Introduction

With the rapid advancement of the internet and digital technologies, tourism live streaming (TLS) has emerged as a dynamic mode of information dissemination that surpasses traditional text and images. Unlike static information transmission, TLS leverages real-time interaction, enabling viewers to acquire instant tourism information while engaging with streamers1. This interaction enhances viewers’ awareness and emotional connection with tourism destinations, facilitating two-way communication and increasing immersion and participation2. According to data from China Tourism Academy, 2020 was hailed as the “Year of Live Tourism,” with 26.3% of Chinese consumers watching live streaming of cloud tourism. The rapid rise of tourism streamers on short-video platforms like TikTok has demonstrated the immense potential of live streaming in tourism marketing3. For instance, Dong Yuhui’s Lijiang tourism live streaming attracted 5 million viewers and generated over 50 million RMB in sales.

TLS has also shown rapid growth globally, not only in China. According to Streams Charts, in the third quarter of 2024, global viewers collectively watched 19.8 billion hours of live-streaming content, with YouTube leading as the dominant platform, demonstrating the broad reach of live streaming. In the Asia–Pacific region, Ctrip has established a new TLS center in Thailand, leveraging the country’s mature live-streaming ecosystem and abundant resources for tourism content creation to attract international viewers successfully. In 2023, Ctrip launched its “Super Global Travel” live-streaming series in Thailand, selling over 20,000 hotel rooms and significantly boosting inbound tourism demand. Following this, similar live-streaming events in Tokyo, Singapore, Seoul, and Hong Kong further increased tourism demand in these regions. These examples highlight that TLS demonstrates significant marketing potential within the global tourism industry and has become a key tool for engaging viewers and driving tourism-related consumer behavior. It plays a crucial role in attracting target markets and enhancing industry competitiveness4.

TLS effectively shortens the psychological distance between viewers and destinations by integrating exploration, social interaction, and entertainment elements5. It stimulates sensory experiences and tourism desires, providing an innovative approach to scene-based marketing6,7. However, despite its excellent performance in attracting viewers and driving sales, the success of TLS relies not only on content creativity and appeal but also on maintaining a strong and active viewer base. Any minor dissatisfaction or negative experience may lead to viewer attrition, which poses a significant challenge for TLS. Therefore, enhancing long-term viewer engagement and retention has become an urgent issue for live-streaming platforms and content creators to address.

TLS is an emerging research area that has gradually become a focus of academic attention. Existing studies mainly concentrate on aspects such as live streaming experiences8,9, viewer attention behavior10, purchasing behavior11,12, and tourism intentions13, exploring how live streaming enhances viewer interactivity and engagement to boost tourism intentions and purchasing behavior11. These studies offer insights into the positive factors that drive sustained viewer attention. However, research addressing viewer retention remains insufficient despite its critical role in ensuring the sustainability and stability of live-streaming marketing effectiveness14,15. Existing studies mainly focus on analyzing the drivers of continuous viewer attention, such as the interaction between live streamers and viewers8 and the impact of digital availability on impulsive consumption and viewer intentions15. However, there has been limited research on the reverse factors that lead to viewer disengagement. This gap in research has resulted in a lack of a comprehensive understanding of the mechanisms behind viewers’ non-continuous following intention (NCFI), especially in terms of the negative factors contributing to viewer attrition. The factors influencing NCFI are complex and diverse, requiring more in-depth and systematic research to reveal the underlying mechanisms, thereby providing effective retention strategies for live-streaming platforms and content creators.

This study applies the Push–Pull–Mooring (PPM) theory to address this research gap. This study explores the impact mechanisms of the push, pull, and mooring dimensions on viewers’ NCFI by developing a research framework through a literature review and in-depth interviews. The research not only helps fill the gap in the academic literature regarding viewer attrition and retention rates but also provides practical guidance for live-streaming platforms, content creators, and tourism practitioners. It will help them better understand and address the issue of viewer retention, thereby enhancing the market competitiveness of live-streaming content and increasing viewer loyalty and engagement.

Literature review

Conceptual connotations of TLS

As an interactive media form relying on digital platforms, TLS offers a unique tourism experience through live audiovisual content. It emphasizes immersive destination experiences, including local food and culture, stimulating viewers’ interest in visiting the destination in person15,16. Overcoming spatial limitations and enhancing authenticity and interactivity increase viewers’ sense of participation and social interaction and effectively promote the destination8. The academic definition of TLS is still evolving. Hilvert-Bruce et al.17 describe it as an internet-based multimedia entertainment form emphasizing real-time interaction between streamers and viewers. Meanwhile, Su et al.9 focus on the role of tourism streamers in highlighting destination features and stimulating viewer participation.

