Anthropomorphism and transparency interplay on consumer behaviour in generative AI-driven marketing communication

As David Sacks said in his podcast, “Not using Generative AI will be like trying to compete without knowing how to use Microsoft Office 20 years ago” (Sacks, 2023). Nowadays, Generative AI (GAI), conceived as a more specific sub-discipline within AI, has been widely adopted and well-applied to boost productivity. As technological innovations compel marketers to stay ahead of the knowledge curve (Grewal et al., 2019) and bridge the theories and practice, GAI tools, such as ChatGPT Midjourney, have been widely adopted in marketing practices (Aminifard et al., 2024). Mehta et al. (2022) provided robust and generalisable findings on the relationship between AI and user attitudes and behaviour. Mariani et al. (2022) explored AI usage in marketing, consumer research and psychology. Jain et al. (2023) conducted a comprehensive and hybrid literature review on the connections between AI and consumer behaviour. Similarly, Rabby and Chimhundu (2021) examined how AI influences consumer behaviour in digital marketing.

While these studies offered plentiful insights into leveraging AI in marketing, they do not specifically focus on Generative AI (GAI). GAI has pushed businesses to rethink their strategies for engaging customers and building lasting relationships (Mondal et al., 2023). With GAI, marketers can communicate and reach out to consumers swiftly and automatically by creating responses to consumer-generated reviews, making social media marketing campaigns, or writing GAI-generated response emails, among many others. Considering the rapid adoption and application of GAI, it is necessary to investigate how GAI impacts consumer behaviour through marketing communication.

Much of the research has been dedicated to conceptual and theoretical discussions (Kshetri, 2024; Dwivedi et al., 2023; Mondal et al., 2023), with relatively few empirical studies. The limited empirical work has primarily focused on one type of marketing communication or a few characteristics of GAI (Ho, 2024; Konya-Baumbach et al., 2023; Li and Wang, 2023; Jiang et al., 2022). We also acknowledge that there exist prior studies about GAI’s impact on consumer intention (Chakraborty et al., 2024; Chaisatitkul, 2023; Sohn, 2020) but studies about underlying mechanisms of such effect are still lacking.

Building on the theory of planned behaviour (TPB), which emphasises the role of the psychological process towards behavioural intention and actual consumer behaviour is highlighted (Laheri, 2024; Tiwari, 2023; Vabø and Hansen, 2016; Ajzen, 2011), this study seeks to investigate how AI-generated marketing communication affects consumer behaviour intention through psychological impact of generative AI.

Following the suggestions from Paul et al. (2023), this paper has two primary aims. First, it seeks to expand the breadth and depth of the existing studies by examining how two attributes of GAI-produced marketing (i.e. anthropomorphism and transparency) influence consumer behavioural intent and consumer behaviour intent (i.e. purchase intention) through consumers’ subjective perceptions and reactions. Anthropomorphism refers to the perception that an entity possesses human characteristics. It is a fundamental psychological response shaped by cognitive representations typically acquired during childhood (Bjorklund and Causey, 2017). As GAI is increasingly used to complement and/or to replace a human agent in a marketing interaction, these systems are designed to appear more human-like (Islam et al., 2024; Kshetri, 2024). Consequently, the distinction between human and AI agents has become increasingly blurred, raising the issue of GAI transparency (Kim et al., 2024).

Inspired by Kirkby et al. (2023), this study also investigates whether disclosure that content is AI-generated negatively moderates the relationship between anthropomorphism and consumer behavioural intent. This inquiry reinforces the relevance of anthropomorphised marketing communication. By analysing these relationships, the study aims to deepen understanding of psychological mechanisms underlying consumer responses to GAI-generated marketing content. Second, two forms of GAI-based marketing communication, dynamic interactions (e.g. chatbot conversations) and less dynamic contexts (e.g. promotional sales campaign poster advertisements), were used to test the generalisability of the findings.

This research contributes to the literature in several aspects. First, it reveals how consumers react to GAI-produced content, identifying a chain of psychological effects triggered by exposure to GAI-generated marketing communication. In particular, the study provides empirical evidence of how anthropomorphism and transparency in GAI-generated marketing communication influence purchase intention through perceived social presence and consumer trust. Second, it addresses the gap in the literature by empirically investigating the interaction between transparency and anthropomorphism in GAI-produced marketing communication. Third, it confirms the generalisability of leveraging AI-generated anthropomorphised content across various marketing communication circumstances.

