Due to their growth background (Liu, 2023) individuals in Generation Z are technology-savvy people (Dania et al., 2023; Wang et al., 2023), using technology information systems in every aspect of their lives (Liu et al., 2023; Sillero Sillero et al., 2023). Therefore, capturing the attention of Generation Z, a promising consumer target (Amatulli et al., 2023), could yield substantial benefits, as they are active online users (Alves de Castro, 2023) and engage in online shopping (Alshohaib, 2024), comprising 30% of the world’s population (Liu, 2023). However, achieving these objectives presents multifaceted challenges, demanding innovative and nuanced marketing approaches.

Previous research either investigated how flow affects behavior intention (Chang and Zhu, 2012; Kim and Han, 2014; Pelet et al., 2017; Wu et al., 2020; Xu et al., 2021; Hyun et al., 2022; Kim, 2022; Zhang et al., 2023) or how advertising variables relate to each other (Jung, 2017; Tran, 2017; Zarouali et al., 2017; Redondo and Aznar, 2018; Abbasi et al., 2021; Niu et al., 2021; Sharma et al., 2021; Alzaidi and Agag, 2022; Lütjens et al., 2022; Lim et al., 2023). Online shopping is more prone to impulse purchases compared to traditional shopping (Wu et al., 2020; Dimara and Parahiyanti, 2024). Existing literature indicates that there is a link between perceived enjoyment and online impulse buying (Wu et al., 2020). In this context, Generation Z users’ perceived entertainment from social media usage (Wijaya et al., 2020) may influence online impulse buying.

TikTok had 1.5 billion monthly active users in 2023 (BusinessofApps, 2024). This platform has become deeply entrenched in the fabric of individuals’ experiences, as an ad on TikTok today could target about 13.6% of all human beings (DataReportal, 2023). In Romania, an important emerging market for the EU, TikTok recorded an advertising sales peak of $1.29 million in March 2023 (Statista, 2023). Almost 65% of Romanian TikTok users are aged between 18 and 24 (Start.io, 2023), corresponding to Generation Z users.

Flow, also known as “optimal experience” (Csikszentmihalyi and LeFevre, 1989; Pelet et al., 2017), was first introduced and approached as a theoretical concept by Csikszentmihalyi (1975), defining it as a “holistic sensation that people feel when they act with total engagement” (Csikszentmihalyi, 1975, p. 36).

Many current studies view flow as a one-dimensional concept (Lavoie and Zheng, 2022; Obadă and Dabija, 2022). Researchers measure it with scales that have three, four, or eight items (Xu et al., 2021; Hyun et al., 2022; Kim, 2022; Zheng, 2023). As flow theory suggests, there is one underlying construct (Lemmens and von Münchhausen, 2023), and this paper’s authors also adopted the unidimensional measurement scale.

Defined by their seamless integration into daily life, social media platforms have become ubiquitous communication tools within the digital landscape. Staying connected is an important aim and benefit of social media use (Miranda et al., 2023). Research confirms that social media is a source of pleasure, also expressed as enjoyment (Pelet et al., 2017; Wu et al., 2020; Qin et al., 2022; Miranda et al., 2023; Roberts and David, 2023; Brailovskaia and Margraf, 2024). Thus, following Csikszentmihalyi’s (1975) study, researchers also adapted the flow concept for the social media context (Pelet et al., 2017; Wu et al., 2020; Xu et al., 2021; Kim, 2022; Brailovskaia and Margraf, 2024).

Chang and Zhu (2012) argue that enhancing the flow users experience is crucial, as it significantly impacts their satisfaction on social networking sites. Several other authors (Kim and Han, 2014; Pelet et al., 2017; Xu et al., 2021; Hyun et al., 2022) supported the same idea, showing that flow, as part of the online experience, has significant positive effects and is a predictor of consumer purchase intention, suggesting that technological knowledge should be combined with flow aspects (Hyun et al., 2022); and flow also enhances social media regular use (Pelet et al., 2017) and online impulse buying (Wu et al., 2020).

