058
CD	0.944	0.928

Note. ms = model vs. saturated; bs = baseline vs. saturated; RMSEA = root-mean-square error of approximation; CI = confidence interval; AIC = Akaike information criterion; BIC = Bayesian information criterion; CFI = comparative fit index; TLI = Tucker–Lewis index; SRMR = standardized root-mean-square residual; CD = coefficient of determination.

shows, both models demonstrated acceptable fit indices. Although chi-square tests were significant (χ²_ms (312) = 881.437 and 924.563, p < 0.001), this is common in large samples. Root mean square error of approximation values are 0.069 and 0.075, close to recommended thresholds; comparative fit index (0.932 and 0.923) and Tucker–Lewis Index (0.917–0.908) approached or exceeded 0.90; standardized root mean square residual values were below 0.08 (0.051 and 0.058); and CD values remained high (0.944 and 0.928). Collectively, these indicators suggest that both type-specific measurement models achieved satisfactory fit and explanatory adequacy [93]. All model fit indices were cross-checked with WLSMV estimation, showing consistent and marginally improved outcomes. Furthermore, Harman’s single-factor test indicated no substantial common method bias, as the theoretical six-factor model fit the data significantly better than a single-factor model (Δχ² = 1966.2 and 2071.2; p < 0.001), confirming the discriminant validity of the constructs.

Table 5. Hypothesis test results.

	Structural Path	Type	β	p-Value	z-Score	Standard Error	Bootstrap 95% CI	Result
H1	AIL → EM	T	0.521	0.000	6.89	0.076	(0.374, 0.668)	Supported
		E	0.548	0.000	7.12	0.077	(0.396, 0.700)	Supported
H2	EM → IA	T	0.235	0.008	2.67	0.088	(0.062, 0.408)	Supported
		E	0.220	0.015	2.43	0.091	(0.042, 0.398)	Supported
H3	EM → PI	T	0.087	0.298	1.04	0.084	(−0.078, 0.252)	Not Supported
		E	0.064	0.461	0.74	0.086	(−0.102, 0.230)	Not Supported
H4	AIL → IU	T	0.592	0.000	7.58	0.078	(0.439, 0.745)	Supported
		E	0.567	0.000	7.21	0.079	(0.412, 0.722)	Supported
H5	IU → IA	T	0.320	0.000	4.38	0.073	(0.177, 0.463)	Supported
		E	0.340	0.000	4.59	0.074	(0.195, 0.485)	Supported
H6	AIL → SC	T	0.508	0.000	6.45	0.079	(0.354, 0.662)	Supported
		E	0.534	0.000	6.78	0.079	(0.379, 0.689)	Supported
H7	SC → IA	T	0.185	0.045	2.01	0.092	(0.005, 0.365)	Supported
		E	0.165	0.076	1.77	0.093	(−0.018, 0.348)	Not Supported
H8	AIL → IA	T	0.156	0.067	1.83	0.085	(−0.011, 0.323)	Not Supported
		E	0.142	0.089	1.70	0.084	(−0.022, 0.306)	Not Supported
H9	AIL → PI	T	0.098	0.185	1.33	0.074	(−0.046, 0.242)	Not Supported
		E	0.165	0.042	2.03	0.081	(0.006, 0.324)	Supported
H10	IA → PI	T	0.618	0.000	8.12	0.076	(0.468, 0.768)	Supported
		E	0.564	0.000	7.45	0.076	(0.415, 0.713)	Supported

Note: β = standardized path coefficient; AIL = artificial intelligence literacy; EM = emotion value; IA = information adoption; PI = purchase intention; IU = information usefulness; SC = source credibility; T = Tangible product; E = Experiential service.