From a managerial perspective, this paper helped marketers better understand how GAI affects consumers’ attitudinal and behavioural reactions. It offers practical guidance on designing effective marketing communication strategies to optimise consumer experience, build consumer trust and enhance purchase intent (Torres‐Moraga et al., 2008).

Kühne and Peter (2023) explored the multidimensional nature of anthropomorphism in human-robot interactions, providing greater conceptual clarity. When applied to marketing, anthropomorphism refers to attributing human-like physical features, motivations, behaviours, emotions and mental states to non-human agents or objects (Waytz et al., 2010; Eyssel et al., 2011). Drawing on the uncanny valley theory, previous studies often identified an adverse effect of robot anthropomorphism on consumer attitudes towards robots (Mende et al., 2019). However, Chung et al. (2023) found that the optimal level of verbal anthropomorphism in social robots varies depending on privacy concerns. This suggests that tailoring verbal anthropomorphism can enhance the consumer experience.

Many studies have examined the effects of anthropomorphism on consumers’ cognitive and attitudinal reactions. Karimova and Goby (2021) found that appropriate anthropomorphism and personality traits influence consumer preferences. Similarly, Xiao and Kumar (2021) showed that anthropomorphism fosters consumer trust, attachment and willingness to engage. Ho (2024) revealed that anthropomorphism strengthens willingness to pay a premium when consumers prioritise hedonic (emotional and novelty) values. While existing research demonstrated that anthropomorphism could lead to favourable marketing outcomes, such as positive attitudinal and behavioural reactions, Konya-Baumbach et al. (2023) noted that studies on the effectiveness of anthropomorphism in marketing contexts remain limited.

This study adopts the social exchange theory (SET) as one of the theoretical foundations to investigate the effect of anthropomorphic generative AI content in marketing. Emerging in the late 1950s, SET has been widely used to understand interpersonal relationships across various fields, such as psychology, sociology and marketing (Homans, 1958; Blau, 2017). The core concept of SET is that relational parties weigh rewards and costs during interactions, ultimately seeking to maximise their reward (Saks, 2006). The theory also suggests that social interaction is a reciprocal exchange of activity (Molm, 1991). Positive actions from one party typically elicit favourable responses from the other, reinforcing relational bonds. SET emphasises that reciprocal exchanges foster trust and commitment between parties (Hsiao et al., 2023). Perception of equity and fairness also plays a crucial role in building and maintaining trust.

SET has been effectively applied in marketing research to explain how factors like perceived value, emotional engagement and reciprocity affect trust (Ghafari et al., 2019; Vosta and Jalilvand, 2022), behavioural intention (Farhana, 2021; Holthausen, 2010; Hsiao et al., 2023; Kim et al., 2022). Kim et al. (2022) confirmed, using SET, that perceived social presence positively affects trust, echoing findings by Konya-Baumbach et al. (2023). In the next section, we discuss further how such social presence may influence consumers’ perceptions of GAI.

Social presence theory (SPT) was initially developed by Short et al. (1976) and refined by Gunawardena and Zittle (1997). It defines social presence as the degree to which a communicating agent is perceived as a “real person” in mediated communication. Social presence is influenced by verbal and non-verbal, such as facial expressions, vocal cues, gestures, friendliness, expertise and physical appearance (Gunawardena and Zittle, 1997). In marketing communication, these cues can comprise names, profile pictures, emoticons and human-like linguistic elements (Konya-Baumbach et al., 2023). An artificial marketing agent can mimic social presence through anthropomorphised content (Araujo, 2018). High-level anthropomorphism means that the technology mimics human-like behaviour, and arguably, this can create a situation where the customer interacts with a human instead of a technology. Previous studies also found that human-like contents and/or interactions generated by the GAI can create a sense of human presence (Sarraf et al., 2024). Research has demonstrated that social presence can enhance trust and enjoyment in consumer interactions (Hassanein and Head, 2007; Qiu and Benbasat, 2009; Ogonowski, 2014; Schuetzler et al., 2020).

Following Bleier et al. (2019), who emphasised leveraging social presence to optimise customer experience, this study investigates whether anthropomorphism in AI-generated marketing communication fosters trust through social presence. In this study, social presence refers to the perceived social presence of the entity delivering GAI marketing communication. Specifically, we investigate these effects within AI-generated communication for insurance products (Konya-Baumbach et al., 2023).