Researchers delved into a diverse array of online environments and social media platforms. For example, Roberts and David (2023) conducted an examination and comparison of flow experiences reported on various social media platforms, highlighting that TikTok users experience higher levels of overall flow compared to Instagram users.

This study employed a quantitative methodology to collect data from primary sources, Generation Z TikTok users, through an online survey based on a questionnaire. The overall research design approach involved conducting a correlational study in a single country. Because we analyzed the variables using only one sample of subjects, the study was cross-sectional. The decision was based on our research goal and objectives, which aimed to understand the relationship between the variables in the proposed research model. The authors’ choice to focus their inquiry on the emerging market in Romania is well-founded, given that at the beginning of 2023, TikTok had 7.58 million users aged 18 and older in Romania, and its advertisements reached 46.8 percent of all adults aged 18 and older in this country (DataReportal, 2023). Moreover, in the last three years, TikTok’s advertising sales have increased considerably, reaching its highest point of $1.29 million in March 2023 (Statista, 2023). Hence, Romania serves as a significant global benchmark for testing the proposed model, offering a valuable research context for both theoretical insights and practical applications.

The study participants were Generation Z TikTok users from Romania. We established an age criterion in our online screening questionnaire with the explicit intention of identifying individuals from Generation Z, since this was our research population. We also used filter questions to gather pertinent information, such as age, TikTok usage time, and exposure to TikTok ads.

We developed the online questionnaire by adapting and modifying the unidimensional scales from the literature, which demonstrated excellent psychometric properties in terms of prior reliability and validity (see Table 1). The scales for Motives for TikTok Use (MTU) (construct comprising 7 items), Perceived Ad Intrusiveness (PAI) (construct comprising 2 items), Attitude towards Advertising (ATT) (construct comprising 3 items), Ad Credibility (AC) (construct comprising 3 items), Ad Value (AV) (construct comprising 3 items), Online Flow Experience (OFE) (construct comprising 5 items), and Online Impulse Buying (OIB) (construct comprising 3 items) were all reflective. To evaluate each construct, participants were requested to indicate their degree of agreement with a range of statements via 5-point Likert scales ranging from 1 (strongly disagree) to 5 (strongly agree).

Before administering the questionnaire, we ensured that the items were readily understood by the participants and ran a pilot test involving 120 respondents who were TikTok users from Generation Z. We provided the participants with a draft of the online questionnaire containing all the items, asking them to carefully read them. They highlighted any unclear word or sentence and provided an alternative. The purpose of pre-testing was to evaluate their language comprehension and identify any potential instrumental errors. We made minor modifications to the initial questionnaire design based on the findings of the pilot study.

In December 2023, the researchers collected data from Romanian TikTok users from Generation Z who agreed to participate in the study by clicking the anonymous link in the online questionnaire, using a snowball sampling technique. The first section of the questionnaire described the study and included an online consent form, emphasizing the confidentiality and anonymity of the responses and stating that participation was voluntary. In addition, we employed filter questions to gather pertinent information, including age, time of TikTok usage, and exposure to TikTok ads. We also implemented control questions to ensure response quality and mitigate the drawbacks of online surveys. Sharing the study’s hyperlink with TikTok users gathered a total of 2,928 responses. After an initial assessment of the data quality, we retained 2,406 valid responses for further analysis. We excluded incomplete questionnaires, more specifically those that failed to satisfy the study’s criteria or lacked control questions.

We avoided gathering any data that may potentially reveal the identity of the participants, including their ID number, physical attributes, physiological traits, psychological characteristics, economic status, cultural background, or social qualities (i.e., any personal data). Furthermore, we required parental consent for respondents under the age of 18; therefore, we included an initial section in the online questionnaire to obtain their permission. Overall, we adhered to national laws, specifically Law 677/2001 and Law 206/2004, which govern ethical standards in scientific research in Romania, ensuring the protection of the participants’ rights.