presents the standardized coefficients and significance levels for direct paths estimated in the two type-specific SEMs. AI literacy significantly and positively influences emotional value in both type-specific models (βT = 0.521, βE = 0.548; p < 0.001), supporting H1. Emotional value, in turn, positively affects information adoption (βT = 0.235, βE = 0.220; p < 0.05), confirming H2. AI literacy also significantly increases information usefulness (βT = 0.592, βE = 0.567; p < 0.001) and source credibility (βT = 0.508, βE = 0.534; p < 0.001), supporting H4 and H6. Information usefulness significantly and positively influences information adoption in both AISV contexts (βT = 0.320, βE = 0.340; p < 0.001), supporting H5. Source credibility only significantly and positively affects information adoption within the tangible product model (βT = 0.185, p < 0.05), while the effect is non-significant for experience service, partially supporting H7. The direct effect of AI literacy on information adoption is non-significant (βT = 0.156, βE = 0.142; p > 0.05), failing to support H8. Information adoption strongly predicts purchase intention (βT = 0.618, βE = 0.564; p < 0.001), confirming H10, while emotional value’s direct effect on purchase intention is non-significant (H3 is not supported). AI literacy’s direct effect on purchase intention is significant only within the experiential service context (β = 0.165, p = 0.042), partially supporting H9. These results are illustrated by the final structural models with standardized coefficients and significance levels, as shown in Figure 1 and Figure 2.

As a supplementary analysis, mediation effects were tested using the Delta method, Sobel test, and Monte Carlo simulation (Appendix B). The indirect paths from AI literacy to information adoption through emotional value, information usefulness, and source credibility are significant, suggesting that AI literacy influences information adoption mainly through these value perceptions. Among the three mediators, information usefulness shows the strongest indirect effect, followed by emotional value and source credibility, and this pattern is consistent across both AISV contexts. The indirect effect of AI literacy on purchase intention via information adoption is non-significant (p > 0.05), whereas the indirect effect of emotional value on purchase intention via the information adoption pathway is significant (p < 0.001), highlighting information adoption as a key mediator in both models.

5. Discussion

5.1. How AI Literacy Shapes Consumer Perception and Information Adoption from AISVs (RQ1)

This study’s findings suggest that consumer AI literacy does not directly affect AISV information adoption (H8 not supported) but operates through three parallel mediating paths: information usefulness, emotional value, and source credibility. This challenges the assumption that “understanding AI technology necessarily leads to AI technology acceptance” [94] and highlights consumers’ prioritization when evaluating AISVs: utilitarian value outweighs emotional and trust factors. These results establish value perceptions as key mediators between technological cognition and behavioral adoption.

The emotional value pathway confirms that AI literacy enhances AISV information adoption by increasing emotional value perception (H1, H2 supported). This effect follows a two-stage cognitive emotional transformation: Consumers with higher levels of AI literacy are better able to reduce cognitive load and AI anxiety [94], then convert this cognitive advantage into an emotional appreciation of AISV content. This finding addresses Coffin’s [25] question of whether consumers will exchange choice autonomy for AI convenience, showing that AI evaluation depends not only on technology itself but also on consumers’ AI literacy. By validating AI literacy as a key antecedent of emotional value perception, this study clarifies the formation of users’ emotional responses to AISVs, complementing prior research that reported divergent reactions to AIGC without explaining their origins [28,47].

However, the hypothesis that emotional value directly impacts purchase intention (H3) is not supported, revealing a theoretical nuance. While prior studies suggest that emotional engagement from online advertisements can positively influence attitudes and purchase intention [10], our findings show that emotional value primarily promotes information adoption. Only through this mediating pathway does AIGC influence final purchase decisions, highlighting its role in driving initial engagement and information processing rather than directly affecting purchase behavior.

This study identifies consumer AI literacy enhances AISV information adoption primarily by increasing perceived usefulness (H4, H5). This aligns with prior research emphasizing perceived usefulness as a key determinant in technology acceptance [94]. AI advertising technologies are highly efficient—capable of generating 20,000–30,000 copies daily, 50–60 times the output of humans—while leveraging data-driven approaches, tool support, and parallel processing [47]. Our results show a significant positive correlation between AI literacy and perceived AISV usefulness, which can be explained through the “machine heuristic”: Consumers perceive AI as more accurate in objective, quantifiable tasks [31]. In utilitarian consumption contexts, Higher AI literacy tends to facilitate the application of this heuristic, attributing greater functional value to AISV content [68]. To the best of our knowledge, this study is the first to empirically link AI literacy, the machine heuristic, and perceived usefulness, explaining how transparency disclosure enhances marketing effectiveness: AI literacy activates the machine heuristic, which increases perceived usefulness and, in turn, promotes information adoption, providing a theoretical basis for AI marketing design.