Consumer trust refers to the expectations held by the consumer that the service provider is dependable and can be relied on to deliver on its promises (Sirdeshmukh et al., 2002). This paper employs the trust construct proposed initially by Mayer et al. (1995) and adapted in various contexts, such as by Benbasat and Wang (2005). Kim et al. (2021) defined trust as the extent to which consumers perceive the ability, reliability and integrity of the entity with whom they interact. Trust is a critical factor influencing users’ attitudes and behavioural intentions (Tiwari, 2023). This study treated consumer trust in the entity delivering GAI marketing communication as a first-order construct by aggregating its theoretical dimensions (competence, benevolence and integrity) into a single composite measure (Mayer et al., 1995). While other research often conceptualised trust as multidimensional, we followed previous research that supports treating it as a single construct for simplicity and clarity in analysis (McKnight et al., 2002; Pavlou and Gefen, 2004; Benbasat and Wang, 2005). By combining the dimensions, we aimed to capture the overall essence of consumer trust as perceived by participants.

Behavioural intentions reflect a consumer’s willingness to maintain a sustainable relationship with the service provider (Zeithaml et al., 1996; Gounaris et al., 2007). They include engaging with a service, recommending it, or purchasing. In this study, purchase intention (PI) is a focal component of behavioural intentions. PI represents a consumer’s willingness to purchase a particular product or brand at a specific time or situation (Lu et al., 2014). Based on the TPB, PI is widely recognised as a predictor of actual purchasing behaviour (Tiwari, 2023; Laheri, 2024).

By integrating SPT, prior research highlights trust as a mediator between perceived social presence and behavioural intentions, such as purchase intention (Lu et al., 2016; Jiang et al., 2019; Kim et al., 2021; Yeboah and Afrifa-Yamoah, 2023). Our research builds on SET and SPT; it aims to reconcile existing empirical research findings on the relationship between anthropomorphism, social presence and trust in GAI marketing communication. Specifically, this study suggests that anthropomorphism, defined as more human-like cues, enhances perceived social. Enhanced social presence fosters stronger social connections, facilitating consumers’ engagement and relationship with marketing communication. This strengthens trust in the GAI agent and promotes favourable behavioural intentions, such as purchase intention (c.f., Jiang et al., 2022). Based on this, we propose the following hypotheses:

Transparency is crucial in GAI usage in governance, compliance, legal questions and customer interaction (Dwivedi et al., 2023). Disclosing AI’s involvement is an additional and important facet of this broader openness (Kirkby et al., 2023). While previous studies have primarily focused on conceptual discussions surrounding AI ethics (Hacker et al., 2023; Franzoni, 2023) or specific case analyses of GAI applications (Beerbaum, 2023), empirical research remains limited. Kirkby et al. (2023) proposed a model to explore the impact of transparency on brand image and consumer attitudes. Their findings revealed that text disclosed as AI-generated is not perceived as less authentic than that disclosed as human-written and has no adverse effect on brand authenticity and brand attitude. Similarly, Creyer (1997) highlighted the importance of a firm’s ethical behaviour during consumers’ purchase decisions, suggesting that transparency could influence consumer behaviour.

Transparency and disclosure of information are fundamental elements of sociocultural life (Schudson, 2015). Based on this, transparency in marketing communication may influence perceptions of social presence by revealing the identity behind the message. This suggests that transparency could moderate the effect of anthropomorphism on consumer trust in the entity delivering GAI marketing communication and on purchase intentions through perceived social presence. Drawing from SET, principles of equity and fairness, such as transparency regarding AI usage, encourage consumers to trust and sustain relationships with the other party. Therefore, this paper deduces that transparency reinforces the positive impact of trust on purchase intention. Based on the theoretical framework above, we propose the following hypotheses:

A conceptual model of all hypotheses can be found in Figure 1.

Following Paul et al. (2023), this paper investigates how two aspects of GAI-produced marketing communication (anthropomorphism and transparency) affect consumer trust in generative AI-generated marketing communications and purchase intention. The study evaluates these effects across two distinct types of marketing communication: chatbot, representing a more interactive medium and poster, representing a less interactive medium. To ensure the robustness of our findings, we tested the effects of anthropomorphism and transparency in these two types of media. The interactive chatbot medium allowed for dynamic customer service interactions, while the static poster medium represented a sales campaign with no interactive components.

We have framed this research within the insurance industry for three key reasons. Firstly, effective marketing communication is critical in the insurance industry as it involves selling products, building trust and fostering long-term relationships with customers (Marcos and Coelho, 2018). Secondly, inspired by Vieira et al. (2018), this study focused on utilitarian consumption, which has received less attention than hedonic shopping. Finally, following Jain et al. (2023), this research situates its empirical investigation within a specific context to manage experimental complexity and enhance feasibility.