TLS relies on digital platforms, such as YouTube, Instagram, and TikTok, which provide technical support and interactive spaces for tourism-related live streaming and have become the core medium for showcasing tourism content5. Platforms refer to internet-based digital spaces that provide the technological foundation for real-time interaction, information exchange, and social functions between content creators and viewers18,19. This study views TLS as a way to showcase tourism scenes vividly using digital platforms. By offering immersive virtual tourism experiences, TLS can enrich viewers’ engagement and enhance their loyalty and retention. Additionally, it effectively increases the visibility and promotional impact of destination services.

The literature review of viewer behavior in TLS

Research into TLS viewer behavior primarily examines the drivers behind viewing, touring, and purchasing decisions. Liu et al.20 identified key motivational factors, including social belonging, media engagement, and relaxation, that shape viewers’ viewing motives through interviews and empirical research. Studies suggest that live streaming offers a dynamic, interactive visual experience that differs from static visual media5,13, allowing viewers to virtually follow the trajectory of tourists in real-time and stimulating “online unconscious motives.”

Tools like the Stimulus-Organism-Response model and motivation theory have been employed to study tourists’ engagement with tourism and purchasing motives. Li et al.20 found that customer participation enhances purchasing intentions, while Zhang et al.13 noted that destination image and interaction can build trust and influence tourism intentions. Additionally, the unique features of tourism e-commerce live streaming—interactivity, authenticity, and entertainment—along with traffic experience and trust positively influence purchasing intentions20. Yu and Xie6 discovered that entertainment atmosphere and celebrity appeal in tourism e-commerce could boost viewer purchasing intentions via place attachment.

Despite these findings linking live streaming features to viewer motives, current research fails to explain the reverse perspective of why viewers disengage from live streaming. A holistic theoretical model to account for non-continuation behavior is missing, and there is a dearth of empirical studies on this aspect, hindering a thorough comprehension of viewer behavior complexity in TLS.

Research on NCFI

Definition of NCFI

In TLS, viewers’ attention behavior is a one-way participation process that does not require approval from the streamer16. Habits often drive this shift in participation mode and can lead to discontinuous use or self-regulation behaviors, similar to NCFI21. However, NCFI in TLS has specific characteristics distinguishing it from disengagement behaviors on other social media platforms. Specifically, NCFI in TLS differs significantly from disengagement behaviors on traditional social media regarding non-continuous following motivation, tolerance for non-instant interaction, and expectations for personalized feedback. Detailed content is presented in Table 1.

Table 1 Comparison of NCFI in TLS and disengagement on traditional social media.
Full size table

Based on the above characteristics, NCFI in TLS mainly manifests as viewers reducing their watching time, unfollowing, or switching to other live-streaming channels when they perceive that the content does not meet their expectations or that the interactive experience is poor. These behaviors reflect viewers’ psychological and behavioral changes under high interaction costs and personalized needs.

The literature review of NCFI

Research on social media NCFI investigates viewer detachment, usage cessation, or platform-switching behaviors. Findings suggest that reciprocity23, information overload24, and viewer exhaustion21 influence social media engagement. Viewer transfer studies differentiate between migrations between various media and shifts between products or services within the same platform. For instance, Hsieh et al.25 examined the transition from blogs to microblogs. Kwak et al.26 identified information, relationship longevity, reciprocity, and network overlap as key factors in microblog viewer migration. Zhang et al.27, from a social exchange perspective, argued that perceived costs and benefits drive NCFI on WeChat subscriptions.

However, research on NCFI in TLS remains under-researched. Unlike other live content formats, TLS focuses on environmental interaction, destination showcasing, and viewer engagement with tourist experiences, cultivating a community around shared interests rather than purely transactional interactions2,15. This distinct context calls for in-depth exploration into the determinants of NCFI specific to TLS.

PPM theory

The PPM theory provides a comprehensive explanatory framework for forming individual behavioral intentions by identifying push, pull, and mooring factors that influence behavior change28. This theory has been widely applied in marketing management and has been further extended in the field of information technology. For example, Fu et al.29 studied Facebook user churn by considering fatigue and dissatisfaction as push factors while viewing personal norms and habits as mooring factors. Tang et al.30 developed a non-continuous motivation model for the brand Weibo, with information quality and service as push factors, alternative attractiveness as pull factors, and perceived non-compliance costs as mooring factors. Zhang et al.31 used the PPM theory to analyze bloggers’ switching intentions, identifying viewer satisfaction as the main push factor, alternative services as pull factors, and switching costs as mooring factors. These studies indicate that the PPM theory is adaptable to various contexts and helps analyze complex user behavior shifts.