The results support H6, indicating that higher AI literacy significantly enhances consumers’ perceptions of the AISV vloggers’ credibility. Previous work has suggested that AI transparency disclosure enhances perceived objectivity in advertising [95] and signals brand openness and value alignment [27,74]. Our findings extend this framework to content creator evaluation. The study reveals that technological competence in the AI-generated marketing context is reshaping the mechanism of consumer trust formation. When consumers possess a deep understanding of AI’s capabilities and limitations, they can more accurately recognize its functional boundaries and strengths [16], thereby interpreting vloggers’ transparent and skillful use of AI technologies as signals of honesty and professional expertise. Contrary to the traditional assumption that AI involvement undermines authenticity and trust [25], the findings demonstrate that AI literacy does not erode but rather reinforces the foundation of trust by strengthening consumers’ comprehension of technological transparency and operational proficiency.

H7 was supported, revealing that the effect of source credibility on information adoption presents context-specific variation: this effect is significant only in tangible product AISVs but not in experiential service AISVs. This finding suggests that the effectiveness of the trust mechanism depends on the verifiability of the evaluation target and consumers’ cognitive orientation. Prior research has pointed out that products possess functional attributes that can be owned and compared, whereas services center on subjective experiences and emotional interactions [33]. Within this framework, the trust pathways activated by AISVs appear to operate differently across contexts. For tangible products, AI technologies can intuitively demonstrate product characteristics and performance, enabling consumers to judge the authenticity of information based on visual cues and to perceive vloggers’ professional presentation as a reliable signal. This observation aligns with existing research emphasizing the central role of source credibility in shaping consumer trust and information adoption [18,60]. In contrast, in experiential service contexts, evaluation standards rely on subjective experiences and affective response. Even when vloggers are perceived as highly credible, consumers are unable to determine whether their own experiences will correspond to what is shown, making trust less likely to translate into adoption behavior. The nonsignificant results of H7 for the service type indicate that when purchase decisions are based on subjective feelings rather than functional verification, the effect of source credibility within the trust pathway becomes substantially weakened. These findings reveal the contextual boundaries of AI-driven trust mechanisms. When decisions rely on rational cognition and objective information—as in product contexts—source credibility serves as a critical persuasive cue that facilitates AISV adoption. However, when decisions are guided by emotional experience—as in service contexts—the influence of this pathway diminishes, and trust factors become secondary in shaping adoption intentions.

In traditional ELM theory, source credibility functions as a peripheral cue used by consumers in low-cognitive-investment states [55]. In the AISV context, however, highly AI-literate consumers evaluate source credibility more deeply. Their understanding of AI allows them to interpret vloggers’ transparent and skillful use of AI as evidence of honesty and professional competence. This cognitive processing elevates source credibility from a peripheral cue to a factor requiring thoughtful analysis, challenging and extending traditional persuasion frameworks in the AI content era. This study shows that highly AI-literate consumers assess creators’ honesty and professional capabilities by evaluating technological transparency in AISVs, establishing a new paradigm of credibility assessment based on technological literacy.

Overall, this study identifies three parallel pathways through which AI literacy influences AISV information adoption: emotional value, information usefulness, and source credibility. AI literacy serves as a cognitive advantage that enhances emotional appreciation for AISV content, promoting adoption. The “machine heuristic” pathway—where high AI literacy increases perceived information usefulness—exerts the strongest effect, while the source credibility pathway demonstrates that highly AI-literate consumers develop trust in algorithmic integrity through perceptions of technological transparency. These findings clarify how AISV marketing content gains consumer acceptance. However, the conversion from information adoption to actual purchase decisions remains to be examined.

5.2. Direct and Indirect Effects of AI Literacy on Purchase Intention (RQ2)

This study identifies multiple mediating mechanisms through which AI literacy influences purchase intention, based on SEM analysis. The results confirm that AISV information adoption significantly influences purchase intention (H10 supported), reinforcing its role as a key antecedent of consumer behavioral intentions [5]. AI literacy enhances information adoption by shaping consumers’ three-dimensional value perceptions of AISV—emotional value, information usefulness, and source credibility—which in turn promote purchase intention; this supports the view that digital literacy strengthens consumers’ critical understanding of marketing strategies [29].