We conducted our pre-test through an online survey with 100 valid responses (M_age = 30.0; 52% female). Prolific, an online data-collecting platform, was used to recruit the participants for the survey. Prolific has been widely adopted in research for its ability to provide diverse respondents across various consumer groups and countries (see Atalay et al., 2023; Beisecker and Schlereth, 2024; Cheng and Orazi, 2023; Guo et al., 2024; Philipp-Muller et al., 2023). In this study, to ensure external validity, in both the pretest and the main study, we did not discriminate against any consumer group above the consent age (18 years) and willing to fill in the survey. Participants were randomly assigned to one of eight experimental conditions to evaluate the effectiveness of anthropomorphism and transparency manipulations in two marketing communication media, i.e. poster and chatbot (see Table 1). Detailed manipulation materials are presented in eight groups as outlined in  Appendix.

To test the validity of anthropomorphism manipulation, participants rated the perceived anthropomorphism of the marketing communication party (chatbot or poster) using a 7-point scale (1 = strongly disagree, 7 = strongly agree) in response to statements such as “The other party… feels human-like” and “used language that felt natural and human-like.” A one-way ANOVA revealed a significant effect of condition on perceived anthropomorphism [F(7, 93) = 2.905, p = 0.009, η² = 0.179], confirming that the manipulation influenced participants’ perceptions as intended. The perceived anthropomorphism was significantly higher in the high anthropomorphic conditions (M = 5.08, SD = 1.29; M = 5.27, SD = 1.20) than in the less anthropomorphic conditions (M = 3.13, SD = 1.67; M = 3.98, SD = 1.82), supporting the validity of our experimental stimuli.

To validate transparency manipulation, participants rated the perceived transparency of the chatbot or ad using seven-point scales (1 = strongly disagree, 7 = strongly agree) in response to statements such as “The other party… informed me about the AI nature of text generation” and “stated its AI nature at the beginning of our interaction”. Similar to anthropomorphism, a one-way ANOVA also shows a significant effect of the manipulated conditions on the perceived level of transparency [F (7, 93) = 4.066, p < 0.001, η² = 0.234], confirming that transparency manipulation influences participants’ perceptions. Moreover, supporting the validity of our experimental stimuli, descriptive statistics showed that transparent conditions have a higher mean level of transparency (M = 5.365, SD = 1.685 and M = 4.017, SD = 1.623) than non-transparent conditions (M = 3.134, SD = 1.657 and M = 2.312, SD = 1.353).

After confirming the manipulation, we collected a larger data set to test our hypotheses. A total of 444 participants were recruited via Prolific. Consistent with the pre-test, all respondents voluntarily completed the survey and were above the age of consent (18 years). Prolific performed the initial screening, targeting participants with approval rates higher than 80% to ensure reliability. In addition, there were embedded attention checks that aimed to detect inattentive respondents and affirm the eligibility of participants (Waites and Ponder, 2016; Abbey and Meloy, 2017). Specifically, participants were asked to select “insurance” among four options (Banking, Health, Insurance and Education). Eight participants failed this manipulation check. After filtering these responses, 436 responses were valid (M_age = 31.52, 53.2% females). Of the participants, 82.7% were between 18 and 40 years old (n = 367), with the youngest being 18 years old (n = 3) and the oldest being 78 (n = 1). Regarding education, the highest level of completion was a PhD (n = 3) and 25.2% of the participants had a master’s degree. In terms of employment, 58.8% were employed full-time (n = 261), 16.4% had part-time jobs (n = 73) and 18.5% were seeking opportunities (n = 82).

The study used a 2 (high anthropomorphism vs low anthropomorphism) × 2 (transparency vs non-transparency) × 2 (chatbot vs sales campaign poster) between-subjects experimental design. Participants were randomly assigned to one of the eight experimental conditions, as outlined in Table 2. The distribution across the groups was as follows: group 1 (n = 53, 12.2%), group 2 (n = 51, 11.7%), group 3 (n = 52, 11.9%), group 4 (n = 58, 13.3%), group 5 (n = 53, 12.2%), group 6 (n = 56, 12.8%), group 7 (n = 57, 13.1%) and group 8 (n = 56, 12.8%). In each group, participants were exposed to marketing communication content with corresponding characteristics and then asked to rate survey questions regarding consumer behaviour on a seven-point scale (1 = strongly disagree and 7 = strongly agree). The survey also included further questions on personality habits that did not involve privacy-sensitive information and collected demographic information.

The independent variable is anthropomorphism, which was manipulated in a binary way: high anthropomorphism means that the marketing communication content used human-like cues, while low anthropomorphism used machine-like cues. The dependent variable, consumer behavioural intention, was measured as purchase intention. The mediators include the level of perceived social presence and trust towards the entity delivering GAI marketing communication.