This study uses the PPM theory as a theoretical framework to analyze NCFI in TLS, demonstrating its applicability in several aspects. First, the PPM theory helps integrate the influence of various stakeholders, especially the original live-streaming accounts, alternative live-streaming accounts, and individual viewers. These factors shape viewers’ behavior and affect their retention through push, pull, and mooring mechanisms2,32. Second, the PPM theory can explain the effects of these various influencing factors on forming NCFI, such as positive promotion or negative hindrance. This analytical framework helps uncover the distinct roles of each stakeholder and their combined effects on viewer behavior, making it particularly suitable for the complex behavioral patterns of viewers in the TLS industry. Finally, the PPM theory considers external situational factors. It incorporates the viewers’ internal psychological factors33, scientifically explaining how factors like content preferences, interactive experiences, and personalized live-streaming matching collectively influence NCFI. This multi-stakeholder and multi-effect-based analysis approach aligns well with the complexity of viewer behavior in TLS, offering more comprehensive strategic guidance for practical operations.

In conclusion, the PPM theory provides a solid theoretical foundation for this study. It can integrate multidimensional factors while capturing the high dynamism and individual psychological drivers unique to TLS. Its application helps to better understand behavioral shifts in the TLS context and provides theoretical support and practical insights for related research and practice.

Influence factors of NCFI in TLS

By reviewing the literature and observing reality, this study identifies the potential influencing factors: psychological contract breach32, viewer-live streamer social distance34, alternative attractiveness35, customer complaining behavior36, and perceived controllability37. Based on the PPM theory, the above factors are classified to analyze the causes of NCFI in TLS as systematically as possible.

Push factors

Psychological contract breach

Psychological contract breach is a key push factor influencing viewers’ NCFI in TLS. This violation occurs when viewers form negative perceptions of their live-streaming experience32. Rousseau first proposed the concept of psychological contract38, suggesting that when consumers feel that a service provider has failed to fulfill an implicit agreement, a violation occurs, leading to a decline in loyalty. Due to the complexity of psychological contracts, unmet expectations increase the likelihood of contract violation39.

Psychological contract breach consists of transactional and relational psychological contract breach40. A transactional psychological contract breach refers to situations where there is a violation or failure to meet expectations in the economic exchange between viewers and streamers. Specifically, in TLS, a streamer may recommend a product from a tourist destination during a live session, claiming it is high quality and unique. However, if viewers purchase the product and find its quality is subpar and does not match the streamer’s description, they may feel deceived, leading to a crisis of trust. Furthermore, suppose viewers purchase tourism products or tickets through the live-streaming platform and later discover that the prices are higher than those on other platforms. In that case, they may believe that the streamer has failed to fulfill their promise of offering “discounts and convenience,” which further diminishes trust in the streamer.

On the other hand, a relational psychological contract breach refers to the breakdown of trust and reciprocal relationships between viewers and streamers on a social and emotional level40. For example, an emerging streamer may initially establish sincere and intimate interactions with viewers, presenting a humble and relatable image that fosters a sense of connection. However, as the streamer’s popularity grows, they may exhibit an arrogant attitude, reducing or neglecting interactions with viewers. This shift can create a sense of increased social distance, leading to disappointment and a crisis of trust.

In TLS, a psychological contract breach significantly undermines viewers’ trust in the streamer, and trust is crucial for maintaining continuous viewer engagement. Viewers may reduce their viewing frequency and switch to alternative content when a violation occurs, further destabilizing the live streaming ecosystem36,40.

Customer complaining behavior

Scholars generally consider customer complaining as a part of service failure, referring to the actions or non-actions displayed by customers due to dissatisfaction41. In the context of TLS, customer complaining behavior can be defined as a series of behavioral or non-behavioral reactions triggered by perceived dissatisfaction during the viewing process of TLS viewers36.

Perceived controllability

Perceived controllability is crucial for consumers to assess service failure causes, indicating their belief in the manageability of service issues42. In TLS, operators facilitate vibrant tourism product presentations and foster trust between live streamers and viewers. Perceived controllability pertains to viewers’ views on live streamers’ power to prevent a psychological contract breach. High perceived controllability suggests that viewers attribute live streaming failures to the streamers’ lack of ability, believing that the streamer possesses the necessary conditions and capabilities to avoid failure. Conversely, when the perceived controllability is low, viewers are more likely to attribute live streaming failures to external factors beyond the streamer’s control, considering these factors to be outside the streamer’s capacity to manage and thus assigning only secondary responsibility to the streamer37.

Mooring factor: viewer-live streamer social distance

Per the PPM theory, mooring factors are personal or social elements that can sway viewer behavior32, impacting NCFI despite evident push and pull factors. In TLS, the viewer-live streamer social distance characterized by similarity, familiarity, and closeness can shape viewers’ connections to the content43. Originally denoting social class disparities, social distance now informs research on norms, emotions, and interactions. Helfgott and Gunnison43 interpret social distance emotionally, while Liviatan et al.44 suggest that familiarity fosters identification. Thus, as a measure of similarity, social distance influences interpersonal closeness. This study examines the viewer-live streamer social distance to understand its role in interaction and as a mooring factor in NCFI.