The analysis shows that emotional value affects purchase intention only indirectly through information adoption, revealing an “emotional purchase conversion mechanism”: While emotional value stimulates initial interest, visual expressiveness enhances emotional experience but does not necessarily strengthen consumers’ cognitive assessment of message reliability [59], and thus cannot directly drive purchase decisions. Consumers must assess content utility and relevance via information adoption to form purchase intentions. Higher AI literacy reduces cognitive load when processing AISV content [96], allowing consumers to perceive value more clearly and evaluate information more efficiently, facilitating the conversion from emotional engagement to purchase decisions.

Regarding AI literacy’s direct effect on purchase intention (H9), a descriptive difference is observed across AISV types: the direct path from AI literacy to purchase intention was evident in the experiential service model but not detected in the tangible product model. This indicates that the mechanisms linking AI literacy to purchase behavior vary contextually across AISV categories, forming the basis for further investigation.

5.3. Relationship Patterns Under AISV Types (Products and Services) Differences (RQ3)

Separate SEM revealed that AI literacy consistently influences the three mediating variables—emotional value, information usefulness, and source credibility—across both AISV contexts. AI literacy acts as a metacognitive ability that enhances consumers’ cognitive assessment of AIGC, independent of product category. However, a descriptive variation emerged in the direct path from AI literacy to purchase intention: the effect appeared significant in the experiential service model but was not observed in the tangible product model. This indicates that AI literacy’s effect on purchase decisions varies by AISV types, warranting theoretical exploration of the underlying cognitive mechanisms.

Grounded cognition theory and dual mechanisms of technological transparency help explain this phenomenon. Grounded cognition theory suggests that cognitive processes are rooted in interactions between the body and environment [97], with sensorimotor systems and mirror neurons engaged during concept understanding and action observation [98]. Experiential AISVs (e.g., fitness, dance, sports guidance) rely on bodily perceptual engagement. Highly AI-literate consumers, understanding AI generation mechanisms, can overcome technological barriers, accurately interpret visual-action simulations, map AISV content to potential bodily experiences, and form direct purchase intentions.

Moreover, AI marketing faces challenges owing to algorithmic opacity, with low explainability reducing adoption willingness [99] and transparency deficits contributing to consumer resistance [74]. For highly AI-literate consumers, this opacity transforms into a cognitive advantage. Explainable AI systems enhance comprehensibility, allowing users to understand model operations and content generation [99]. Combining this transparency with existing technical knowledge, highly AI-literate consumers shift focus from questioning AI content authenticity to evaluating how effectively AISVs convey experiential attributes. This cognitive advantage, coupled with neural simulation capabilities from grounded cognition, explains why AI literacy appears to bypass traditional information adoption pathways and is observed to directly influence purchase intention in experiential service contexts.

Relative to experiential service contexts, actorless tangible product-oriented AISVs primarily emphasize functional specifications and physical attributes, engaging bodily perception simulation to a lesser extent. In this context, even high AI literacy appears to provides limited advantage; consumers struggle to map AISV content to bodily experiences and thus are less likely to form direct purchase intentions. The finding also shows that AI literacy has a non-significant effect on information adoption in the tangible product model, indicating its limited role in this domain. These patterns define the boundary conditions of AI literacy’s influence on purchase decisions: in experiential service contexts, AI literacy appears to bypass traditional information-adoption pathways and is observed to directly affect purchase intention.

This result challenges two existing perspectives. First, contrary to Gursoy et al. [65], who concluded that AI anthropomorphic features do not enhance task performance or service quality, we find that consumers with higher levels of AI literacy tend to shift their evaluation from anthropomorphism to functionality, effectively overcoming the “uncanny valley” [100]. Similarly, research on virtual AI-driven streamer has shown that perceived anthropomorphism can positively influence consumers’ flow experience, which in turn fosters impulsive purchase intention [101]. This cognitive shift reflects the evolution of technology acceptance models, particularly within experiential service contexts. Second, we question Wu et al.’s [47] claim that AI content creation is most effective for rational appeals and utilitarian products. When AI literacy is considered, greater consumer literacy appears to be associated with higher acceptance and more direct engagement with experiential service AISVs, suggesting that AI content applicability extends beyond purely functional domains.