We analyzed the data collected in the study using both descriptive and inferential statistics. We obtained the descriptive statistics for the measured variables, including means and standard deviations, using the SPSS software, version 23. We performed the inferential phase of our study by employing structural equation modelling (SEM) with Smart PLS 4 to test the hypotheses. Considering that our objective was theory-building, the SmartPLS software is appropriate for this type of investigation (Ringle et al., 2015). Structural Equation Modelling (SEM) is a valuable analytical method that allows for the simultaneous execution of factor analysis and path analysis. Following Kline’s (2016) recommended two-step modelling technique, we initially assessed the outer model (i.e., measurement model), followed by testing the direct relationships in the inner model (i.e., structural model).

In terms of relevant demographic information, participant ages ranged from 15 to 27 years (M = 21.23; SD = 2.31). Most respondents were between 20 and 22 years old (55.37%; n = 1,086). In terms of gender, 63.30% (n = 961) identified as female, while 36.70% (n = 562) identified as male. A majority (52.54%; n = 1,264) graduated from high school and resided in an urban area (64.30%; n = 1,547). Over half of the respondents (55.10%; n = 1,325) reported low income, and the majority (51.70%; n = 1,204) were students. About half of them (49.80%; n = 1,198) claimed to spend 2 to 3 h a day on TikTok.

We evaluated the measurement models in the initial phase using SmartPLS 4.0’s structural equation modelling capability. We tested every reflective construct from the conceptual model for internal consistency and validity. Table 2 provides the item loadings, reliability indicators, average variance extracted (AVE), and VIF values. The loadings exceed the minimum threshold of 0.70 (Hair et al., 2010), indicating the presence of convergent validity in all the measured items. The minimum and maximum values lie within the range of 0.755 (OFE5 item) and 0.942 (PAI2 item), thereby satisfying the minimum criteria. We evaluated reliability using Cronbach’s α, with a threshold of 0.7 or greater considered satisfactory for confirmatory analysis (Henseler and Sarstedt, 2013). All reliability scores exceed 0.7, thereby confirming the internal consistency of the model. All AVE values surpassing 0.5 signify a sui model and support the convergent validity of the constructs (Chin, 1998). When the composite values exceed 0.7, the composite reliability (CR) indicates adequate reliability of the constructs (Hair et al., 2010). Additionally, we assessed the level of collinearity between the items in the measurement model. The dataset does not demonstrate multicollinearity, as the highest VIF value among all items is 2.769 (OFE3 item), which is lower than the collinearity analysis threshold of 3.3 as established by Hair et al. (2017). Consequently, the existence of common method bias (CMB) was not an issue.

We have assessed discriminant validity for each concept using the Fornell-Larcker (Table 2) and Heterotrait-Monotrait (Table 3) criteria. Following the Fornell-Larcker standards (Hair et al., 2010; Henseler and Sarstedt, 2013), the AVE value for each latent variable should be greater than the correlation coefficient between that variable and all of the other variables.

To ensure that the concepts are not conceptually identical, we applied the Heterotrait–Monotrait criterion. All construct values in the present study are found to be lower than the critical value of 0.9 (Henseler et al., 2015), which provides evidence for the constructs’ discriminant validity.

After evaluating the measurement model, in the next phase, we employed the bootstrap method to test the hypotheses and determine relationships between the latent variables. To properly assess the structural model, it was important to examine the constructs’ collinearity. The inner model’s highest VIF values are 1.744 (AC → AV and ATT → AV), which is below the critical value of 3.3 (Hair et al., 2017). This suggests that there is no multicollinearity among the constructs. The square root mean residual (SRMR) was 0.046, which meets the minimum threshold of 0.08 and indicates that the saturated model’s fit quality is acceptable. As highlighted in Figure 2, Motives for TikTok Use explain 0.3% (R² = 0.003) of the variance of Perceived Ad Intrusiveness. Furthermore, 44.2% (R² = 0.442) of the variance of Ad Credibility is explained by Attitude toward Advertising and Motives for TikTok Use, while Perceived Advertising Intrusiveness explains 10% (R² = 0.100) of the Attitude toward Advertising. Furthermore, Attitude toward Advertising and Ad Credibility account for 62.7% (R² = 0.627) of the variance of Ad Value. Motives for TikTok Use and Ad Value explain 42.6% (R² = 0.426) of the variance of Online Flow Experience. Finally, 27.4% (R² = 0.274) of the variance of Online Impulse Buying is explained by the Online Flow Experience and Ad Credibility. Also, the model has good predictive power for the endogenous latent variable since the Q² values of the exogenous variables are >0 (PAI = 0.002, ATT = 0.008, AV = 0.017, OIB = 0.025, AC = 0.031, and OFE = 0.415) (Hair et al., 2017).