Pull factor: alternative attractiveness

In TLS, the pull effect is the allure that draws viewers to engage with different live streamers. Alternative attractiveness significantly influences viewers’ propensity to switch, with limited options leading to continued viewing of the current live streamer45. Ping46 regarded alternative attractiveness as consumers’ evaluation of choices, while Jones et al.47 considered it to reflect viewers’ positive attitudes toward competitive substitutes in the market. Dai and Deng48 pointed out that alternative attractiveness is the degree to which viewers are attracted to other platforms that provide similar functions while using the current service. Therefore, this study defines the alternative attractiveness of TLS as the satisfaction viewers gain by following other TLS49.

Study I: the attribute dimensions of alternative attractiveness in TLS

This study performed with relevant guidelines. The concept of alternative attractiveness in TLS remains underexplored, with a deficiency in measurement tools. At the same time, it has been characterized by various dimensions in other research areas, such as cost, variety, and service quality in online retail50. Its application in live streaming requires further investigation due to its distinct features like shared interests5,15, immersive experiences, and interactive depth.

Study I aims to identify the attributes of alternative attractiveness unique to TLS. Employing qualitative methods and utilizing Nvivo.11 for data coding, this study seeks to establish a basis for accurately measuring and applying alternative attractiveness in future research, enhancing our understanding of how viewers perceive the allure of alternative TLS.

Data sources

This study’s data collection combined face-to-face and online interviews. Participants were viewers who had long followed specific TLS streamers and had strong language skills. The platforms involved in the interviews included WeChat, Xiaohongshu, and Douyin. Participants were recruited by joining fan groups, interacting with related social media posts, and utilizing snowball sampling. This recruitment method effectively targets viewers with extensive experience in watching TLS, ensuring they have a deeper understanding of TLS and providing more reliable and high-quality data support for the study.

The sample selection followed the principles of qualitative research, emphasizing the depth of the sample and data saturation51. We selected individuals with a deep understanding and personal experience of TLS, particularly those who had initially followed a specific streamer and later transitioned to following other streamers (for example, from following “Fang Qi Kiki” to following “Planet Research Institute”). Participants’ ages ranged from 20 to 47 years, and their professional backgrounds included students, teachers, and employees of private enterprises. Approximately 87.5% of the respondents held a university degree or higher. Regarding gender, 41.6% were male and 58.4% were female. Thus, the sample represents gender, age, and professional background (see Table 2).

Table 2 Basic information of participants (N = 24).
Full size table

However, snowball sampling may lead to insufficient representativeness of the sample in specific characteristics, as it may be biased toward participants with more closely connected social networks, thus affecting the diversity of the sample52. To address this issue, the researchers focused on sample heterogeneity during the initial design phase by engaging with multiple fan groups of different streamers and conducting extensive online social interactions to encompass a more diverse range of viewers. Additionally, theoretical saturation was strictly followed during data collection, meaning data collection ceased when no new insights were provided51,52, to maximize sample diversity and the credibility of the research conclusions.

The interview guide was developed based on extensive live-streaming observation and literature review and was refined after initial interviews with three experienced participants. The study was conducted from March 2 to April 28, 2023, and involved 24 participants. The interviews primarily focused on recalling viewing experiences, evaluating the attractiveness of live-streaming and streamers, and assessing live-streaming quality standards. The interviews followed a semi-structured format, with questions covering participants’ tourism experiences, live-streaming viewing habits, preferred streamers, viewing durations, reasons for ceasing to watch, and motivations for switching to other streamers.

After the interviews, all recordings were transcribed, and a portion of the data was randomly allocated for coding (2/3) and saturation checks (1/3) to ensure comprehensive and in-depth data analysis. The average duration of each interview was 40 min, and approximately 180,000 words of interview text were generated.

Text analysis

Open coding

This study employed open coding to initially analyze interview data, meticulously examining responses from 2/3 of participants. Through line-by-line analysis and labeling, conceptual categories were identified, and similar elements merged into broader themes. After filtering out rare and contradictory concepts, the study culminated in 35 distinct concepts and 21 coherent categories (Table 3).

Table 3 Examples of initial concepts and categories formed by open coding.
Full size table

Axial coding

Axial coding involves cluster analysis, exploring the relationships between each category and establishing connections among different categories. As the relationships become clear, subcategories and main categories are formed, as shown in Table 4.

Table 4 Subcategories and corresponding main categories are formed by axial coding.
Full size table

Through logical analysis, this study identified nine subcategories: professionalism, credibility, charisma, visual pleasure, auditory pleasure, emotional resonance, timeliness, freshness, and diversity, and three overarching categories: competence of tourism live streamers, enjoyment of tourism viewing, and attractiveness of tourism topics.