The first moderator is transparency, where transparency refers to disclosing that GAI generated the marketing communication message, and non-transparency means no notification about the message generator. The second moderator is the communication scenario in a chatbot or a sales campaign poster. Perceived social presence was measured by asking participants to rate their agreement with statements on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree), adapted from Kumar and Benbasat (2006). The mean of these ratings was used to measure perceived social presence. Similarly, trust towards the entity delivering GAI marketing communication and purchase intention were measured by participants’ agreement with statements on the same seven-point scale, adapted from previous research (Benbasat and Wang, 2005; Li and Wang, 2023; Huh et al., 2023).

The control variables were selected based on extant theories. Researchers widely recognise personality as a critical factor influencing consumer behaviour (Gangai and Agrawal, 2016). Sarker (2013) highlights the need for marketers to understand consumers’ personalities and traits and how these traits shape their behaviour. We thus selected consumption habits and trust disposition as control variables.

Consumption habits control whether people prefer to make purchase decisions independently or to consult others, which affects how much social presence people like in their purchase decision-making process. Trust disposition, as adopted by Lu et al. (2016), reflects an individual’s general propensity to trust or distrust others. Factors such as past experiences, cultural environment and education shape this trait, which significantly impacts trust formation (Cheng et al., 2019; Wang et al., 2022). Its inclusion ensures that pre-existing general trust tendencies do not influence variations in trust. Finally, the tendency to trust AI-generated content was controlled, based on Soni (2023), who found that uncertainty regarding how consumers perceive AI-generated content might hinder AI adoption.

The core constructs were evaluated for internal consistency, reliability and validity using Cronbach’s alpha and item-to-total correlations (Hajjar, 2018). The measurement of social presence, trust and purchase intention demonstrated high reliability (Cronbach’s alpha = 0.926, 0.944, 0.971). All corrected item-total correlations for items within each scale were above 0.30, indicating that all items were strongly correlated with the total score, further supporting reliability (see Table 3).

The correlation matrix in Table 4 highlights several significant relationships. Social presence (SP) showed a strong positive correlation with both trust (r = 0.689, p < 0.001) and purchase intention (PI) (r = 0.690, p < 0.001), indicating that higher social presence is associated with increased trust and purchase intention. Social presence also positively correlated with the tendency to trust GAI (r = 0.445, p < 0.001) and weakly correlated with trust disposition (r = 0.107, p = 0.021). Trust demonstrated significant positive correlations with purchase intention (r = 0.780, p < 0.001) and tendency to trust GAI (r = 0.543, p < 0.001), suggesting that greater trust leads to higher purchase intention and aligns with participants’ willingness to trust AI-generated content. Anthropomorphism showed significant positive correlations with social presence (r = 0.640, p < 0.001), trust (r = 0.523, p < 0.001) and purchase intention (r = 0.564, p < 0.001), reinforcing its role in influencing these variables.

The analysis revealed a moderate positive correlation between anthropomorphism and the tendency to trust GAI (r = 0.369, p < 0.001) and a weaker correlation with trust disposition (r = 0.122, p < 0.001). Transparency positively but weakly influenced social presence (r = 0.193, p < 0.001), trust (r = 0.164, p < 0.001) and purchase intention (r = 0.221, p < 0.001), suggesting a narrow but significant relationship. Interestingly, consumption habits did not significantly influence anthropomorphism, social presence, trust, or purchase intention, except for a weak positive correlation with trust disposition (r = 0.294, p < 0.001). Finally, the tendency to trust GAI positively correlated with trust (r = 0.550, p < 0.001) and purchase intention (r = 0.543, p < 0.001), highlighting the role of general GAI trust tendency in enhancing behavioural intentions. Trust disposition also showed positive correlations with trust (r = 0.128, p = 0.004) and purchase intention (r = 0.157, p < 0.001), justifying its inclusion as a control variable to ensure that pre-existing trust tendencies do not influence variations in trust.

After confirming the core constructs’ internal consistency, reliability and validity, we performed ordinary least squares regression-based path analysis to test all hypotheses. We first conducted a one-way ANOVA to test H1, which examines the direct relationship between anthropomorphism and purchase intention (PI) [see Table 5(a)]. The results revealed a significant main effect of anthropomorphism on purchase intention, F (1, 434) = 202.43, p < 0.001, η² = 0.318. Specifically, participants in the high anthropomorphism condition (M = 5.21, SD = 1.45) reported significantly higher purchase intention than those in the low anthropomorphism condition (M = 3.03, SD = 1.38). These findings support H1.