Overall, consumer cognition of AI capabilities appears to be evolving from mechanical and rational toward experiential and emotional dimensions, with AI literacy seemingly facilitating this transition. By clarifying the boundary conditions of AI literacy’s effect on purchase decisions, this study offers a new perspective on the context-specific impact of AI content marketing and provides a theoretical foundation for optimizing AI marketing strategies across AISV types (tangible products and experiential services).

Finally, this study incorporated age, gender, and vlog viewing frequency as control variables to account for demographic and media-usage heterogeneity (Appendix C). The results reveal a coherent and theoretically meaningful pattern. Age exerts significant negative effects on both AI literacy (β = −0.185, p = 0.012) and information adoption (β = −0.142, p = 0.028), indicating that older users face greater cognitive and adaptive barriers when engaging with AI-related information—a finding consistent with the literature on the AI divide [21]. Gender shows no significant impact on adoption intention but demonstrates a positive effect on purchase intention (β = 0.156, p = 0.018), suggesting that the predominantly female user base of short-video commerce platforms (e.g., TikTok Mall) may contribute to higher observed purchase activity [15]. Vlog viewing frequency positively influences both AI literacy (β = 0.234, p = 0.003) and adoption intention (β = 0.167, p = 0.021), implying that more frequent platform use strengthens algorithmic learning and personalization mechanisms. As a result, users are more likely to be exposed to AI-recommended [77] and AISV, which, through repeated interaction, fosters greater familiarity and enhances adoption tendencies. Collectively, the control-variable results highlight that demographic and behavioral heterogeneity constitutes an underlying layer shaping user responsiveness in AISV ecosystems. Age-related disparities appear to constrain AI literacy and attenuate engagement depth, whereas gender-linked variation reveals the socially contextualized nature of purchase intention within algorithmic commerce environments. Moreover, viewing frequency operates as an enabler of algorithmic familiarity; through iterative exposure, users internalize AI-recommendation logic and develop heightened receptivity to AISV. Recognizing these patterned influences extends the explanatory scope of the present model beyond its structural relationships while delineating potential moderating boundaries for subsequent inquiry.

In summary, this study elucidates the multi-level mechanisms by which AI literacy shapes AISV decision-making. The finding related to RQ1 shows that AI literacy influences information adoption through three parallel pathways: emotional value, information usefulness, and source credibility, with the information usefulness pathway, based on “machine heuristics,” exhibiting the strongest effect and emphasizing consumers’ prioritization of utilitarian content. The finding regarding RQ2 suggests information adoption as a key antecedent to purchase intention and demonstrates that the “emotion–purchase conversion mechanism” requires mediation through information adoption. RQ3 indicates that AISV type serves as a contextual boundary: the direct path from AI literacy to purchase intention was observed in experiential service AISVs but not detected in tangible product AISVs, where the effect operated primarily via information adoption. These contextual insights reveal the situational dependency of technological literacy’s impact on consumer decision-making, enriching digital marketing theory by clarifying cognitive affective behavioral conversion mechanisms. Overall, this study defines both the scope and boundaries of AI literacy’s influence, offering a systematic theoretical framework for understanding consumer behavior in emerging AI-driven marketing environments.

6. Theoretical Implications

This study extends the concept of digital immigrants by introducing the consumer classifications of AI immigrants and AI natives, based on differences in AI literacy. It develops the AILPDM, which traces how AI literacy influences information adoption via AISV perceived value and ultimately shapes purchase intention. This study makes three key theoretical contributions. First, by constructing the AILPDM, it addresses limitations of the traditional VAM by incorporating consumer AI literacy as a core antecedent, establishing the complete causal chain: AI literacy influences value perception, which facilitates information adoption and ultimately leads to purchase intention. Unlike traditional VAM, which treats technology solely as a non-monetary cost that reduces perceived value [51], AILPDM highlights the dual nature of technological understanding: initial learning represents a cost, whereas mastered technological capability becomes a cognitive resource that enhances AISV value perception. This distinction between “acquisition process” and “existing capability” strengthens the model’s explanatory power for AI content evaluation. Empirical results confirm that AI literacy significantly affects emotional value and information usefulness, validating AILPDM’s relevance for understanding consumer decision-making in the AI era.