Selective coding

Selective coding was applied to distill a central category that encapsulates the primary themes, culminating in the refinement of “alternative attractiveness” as the core concept, supported by three key dimensions: competence of tourism live streamers, enjoyment of tourism viewing, and attractiveness of tourism topics.

The competence of tourism live streamers is defined by live streamers’ professional qualities, such as knowledge, integrity, and humor, which captivate viewers and build trust. In an increasingly standardized industry, these attributes secure more excellent viewer favor. Live streamers like Dong Yuhui, lauded for “knowledge-based sales” and recognized as a “cultural live streamer,” have attracted a significant following by providing informative and culturally rich content, as opposed to those who rely on empty sensationalism.

Enjoyment of tourism viewing refers to the satisfaction and pleasure viewers experience from live streaming, which provides more than information and shopping. It offers a comforting “cloud tourism” experience. The interactive elements of these streams, engaging viewers visually, audibly, and emotionally, can immerse viewers and enhance their willingness to continue following. However, poor-quality streamers can prompt viewers to seek more enjoyable alternatives, as seen in cases where viewers have switched due to unsatisfactory video quality.

The attractiveness of tourism topics reflects the relevance of streaming content to viewers’ interests. Content that consistently engages viewers encourages long-term following, while repetitive or monotonous content can lead to disinterest and a quest for diverse and trending streams. Viewers may tire of streams featuring similar attractions and seek a variety of destinations to sustain their interest.

Theoretical saturation check

This study, through the coding analysis of the remaining 1/3 of the interview materials, confirmed that the three attribute dimensions of alternative attractiveness in TLS (competence of tourism live streamers, enjoyment of tourism viewing, and attractiveness of tourism topics) have reached theoretical saturation, without revealing new categories or concepts.

Study II: quantitative analysis of the causes of NCFI in TLS

Research hypotheses

Viewers’ NCFI in TLS mainly manifests as viewers reducing their watching time, unfollowing, or switching to other live-streaming channels. The PPM theory consolidates factors into push, pull, and mooring effects, influencing the viewers’ decision to follow. This study, grounded in PPM theory and informed by in-depth interviews, has delineated the dimensions of alternative attractiveness and developed a model encompassing push (psychological contract breaches), pull (alternative attractiveness), and mooring (viewer-live streamer social distance) factors, as depicted in Fig. 1.

Fig. 1
figure 1

Research model of causes of NCFI in TLS based on PPM theory.

Full size image

Hypotheses related to push factors

  1. (1)

    The impact of psychological contract breach on NCFI

    The success of TLS hinges on sustained viewer engagement, which is governed by an unwritten psychological contract encompassing transactional and relational aspects. According to social exchange theory, unmet expectations can lead to the termination of relationships and provoke negative responses, altering cognition, emotions, attitudes, and behavior and potentially leading to disloyalty and adverse actions40,53. Thus, when viewers perceive a psychological contract breach in TLS, it may erode trust and incite negative sentiments, increasing the likelihood of discontinuing their followership. Therefore, the following hypotheses are proposed:

H1a

Transactional psychological contract breach is positively related to viewers’ NCFI in TLS.

H1b

Relational psychological contract breach is positively related to viewers’ NCFI in TLS.

  1. (2)

    The impact of psychological contract breach on customer complaining behavior

    Research on customer-enterprise relationships has increasingly focused on the behavioral consequences of psychological contract breaches. Bavik et al.54 propose that a psychological contract breach of perceived service quality can significantly impact tourists’ satisfaction and word-of-mouth. Bavik et al.55 note a positive correlation between such breaches and various customer complaints, with increased severity of the breach escalating the propensity for complaining behavior. These findings imply that unmet expectations in live streaming can provoke viewer complaints. Consequently, the study proposes the following research hypotheses:

H2a

Transactional psychological contract breach is positively related to customer complaining behavior.

H2b

Relational psychological contract breach is positively related to customer complaining behavior.

  1. (3)

    The impact of customer complaining behavior on NCFI.

    Enterprises risk incurring customer dissatisfaction through their products or services, potentially sparking a range of complaints as customers seek redress or express their grievances. Research indicates that such complaining behavior is linked to a customer’s likelihood to switch providers and often precedes their disengagement2,56. Huang and Ma36 discovered that direct customer complaints can enhance repurchase intentions, whereas private complaints tend to diminish them. Disgruntled viewers in the online TLS market can swiftly identify alternatives and divert their focus14, suggesting that complaining behavior could foster an intention to discontinue following. Consequently, the study proposes the research hypothesis:

H3

Customer complaining behavior is positively related to viewers’ NCFI in TLS.