Having established a significant direct relationship between the independent variable (anthropomorphism) and the dependent variable (purchase intention), we proceeded to test the mediation effects using the Hayes Process Macro (Hayes, 2015). Mediation testing requires that the direct relationship between the independent and dependent variables is significant, which was confirmed in this analysis (Fritz and MacKinnon, 2007).

To evaluate H2, we used Model 6 of the Hayes Process Macro, which examines the sequential mediating effects. For H3a, we used Model 8 of the Hayes Process Macro to test the moderating role of transparency on the relationship between anthropomorphism and social presence. Finally, for H3b, we used Model 92, which allows one moderator to influence the entire mediation path (Stride et al., 2015).

H2: We investigated the mediating role of social presence (SP) and trust in the relationship between anthropomorphism and purchase intention (PI). The findings in Table 5(b) indicate a significant indirect effect of anthropomorphism on purchase intention through social presence and trust (b = 0.49, BootSE = 0.08, 95% BootCI [0.35, 0.66]). Additionally, the direct effect of anthropomorphism on purchase intention remained significant (b = 0.43, SE = 0.14, p = 0.002, 95% BootCI [0.16, 0.70]), indicating partial mediation. These findings support H2, highlighting the importance of social presence (SP) and trust as mechanisms linking anthropomorphism to purchase intention (see Figure 2).

H3a: This hypothesis suggests that transparency moderates the relationship between anthropomorphism and social presence, where disclosing that marketing communication is AI-generated would decrease perceived anthropomorphism levels and subsequently reduce perceived social presence. To test this, we conducted a moderated mediation analysis for social presence, using Model 8 from PROCESS version 4.2 in SPSS, while controlling for consumption habits, trust disposition and tendency with GAI [see Table 5(c) and Figure 3]. The interaction term (Anthro × Transparency) was significant (b = −0.17, SE = 0.07, t = −2.49, p = 0.013, 95% CI [−0.31, −0.04]), confirming that transparency moderates the effect of anthropomorphism on perceived social presence. Conditional effects analysis showed that anthropomorphism had a more substantial impact on perceived social presence at lower levels of transparency (transp = 2.0, b = 2.17, 95% CI [1.80, 2.54]) compared with higher levels of transparency (transp = 6.0, b = 1.47, 95% CI [1.10, 1.85]) [see Table 5(c)].

H3b: H3, which suggests that transparency moderates the indirect effect of anthropomorphism on purchase intention (PI) through perceived social presence (SP) and trust, was tested using Model 92 in PROCESS version 4.2 (Hayes, 2015) [see Table 5(d) and Figure 4]. The analysis controlled for relevant covariates, including consumption habits, trust disposition and experience with GAI. Results revealed a significant conditional indirect effect of anthropomorphism on PI through SP and trust, moderated by transparency. At low transparency (transp = 2.0), the indirect effect was strongest (b = 0.6011, BootSE = 0.1543, 95% CI [0.3112, 0.9234]). This effect weakened at moderate transparency (transp = 4.5; b = 0.4581, BootSE = 0.0777, 95% CI [0.3184, 0.6232]) and further weakened at high transparency (transp = 6.02; b = 0.3746, BootSE = 0.0938, 95% CI [0.2161, 0.5880]).

The direct effect of anthropomorphism on PI was also conditional on transparency, with significant effects at moderate (b = 0.4763, SE = 0.1385, p = 0.0006) and high transparency levels (b = 0.6925, SE = 0.2062, p = 0.0009). However, transparency did not significantly moderate the paths from SP to trust (p = 0.6573) or trust to PI (p = 0.3546). These findings support H3, highlighting that transparency moderates the overall mediation process. Specifically, the indirect effect of anthropomorphism on PI through SP and trust is strongest at lower transparency levels, weakening as transparency increases.

To test the robustness of our findings across different communication scenarios (chatbot vs poster), we used Model 10 in PROCESS version 4.2 (Hayes, 2015). This analysis evaluated whether the indirect effects of anthropomorphism on Purchase intention (PI) through perceived social presence (SP) and Trust remained consistent across scenarios and levels of Transparency. This was inspired by Argo (2020), who demonstrated that other social presence could affect a consumer’s cognition, affect and behaviour, which echoed the theory of social proof and the bandwagon effect (Talib and Saat, 2017; Bindra et al., 2022). Results confirmed that the pathway from anthropomorphism to PI via SP was significant in both scenarios, with chatbots generally amplifying these effects compared to posters; this is consistent with the results of Kollat and Farache (2017) and Li and Wang (2023), who revealed that different marketing communications have different impacts on consumers.