Second, this study extends the ELM to AI content evaluation. While traditional ELM treats source credibility as a peripheral cue [55], our results show that highly AI-literate consumers process source evaluation centrally, interpreting technological transparency as evidence of honesty and professional competence. By integrating AI literacy, AILPDM clarifies how consumers form new information-processing mechanisms, offering a more precise theoretical framework and enhancing ELM’s explanatory power in AI-driven contexts.

Third, this study identifies systematic differences in AI literacy’s influence across AISV types. For actorless tangible product AISVs, AI literacy affects purchase intention indirectly through information adoption, whereas for avatar-mediated experiential service AISVs, it can directly drive purchase decisions. This finding establishes the contextual boundaries of AILPDM and provides empirical insight into AIGC’s differential effects.

Fourth, this study contributes methodologically by strengthening the measurement of AI literacy. Although this construct has largely been operationalized through self-assessment scales, empirical validation remains scarce. This study represents an early validation of AI-literacy scales in consumer research, thereby extending their applicability and enhancing methodological rigor in AI-driven contexts.

7. Practical Implications

7.1. Implications for Brand Managers

Grounded in the AILPDM framework and the empirical results, the findings underscore that brand managers should engage sponsored vloggers with advanced AISV production competence, as technological proficiency enables comprehensive demonstration of product functions and advantages through AI generated presentations, thereby amplifying the utilitarian value most salient to consumers. Beyond functional communication, emotional and trust dimensions remain essential. Collaborations with vloggers who not only exhibit technical mastery but also evoke resonance and authenticity through narrative expression and self-presentation can strengthen persuasion and enhance the overall effectiveness of AISV marketing. Crucially, in the context of tangible product AISVs, source credibility emerges as a decisive persuasion mechanism. When evaluating potential collaborations, brand managers should emphasize vloggers’ perceived trustworthiness and technical competence, as these attributes translate algorithmic transparency into consumer confidence and, consequently, higher adoption intentions. This alignment between technological capability and credibility formation refines strategic vlogger selection in AI mediated marketing environments.

7.2. Implications for Content Creators

AISV content strategies should follow the consumer value assessment sequence: utilitarian value, emotional value, and trust. Accordingly, creators should construct a three-tier content architecture by establishing usefulness through functional information, incorporating emotional elements to evoke resonance, and building trust via vlogger roles. In avatar-mediated experiential services such as AI virtual fitness coaching, incorporating precision motion markers, muscle-activation prompts, and slow-motion demonstrations enables high-AI-literacy consumers to perform mental simulation, transforming perceived utility into direct purchase intention. Conversely, for tangible product categories such as apparel or home furnishings, dynamic visualization and decision transparency—exemplified by disclosing recommendation logic or providing real-time rendering adjustments—function as diagnostic cues that emphasize product efficacy, alleviate resistance to “black-box” algorithms [99], and strengthen adoption and conversion processes. Beyond functional display, sustainable trust formation depends on aligning technological sophistication with transparent disclosure and authentic presentation. The integration of technical expertise and ethical responsibility constitutes the foundation for long-term credibility in creator–consumer relationships. Furthermore, emerging evidence indicates that AI generation in vlog tags or metadata satisfies informational disclosure norms, whereas omitting explicit references to “artificial intelligence” in product descriptions maintains perceptual fluency and mitigates cognitive resistance. Instead, framing messages through terms such as “cutting-edge technology” or “advanced technology” communicates AI-enabled benefits while mitigating potential reactance, thereby enhancing consumer receptivity and market performance [26].

7.3. Implications for Retailers

Retailers should integrate AI literacy into customer profiling as a cognitive dimension that shapes consumers’ interaction with AI-mediated marketing. Incorporating this factor into segmentation enables the design of differentiated transparency strategies that align technological complexity with user capability. As younger consumers generally exhibit higher AI literacy, they can be targeted with AISV campaigns that demonstrate greater technological transparency and provide detailed explanations of AI generation and algorithmic logic. In contrast, low-literacy users should receive simplified, heuristic cues that maintain clarity without inducing cognitive overload. This alignment between cognitive sophistication and message design enhances perceived control and trust, mitigates AI-related illusions [70], and strengthens both information adoption and conversion efficiency, thereby generating competitive advantage in AI-driven retail contexts. Finally, considering that female consumers display higher purchase intentions, AISV product selection can strategically incorporate preferences salient to this demographic, linking algorithmic personalization with user-centric value creation.