  1. (4)

    The mediating effect of customer complaining behavior on the relationship between psychological contract breach and viewers’ NCFI in TLS

    In TLS, viewers exhibit psychological contract breaches primarily in transactional and relational facets. Service failures can prompt viewers to feel betrayed, leading to dissatisfaction and disappointment, and they often articulate this through complaints, potentially impacting their future engagement57. Research reveals that customers frequently opt for complaint behavior rather than silence post-breach, influencing their withdrawal intentions40. With these insights, the study proposes the following research hypotheses:

H4a

Customer complaining behavior mediates the relationship between transactional psychological contract breach and viewers’ NCFI in TLS.

H4b

Customer complaining behavior mediates the relationship between relational psychological contract breach and viewers’ NCFI in TLS.

  1. (5)

    The moderating effect of perceived controllability on the relationship between psychological contract breach and customer complaining behavior.

    In TLS, the relationship between viewers and live streamers transcends the traditional buyer–seller dynamic, focusing on shared experiences and community enhancement20. When live streaming services encounter problems, viewers may perceive a psychological contract breach. However, due to their emotional and time investment, viewers often seek reasons to continue their engagement. A key element in this process is perceived controllability during a crisis, an essential component of attribution theory58.

Weiner’s attribution theory suggests that viewers, as rational information processors, are influenced by causal inferences59. If viewers believe an issue is within the operator’s control, it may trigger negative emotions and defensive actions, such as reduced trust or purchase intent60. In TLS, viewers assess the live streamer’s responsibility and accountability in the event of a breach. If they view the live streamer as responsible and capable but inactive, this can provoke negative emotions and lead to complaints37,61. Thus, increased controllability may exacerbate viewer complaints. Based on these insights, the study proposes the following research hypotheses:

H5a

The relationship between transactional psychological contract breach and customer complaining behavior is positively moderated by perceived controllability.

H5b

The relationship between relational psychological contract breach and customer complaining behavior is positively moderated by perceived controllability.

Hypotheses related to a mooring factor

Customer relationship management research indicates that strong brand or enterprise relationships can mitigate the negative effects of product failures62. Social distance, measuring interaction, communication, and closeness, influences individual interpretations and perceptions of others63. Liberman et al.34 propose that social distance levels affect the depth of information processing and psychological responses, including attitudes and moral judgments. A smaller social distance encourages consideration of context, relaxed moral criteria, and increased tolerance for unethical actions64, while a more significant social distance emphasizes broader moral principles and higher moral expectations65. Therefore, closer social bonds in TLS may enhance viewer tolerance and loyalty. Based on this, the study proposes the following research hypothesis:

H6

Viewer-live streamer social distance is positively related to viewers’ NCFI in TLS.

Hypotheses related to a pull factor

The research utilized Nvivo.11 to dissect interview data, pinpointing three critical dimensions influencing viewers’ intentions to stop following TLS: competence of tourism live streamers, enjoyment of tourism viewing, and attractiveness of tourism topics. Studies suggest that viewers are inclined to stay with current relationships when satisfied and confronted with unappealing alternatives66. However, the emergence of formidable rivals can undermine viewer loyalty67. In social networking platforms35 and social media18, the allure of alternatives significantly impacts viewers’ inclination to switch, diminishing their intent to persist with existing platforms. Xie et al.68 observed a similar trend with tourism booking sites, where alternative attractiveness decreased viewers’ continued engagement. Consequently, the study proposes the following research hypothesis:

H7

Alternative attractiveness is positively related to viewers’ NCFI in TLS.

Research sample

This study recruited TLS enthusiasts through private messages or comments on platforms such as Douyin, Ctrip, and Mafengwo and identified relevant bloggers using keywords on Xiaohongshu and Weibo. A snowball sampling method was applied through social network platforms to select the sample. The survey was conducted from May 15 to August 13, 2023, with 437 questionnaires distributed and 391 valid responses received, resulting in an effective response rate of 89.47%. According to the sample’s descriptive statistics, the majority of respondents were aged between 20 and 29 years, accounting for 59.60%, and 54.73% of the respondents were female.

Regarding occupation, students and employed individuals comprised 38.36% and 58.83%, respectively. The educational background of the respondents was relatively high, with 69.05% holding a bachelor’s degree or higher. Regarding the number of tourism bloggers, most respondents followed three or more bloggers, accounting for 61.64%, indicating a potential for NCFI. As for the choice of tourism viewing platform, Douyin was the most popular, with about 46.29% of respondents preferring this platform. The sample population was relatively young, highly educated, and more inclined to watch TLS content on Douyin. Table 5 presents the basic characteristics of the sample.