As shown in Table 4, all ten hypotheses (H1–H10) were empirically validated. The Cohen f² effect magnitude represents the influence of exogenous variables on endogenous variables. According to Cohen (2013), effect size values of 0.02, 0.15, and 0.35 are categorized as small, medium, and large, respectively.

Hypothesis 1 posited that Motives for TikTok Use (MTU) have a negative impact on Perceived Ad Intrusiveness (PAI). The findings (β = −0.059; t-value = 2.758; p < 0.006) indicate a weak yet statistically significant negative effect; therefore, H1 is supported.

As deduced from Hypothesis 2, Ad Credibility (AC) was presumed to be positively impacted by the Motives for TikTok Use (MTU). The findings indicate a small but statistically significant relationship (β = 0.080; t-value = 4.815; p < 0.001); thus, H2 is validated.

Hypothesis 3 postulated that there is a positive effect of Motives for TikTok Use (MTU) on the Online Flow Experience (OFE). The results (β = 0.625; t-value = 39.541; p < 0.001) indicate a positive, large effect, leading to support H3.

Hypothesis 4 asserted that there is a negative relationship between Attitude towards Advertising (ATT) and Perceived Ad Intrusiveness (PAI). The outcomes (β = −0.317; t-value = 13.854; p < 0.001) indicate a small yet statistically significant negative impact; thus, hypothesis H4 is validated.

Hypothesis 5 assumed that Attitude toward Advertising (ATT) would be positively associated with Ad Credibility (AC). The results (β = 0.643; t-value = 43.617; p < 0.001) suggest the existence of a positive, large effect; thus, hypothesis H5 is supported.

Hypothesis 6 stipulated that Attitude toward Advertising (ATT) has a positive influence on Ad Value (AV). The findings (β = 0.440; t-value = 22.702; p < 0.001) confirm the presence of a positive, moderate effect, providing support for hypothesis H6.

Hypothesis 7 presumed that Ad Value (AV) is positively impacted by Ad Credibility (AC). The outcomes (β = 0.430; t-value = 21.725; p < 0.001) demonstrate the existence of a moderately positive impact that provides support for hypothesis H7.

Hypothesis 8 proposed that Ad Credibility (AC) will be positively associated with Online Impulse Buying (OIB). The study insights (β = 0.499; t-value = 30.582; p < 0.001) indicate the presence of a moderately positive impact that provides support for hypothesis H8.

According to Hypothesis 9, Online Flow Experience (OFE) will be positively impacted by Advertising Value (AV). The results (β = 0.098; t-value = 5.697; p < 0.001) pinpoint the existence of a small yet statistically significant effect; thus, hypothesis H9 is validated.

Finally, Hypothesis 10 assumed that Online Flow Experience (OFE) exerts a positive influence on Online Impulse Buying (OIB). The findings of the study (β = 0.083; t-value = 4.422; p < 0.001) highlight the presence of a small yet statistically significant impact; therefore, the H10 hypothesis is confirmed.

Based on the research assumptions, the results indicate:

As we already argued, previous research focused on the effects of flow on behavioral intention (Chang and Zhu, 2012; Kim and Han, 2014; Pelet et al., 2017; Wu et al., 2020; Xu et al., 2021; Hyun et al., 2022; Kim, 2022; Zhang et al., 2023) or advertising latent variable relations (Jung, 2017; Tran, 2017; Zarouali et al., 2017; Redondo and Aznar, 2018; Abbasi et al., 2021; Niu et al., 2021; Sharma et al., 2021; Alzaidi and Agag, 2022; Lütjens et al., 2022; Lim et al., 2023), without identifying specific outcomes of these relations shaped by Generation Z users. Therefore, our study is the first to investigate these relationships, thus having an important theoretical and practical contribution.