7.4. Implications for Consumers

As AISV evolves into a dominant marketing practice driven by its efficiency and scalability, consumers’ AI literacy becomes a pivotal capability shaping how they interpret and respond to AI-generated content. Conceptualized in this study as a technical dimension, AI literacy enhances consumers’ understanding of generative mechanisms and strengthens their evaluative capacity to assess creators’ ethical compliance—whether AI usage is transparently disclosed, whether messages involve misrepresentation, or whether affective manipulation occurs [16]. Although ethical discernment is not formally subsumed within the definition of AI literacy, technical comprehension functions as its necessary antecedent, enabling consumers to appraise AI outcomes and accountability [24] and consequently to inform trust formation. Improving AI literacy thus empowers consumers to make informed judgments about transparency, explainability, and ethical consistency in AISV content. Such cognitive empowerment reinforces users’ autonomy, facilitates more strategic engagement with algorithmic systems, and enhances their ability to achieve desired outcomes within AI-mediated environments [77].

8. Limitations and Future Research Directions

Despite providing valuable insights into AISV influence mechanisms, this study has notable limitations. First, the study employed a single vlog stimulus that explicitly referenced TikTok. While this approach enhanced ecological validity by reflecting the dominant platform among Chinese users, it inevitably constrained content heterogeneity and limited the generalizability of findings to platforms such as YouTube or Instagram. Likewise, the focus on Chinese Gen Z participants, although analytically meaningful for a globally influential cohort, restricts cross-cultural generalization due to regional differences in AI attitudes, technological trust, and cultural orientations. The predominance of young respondents recruited via WeChat further narrows representativeness, underscoring the need for multi-country replications with diverse demographic structures.

Moreover, in conceptualizing AI literacy, we selectively incorporated the technological application, cognitive understanding, and detection identification dimensions from Carolus et al. [85], excluding the ethical dimension owing to its contextual misalignment with passive AISV consumption. Although technical literacy enables consumers to appraise creators’ ethical conduct, the role of consumers’ own moral frameworks in shaping trust formation remains untested. Future studies could integrate ethical literacy to examine how moral alignment interacts with technical comprehension in influencing perceived authenticity and trust.

Furthermore, given the cross-sectional design, the results should be interpreted as associative rather than causal. Future research could use longitudinal or experimental designs to strengthen causal inference. While the present separate SEM for tangible product and experiential service AISVs, experimental designs assigning participants to predefined product types could provide more controlled evidence regarding how product characteristics govern persuasion outcomes. More nuanced distinctions—such as utilitarian versus hedonic tangible goods or experiential versus functional services—may further clarify heterogeneous consumer responses.

Finally, consistent with prior work, this study operationalized AI literacy through self-assessment measures. Although widely accepted and effective for capturing perceived competence, this approach does not reflect objective proficiency, creating potential discrepancies between perceived and actual literacy. Future research could incorporate objective, performance-based assessments alongside self-reports, thereby producing a more comprehensive and triangulated understanding of AI literacy in consumer contexts.

Collectively, acknowledging these limitations facilitates a more precise delineation of boundary conditions and points toward research designs capable of deepening theoretical and methodological rigor in AISV scholarship.

9. Conclusions

This study demonstrates how consumer AI literacy shapes AISV marketing effectiveness. AI literacy influences information adoption through three pathways—emotional value, information usefulness, and source credibility—with information usefulness exerting the strongest effect. AISV type acts as a critical boundary condition: the direct effect of AI literacy on purchase intention was observed in experiential service AISVs, whereas in tangible product AISVs it functioned primarily through information adoption. This study’s contribution to marketing management lies in systematically revealing how AI literacy drives consumer decision-making, addressing a theoretical gap regarding its role in marketing acceptance. For scholars, AILPDM offers a novel perspective, advancing theory from traditional content response frameworks to multi-level models that account for consumer technological capabilities. For practitioners, the findings inform differentiated AISV strategies: Experiential service marketers should emphasize anthropomorphic elements, while tangible product marketers should enhance technological transparency to build AI trust. This theory-driven approach can improve marketing effectiveness and support sustainable competitive advantage in digital environments.