For H3a, Transparency moderated the relationship between anthropomorphism and SP. Higher levels of Transparency reduced the effect of anthropomorphism on SP, with similar patterns observed for both posters and chatbots. These findings further validated the robustness of H2 and H3a, showing that while the context (chatbot vs poster) and Transparency levels influence the strength of these relationships, the fundamental pathways remain stable [see Tables 6 ].

This study contributes to the growing literature on GAI-driven marketing communication by providing key insights that bridge conceptual gaps and extend existing theories. First, it offers robust empirical evidence demonstrating how anthropomorphism influences consumer behavioural intentions (purchase intention) through an integrated mechanism of psychological (social presence) and attitudinal (trust) pathways. While prior research has predominantly focused on conceptual frameworks (Jarek and Mazurek, 2019; Huang and Rust, 2020; Rabby and Chimhundu, 2021), literature reviews (Mariani et al., 2022; Jain et al., 2023), or isolated communication attributes (Jang et al., 2021; Li et al., 2023; Li and Wang, 2023), this study empirically validates the role of anthropomorphism in driving purchase intention. Examining these mechanisms across multiple communication scenarios addresses the gaps in empirical studies that often focus narrowly on single contexts, such as one communication medium or audience segment (Konya-Baumbach et al., 2023).

Second, this study uniquely explores how the critical attributes of GAI-driven communication (anthropomorphism and transparency) and communication scenarios influence consumer behaviour. The findings reveal that transparency moderates the effect of anthropomorphism on social presence and reduces the subsequent impacts of anthropomorphism on trust and purchase intention in the chain of influence. These moderation effects are particularly evident at higher transparency levels, where the impact of anthropomorphism on social presence weakens. Furthermore, the comparative analysis between chatbot and poster contexts highlights how communication scenarios amplify or weaken these effects. Interactive chatbots consistently strengthened the mediating roles of social presence and trust more effectively than static posters, highlighting how different communication scenarios influence consumer engagement [see Tables 6]. These findings provide valuable insights into how GAI technologies can be applied in specific scenarios and offer practical guidance for marketers to tailor communication strategies across different contexts (Granulo et al., 2021; Song and Lin, 2023; Li et al., 2023).

Finally, this research aligns with calls for empirical validation of GAI’s influence on consumer behaviour (Dwivedi et al., 2023; Paul et al., 2023) while linking its findings to foundational theories. The results reinforce SET by showing that anthropomorphism, through social presence and trust, enhances perceived reciprocity and emotional engagement, driving purchase intentions. Similarly, the findings extend SPT by demonstrating that anthropomorphism fosters social presence, which builds trust and strengthens behavioural intentions like purchase intention. Notably, this study highlights the significant role of consumers’ predisposition to trust AI-based content, showing that individuals more inclined to trust AI are more likely to exhibit positive behavioural intentions. This discovery uncovers cognitive and psychological mechanisms underpinning the relationship between GAI and consumer behaviour, addressing gaps highlighted in previous research (Kang and Kim, 2020; Lee and Park, 2022; Huh et al., 2023).

The study provides valuable insights into the practical implementation of GAI in marketing communication. Consistent with the findings of Konya-Baumbach et al. (2023), it confirms that anthropomorphised content effectively enhances consumer reactions. Moreover, it revealed the effective mechanism of how anthropomorphism leads to positive consumer behaviour intentions through increased social presence and enhanced trust. These findings offer clear guidance for managers to effectively incorporate human-like linguistic and graphical cues into marketing communication. It is more efficient to conduct user segmentation and accurate delivery by displaying generated messages to (potential) consumers with a higher tendency to trust GAI, which leads to more expected results and higher returns on investment (Huang and Rust, 2020). From the manager’s perspective, chatbots can enhance customer experience and operational efficiency by automating response generation and supporting existing human staff (Jang et al., 2021; Verma and Kumari, 2023). The study also provides evidence for the benefits of disclosing that GAI generated their marketing communication. While some ambiguity persists regarding whether such disclosure positively or negatively impacts consumer perception (Kirkby et al., 2023), the study’s results indicate that transparency does not significantly alter the positive effects of anthropomorphism. Disclosure of AI-generated content would not undermine the positive effect of anthropomorphism and would adhere to the ethical standards for GAI (Hacker et al., 2023; Kirova et al., 2023).