The study reveals that motives for TikTok use negatively impact perceived ad intrusiveness. Contrary to previous research suggesting intrusiveness interrupts consumers’ goals (Li et al., 2002), the study suggests that people prefer TikTok ads due to their unique features, such as non-interrupted content, adaptability, user-friendliness, and the ability to swipe away unappealing advertisements (Han, 2020). People perceive advertising as transparent, making it easy for them to learn about the features and effects of the products from other comments (Han, 2020). The study confirms Lütjens et al.’s (2022) finding that the irritation effect decreases over time, likely due to consumers’ habituation to emerging technologies and their evolving sensitivity to diverse touchpoints, a novel relationship in the current research. Users like ads on TikTok (Han, 2020). TikTok’s vertical video ads, which effortlessly integrate with Generation Z’s usage habits and require minimal effort, may be the reason for this (Mulier et al., 2021). Lütjens et al.’s (2022) conclusions are in line with our results, revealing that, as the authors assumed, young users are already accustomed to new technologies and, as a result, users’ attitudes towards ads are already formed before engaging on the social media platform. All these might explain the positive relationships between attitude towards advertising and both Ad Credibility and Ad Value, two relations representing novelties of the current study. This finding also suggests that users’ motivation to use TikTok positively impacts ad credibility and online flow experience, as adapted ads increase credibility and users appreciate these ads, enhancing the overall user optimal experience. Moreover, this study identified a negative relationship between individuals’ attitudes towards advertising and perceived ad intrusiveness. These results are in line with the research conducted by Cortés and Vela (2013), which demonstrates a negative correlation between perceived ad intrusiveness and attitude towards advertising. Tran (2017) and Lütjens et al. (2022) further emphasize that avoidance or irritation, which are linked to intrusiveness, have a detrimental impact on ad attitudes. Additionally, Redondo and Aznar (2018) identify a negative relationship between attitude towards advertising in the online environment and perception of intrusiveness.

According to the research’s findings, ad credibility positively influences ad value, which is consistent with earlier studies on SMS advertising perception by Sharma et al. (2021), mobile advertising by Kim and Han (2014), and in-game pop-up ads by Abbasi et al. (2021). This study provides an important insight into the fact that ad credibility is positively associated with online impulse buying. This study, to the best of our knowledge, is the first to examine this specific relationship. The Alzaidi and Agag (2022) paper, for example, studied the relationship between ad value and purchase intention. From this point of view, this relationship is one of the novelties of the current study.

In the proposed and validated model, the authors found that ad value positively impacts the online flow experience, which in turn positively influences online impulse buying. Following the research directions proposed by Kim and Han (2014) to examine the impact of ad value on flow and the relationship between flow and purchase intention, the study focused on investigating the relationship between online flow experience, ad value, and online impulse buying for the first time. This is another novelty in our study.

The present study also focused on the effect of ad value on online flow and the relationship between ad credibility and online flow experience on online impulse buying. This approach is consistent with Wu et al.’s (2020) suggestions for future research, which outline the need to learn more about online impulse buying through the lens of a certain type of media communication (TikTok), and Qu et al.’s (2023) advice to investigate the role of online flow in influencing impulse buying in scarcity-induced purchases. Therefore, the confirmation of these relationships represents another novelty of the current study.

In conclusion, understanding the factors that drive online impulse buying among Generation Z on TikTok in an emerging market depends on factors such as individuals’ perceptions, attitudes, and motivations. This research is unique as it examines the drivers of online impulse buying from two perspectives: the advertising dimensions, which originate from the communication field, and the motives and flow experience, which stem from the positive psychology field.

Theoretically, this article explores the factors driving online impulse buying on TikTok, particularly from Generation Z’s perspective, including motives for TikTok use, perceptions, and attitudes towards advertising on TikTok and online flow experience. According to the research, when users engage with ads, the motivation to use TikTok does not create intrusiveness, possibly due to the ability to swipe out ads as soon as they appear. Our investigation reveals that attitude towards advertising positively influences Ad Credibility and Ad Value, suggesting that Generation Z users already have an ad attitude due to their experience with social media platforms. The current study revealed, as a new insight, that attitude towards advertising has a positive influence on Ad Credibility and Ad Value. Another novelty of the study is that the authors approached online impulse buying, keeping in mind that prior research focused more on purchase intentions. This approach led to the discovery of a positive relationship between ad credibility and online impulse buying, as well as a positive relationship between online flow experience and online impulse buying. This study provides new insights for future research on the topic.