Table 5 Sample descriptive statistics.
Full size table

Variable measurement

The study adapted empirical research from related fields to create measurement items specific to TLS to ensure the questionnaire’s reliability and validity. The questionnaire is divided into seven sections: psychological contract breach, viewer-live streamer social distance, alternative attractiveness, customer complaining behavior, perceived controllability, NCFI, and demographics. It includes eight items for psychological contract breach, divided into transactional and relational sub-dimensions69; four items for viewer-live streamer social distance57,70; nine items for alternative attractiveness66,71; three items for customer complaining behavior36,56; three items for NCFI72; and three items for perceived controllability73. All latent variables are measured using a 5-point Likert scale.

A pre-survey distributed on TikTok and Ctrip from May 3–10, 2023, gathered 137 responses to assess the questionnaire’s design, accuracy, and the reliability and validity of its constructs. The Cronbach’s coefficient for the variables was above 0.7, and item factor loadings exceeded 0.7, indicating strong reliability and validity. The final questionnaire, refined based on pre-survey results, was distributed through the same platforms.

Common method bias test

This study applied Harman’s single-factor test using exploratory factor analysis in Spss 26.0 for all items to mitigate common method bias. The results identified eight factors with eigenvalues over 1, accounting for 69.772% of the total variance. The primary unrotated factor accounted for 34.866% of the variance, falling below the 50% threshold, suggesting that common method bias was not a significant concern in this analysis.

Reliability and validity test

The study utilized Spss 26.0 and Amos 23.0 for the scale’s reliability and validity assessment. All Cronbach’s coefficients exceeded 0.7, demonstrating strong internal consistency and reliability. Confirmatory factor analysis results indicated an x2/df ratio of 1.334, RMSEA of 0.029, GFI of 0.923, NFI of 0.918, IFI of 0.978, and CFI of 0.978, all of which are within accepted fit criteria, confirming a strong model-data alignment. Variable composite reliability (CR) values were above 0.7, and factor loadings were above 0.70, verifying the scale’s reliability. The average variance extracted (AVE) for all scales was over 0.5, signifying excellent convergent validity (see Table 6). Additionally, the square root of AVE surpassed the inter-variable Pearson correlation coefficients, demonstrating the scale’s strong discriminant validity (refer to Table 7).

Table 6 Results of convergent validity test for all variables.
Full size table
Table 7 Correlation analysis and discriminant validity for all variables.
Full size table

Path analysis and hypothesis testing

This study assessed the structural equation model using Amos 26.0 and maximum likelihood estimation. It showed an excellent fit (x2/df = 1.587, RMSEA = 0.039, GFI = 0.938, CFI = 0.974, RFI = 0.921, NFI = 0.934, IFI = 0.974, TLI = 0.969), suitable for path analysis.

In the main effects outlined in Table 8, TPCB is positively related to viewers’ NCFI in TLS (β = 0.158, p < 0.05), supporting H1a. RPCB exhibits a stronger positive relationship with viewers’ NCFI in TLS (β = 0.245, p < 0.01), supporting H1b. Furthermore, TPCB significantly increases CCB (β = 0.284, p < 0.01), supporting H2a. The effect of RPCB on CCB is even more significant (β = 0.333, p < 0.01), supporting H2b. CCB is also positively related to viewers’ NCFI in TLS (β = 0.141, p < 0.05), supporting H3. Additionally, VLSD positively relates to viewers’ NCFI in TLS (β = 0.213, p < 0.01), supporting H6. AA significantly positively affects viewers’ NCFI in TLS (β = 0.414, p < 0.01), supporting H7.

Table 8 Parameter estimates of structural equation modeling execution results.
Full size table

The mediating effect was tested using Hayes’s74 bootstrapping method, with the number of bootstrap resamples set to 2,000 and the confidence level set to 95%. The results are presented in Table 9 and indicate the following: (1) The mediating effect of CCB on the relationship between TPCB and viewers’ NCFI in TLS is significant (β = 0.040, [0.003, 0.110]). (2) The mediating effect of CCB on the relationship between RPCB and viewers’ NCFI in TLS is also significant (β = 0.047, [0.005, 0.119]). H4a and H4b are supported.

Table 9 The mediating role of CCB in the relationship between PCB and viewers’ NCFI in TLS.
Full size table

PC positively moderates the relationships between TPCB and CCB, as well as between RPCB and CCB. The study found that the impact of TPCB on CCB was weaker under conditions of low PC (simple slope = 0.281, p < 0.01), and the effect intensified under conditions of high PC (simple slope = 0.453, p < 0.01), supporting H5a. Similarly, the positive influence of RPCB was also weaker at low levels of PC (simple slope = 0.213, p < 0.05) and significantly more substantial at high levels of PC (simple slope = 0.467, p < 0.01), supporting H5b. For more details, see Table 10. We further mapped the moderating effects. Figures 2 and 3 show that when PC is high, as the number of TPCB or RPCB increases, CCB will be further strengthened.