This study offers empirical support for applying anthropomorphism to GAI-driven communication across various forms and scenarios. The positive effects of anthropomorphism remain consistent regardless of whether communication occurs through chatbots or static content. These findings underscore the versatility of anthropomorphism in enhancing consumer engagement and trust across diverse contexts.

Overall, the findings solidify the benefits of anthropomorphism and extend its applicability, guiding marketers to use GAI technologies more effectively for a range of tasks. For instance, companies can train large language models to adopt a brand personality, enabling more engaging and human-like (higher anthropomorphism level) customer interactions (Wilson and Daugherty, 2018; Karimova and Goby, 2021). Chatbots can be customised to mirror a customer’s communication style, provide emotional support (Huang and Rust, 2020) and save costs for 24/7 customer service (Jarek and Mazurek, 2019) rather than outsourcing human employees. By aligning marketing objectives with the mechanisms of anthropomorphism, companies can design and implement more impactful and efficient strategies.

This research, while providing valuable insights into the role of anthropomorphism in GAI-driven marketing communication, acknowledges several limitations. First, though the results revealed a positive impact of anthropomorphism on purchase intention through social presence and trust towards the agent, it was examined under controlled circumstances. It is worth noting that many important variables such as consumption habits (whether to prefer the social presence of others during the purchase decision-making phase) and trust disposition were under control, which probably affected the function of social presence in the discovered mechanism. Second, the study did not account for many symbolic characteristics specific to the insurance industry. Although prior research, such as Konya-Baumbach et al. (2023), found no significant differences in the effects of anthropomorphism on trust and purchase intention between hedonic and utilitarian products, supporting the methodological choices here. Future research could benefit from incorporating more industry-specific factors, as Idris et al. (2012) and Barbara et al. (2017) suggested. While focusing on insurance specifics distracted from the primary research questions, including such factors in further analyses could enhance the contextual understanding.

Third, the data set was from the Prolific platform with varying demographic information, which led to inconsistent features of the sample. While the findings appear robust across different demo-sociological perspectives, the study lacks precise segmentation of potential customers, limiting its ability to provide detailed consumer insights. Most research on insurance products is limited to specific regions (Lee et al., 2018; Marmol et al., 2021; Idris et al., 2012), considering the differences among economic levels, cultural environments and personal values. The findings would be more practical and instructive if the study were restricted to specific groups of consumers.

Future research should address these limitations and explore additional factors that shape consumer responses to anthropomorphism in GAI-based communication. Subjective variables such as cognitive engagement (Kumar et al., 2023; Li et al., 2023; Lee and Park, 2022; Huh et al., 2023) and personality traits like construal levels and shared values with GAI (Gangai and Agrawal, 2016; Zidehsaraei et al., 2022; Fikouie, 2022) could Additionally, while social presence partially mediates the relationship between anthropomorphism and consumer trust (Konya-Baumbach et al., 2023), other potential mediators and moderators should be investigated. Factors such as the current state of consumer–brand relationships (Li and Wang, 2023) and perceptions of AI intelligence (Nguyen et al., 2022; Song and Lin, 2023) are likely to play significant roles. Expanding the theoretical framework to account for these dynamics would provide a more comprehensive understanding of how anthropomorphism influences consumer behaviour, enabling more effective and refined marketing strategies.

This study investigated how anthropomorphism and transparency would affect purchase intention through perceived social presence and consumer trust in the entity delivering GAI marketing communication across two marketing communication scenarios: chatbots and posters. The findings revealed that anthropomorphism enhances purchase intention directly and indirectly via the mediating pathways of social presence and trust. The positive effects of anthropomorphism on consumer behaviour intentions were significant and consistent across scenarios, with interactive chatbots amplifying these mediating effects more effectively than static posters. Transparency, defined as the disclosure of AI-generated marketing content, moderated the relationship between anthropomorphism and social presence, reducing its impact at higher transparency levels. However, it did not disrupt the mediation mechanism linking anthropomorphism to purchase intention.

In conclusion, this study provides a comprehensive understanding of how GAI-driven marketing communication influences consumer behaviour, contributing to existing theories while offering practical implications for marketers. By demonstrating the robustness of the mediation mechanism across varying contexts, the findings equip marketers with actionable strategies to design effective communication campaigns leveraging anthropomorphism and transparency in diverse marketing scenarios.

Funding: This publication is part of the project LESSEN with project number NWA.1389.20.183 of the research program NWA ORC 2020/21 which is (partly) financed by the Dutch Research Council (NWO).

Figure A1 

Figure A2 

Figure A3 

Figure A4 

Figure A5 

Figure A6 

Figure A7 

Figure A8