The study has important practical implications. Communication and marketing managers can gain important insights from our investigation, which reveals that users are curious about the next TikTok video and use the platform as a habit, bringing joy, adaptation, and social integration. Furthermore, the motives to use TikTok impact the perception of ads, with highly-rated motives not being considered distracting or disturbing. Practitioners, particularly those targeting Generation Z, can benefit from this insight, as they perceive ads as neither intrusive nor irritating. However, we should conduct studies to identify the motives and scores of other generational cohorts, as higher scores tend to indicate less intrusive ads. High scores on motives to use TikTok also increase the perception of ads as credible and improve the online flow experience, making platform engagement enjoyable, funny, and immersive. The results of our research suggest that companies should focus on content-driven advertising that provides solutions and insights tailored to users’ needs and wants. If users are constantly exposed to content-driven advertising providing solutions and insights adapted to their needs and wants, then a positive attitude is created, leading to the assessment of ads as trustworthy, valuable, and important. This positive attitude contrasts with intrusiveness and enhances the perceived ad value, leading to an enjoyable, immersive experience on the platform (i.e., an online flow experience). Therefore, adapted ad content over time shapes a positive attitude toward advertising, which, in turn, creates a positive perception of ads in terms of credibility and value and an optimal experience on the platform. The positive effect of ad credibility perception on impulsive buying suggests that, if an ad on TikTok is trustworthy, Generation Z users may easily “buy now and think about it later” without giving it much thought, as this aligns with their current mindset. This insight has profound managerial implications, especially in the case of impulsive buying of products (e.g., clothes, perfumes, etc.). Moreover, this consequence is easier to attain on TikTok, supported by the positive effect of the online flow experience over online impulse buying. Finally, we argue that communication and marketing managers need to focus on building the credibility of their communication efforts over time to avoid losing their audience due to users’ tendency to buy online impulsively, even though the TikTok platform has proven to be an enjoyable, funny, and immersive one that encourages impulsive buying.

Future research might employ objective measuring techniques, such as the experience sampling method (ESM), for online flow experience. Also, our research focused only on Generation Z of TikTok users, but future research could explore the perspectives of different audience cohorts, comparing Xers and Baby Boomers of this social media platform with others (e.g., Facebook, X, YouTube, etc.). Other empirical studies could pinpoint if user experience and education have or do not have a relevant role in explaining how individuals respond to perceived invasiveness. Moreover, the negative impact of motives for TikTok use exerted on perceived ad intrusiveness partially approaches a future research direction suggested by Niu et al. (2021), specifically examining factors that influence perceived invasiveness. Furthermore, given the dynamic nature of social media platforms, we contend that further investigation is necessary to comprehend the formation and evolution of attitudes towards advertising. Comprehensive qualitative and quantitative research could potentially provide valuable insights into enhancing the perceived credibility and value of advertisements. Subsequent research provided a comprehensive understanding of the interplay between the findings of this study and the unique cultural and behavioral characteristics of Romanian Generation Z individuals, which could be a useful insight for practitioners operating in emerging markets.

This study contains inherent limitations that could potentially be addressed in future research. We restricted our investigation to Romania, a burgeoning EU market; consequently, the findings are not universally applicable to other social contexts and necessitate interpretation within their specific cultural context. Future studies should take a comparative approach, either by including more countries from Central and Eastern Europe or by comparing developed and emerging countries using probability sampling. Another limitation of the study is the use of a self-administered questionnaire as a method of measuring the variables, which is a form of retroactive assessment.

Finally, this study used cross-sectional data, analyzed by regression, which means these results could be based on spurious correlations. For example, the flow of causality in Figure 1 may in fact be in the reverse direction (impulse buyers justify their behavior by bolstering their attitude toward advertising).