Table 10 The moderating role of PC in the relationship between PCB and CCB.
Full size table
Fig. 2
figure 2

Moderating effect of PC on the relationship between TPCB and CCB.

Full size image
Fig. 3
figure 3

Moderating effect of PC on the relationship between RPCB and CCB.

Full size image

Discussion

Research conclusions

In the digital age, TLS has become a key medium connecting tourist destinations with viewers, significantly influencing the business model innovation and marketing strategies of tourism enterprises and destinations. TLS’s success relies on the viewer’s continuous engagement; achieving growth in traffic and promotional goals becomes challenging without sufficient engagement15.

Based on the PPM theory, this study employs a mixed research approach to explore the reasons behind NCFI in TLS comprehensively. In Study I, a measurement scale for alternative attractiveness was developed, clarifying the scope and components of this variable and providing a theoretical foundation for subsequent research. Study II further investigates how psychological contract breach (which contains two sub-dimensions of transactional and relational), as push factors, significantly influences viewers’ NCFI in TLS, with RPCB having a powerful impact. The path analysis results indicate that psychological contract breach trigger negative emotions and a loss of trust in the viewer, thereby increasing their NCFI in TLS. Specifically, a relational psychological contract breach is more likely to provoke emotional responses and trust divergence from the viewers, making their impact more significant than a transactional psychological contract breach.

Moreover, customer complaining behavior is important to mediate between psychological contract breach and viewers’ NCFI in TLS. The path analysis shows that psychological contract breach directly triggers customer complaining behavior, further exacerbating the viewers’ NCFI in TLS. This mechanism reveals how negative interactions within TLS can amplify viewer disengagement through complaining emotions. Further analysis indicates that perceived controllability significantly moderates the relationship between psychological contract breach and customer complaining behavior. When the viewer perceives the live streaming content or services as more controllable8,75, they are more likely to adopt positive coping strategies and reduce complaining behavior; conversely, lower perceived controllability increases customer complaining behavior.

In addition to a psychological contract breach, alternative attractiveness, and viewer-live streamer social distance, as pull and mooring factors, also influence viewers’ NCFI in TLS. As an external distractor, alternative attractiveness diverts the viewer’s attention and reduces their focus on the live-streaming content. At the same time, when viewer-live streamer social distance is significant, the viewer’s emotional investment in the live-streaming content is lower, thus increasing the likelihood of viewers’ NCFI in TLS2,64.

Theoretical contributions

The main contributions of this study are reflected in the following three aspects:

First, the study constructs and extends the PPM theoretical framework for viewers’ NCFI in TLS, clarifying its underlying mechanisms and boundary conditions. Based on the PPM theory, this study proposes and builds a novel framework to explain the formation mechanisms of viewers’ NCFI in TLS. Most existing studies focus on viewer engagement or following behavior in TLS2,13, overlooking the reverse factors that lead to viewer disengagement. By integrating psychological contract breach and the dynamic relationship between viewers and live streamers, this study fills this research gap and advances the application of the PPM theory in NCFI research33. Specifically, this study investigates how push factors (transactional and relational psychological contract breach), a pull factor (alternative attractiveness), and a mooring factor (viewer-live streamer social distance) work together to influence viewers’ NCFI in TLS. In particular, the mediating role of a customer complaining behavior reveals how a psychological contract breach triggers negative emotions and trust erosion36, further amplifying viewers’ NCFI in TLS. Moreover, the study finds that perceived controllability significantly moderates the relationship between psychological contract breach and customer complaining behavior. When viewers perceive they have control over the live streaming content or service, they are more likely to adopt positive coping strategies and reduce customer complaining behavior. Conversely, lower perceived controllability exacerbates customer complaining behavior and viewers’ NCFI in TLS. This study expands the application of the PPM theory and provides a systematic theoretical framework for understanding NCFI in TLS, deepening our understanding of viewer behavior.

Second, this study refines and deepens the dimensions of alternative attractiveness, advancing the development of related theories. Although alternative attractiveness is identified as a key factor influencing viewers’ continuous attention, previous research has only briefly explored this concept, with unclear definitions and a lack of a multidimensional perspective68. Through in-depth interviews, this study identifies and refines three key dimensions of alternative attractiveness in TLS, addressing gaps in the existing literature on this concept. This refinement provides a clear theoretical framework for subsequent quantitative research and offers a new perspective on how alternative attractiveness influences viewers’ attention and engagement5,13. Unlike traditional studies that mainly focus on advertising content or creative appeal, this study analyzes alternative attractiveness from a multidimensional perspective, advancing the understanding of viewer behavior in information technology and digital marketing76 and providing theoretical support for optimizing live streaming content and improving viewer retention.

Third, this study innovatively shifts to the NCFI perspective, enriching the reverse thinking in TLS research. Most existing literature studies viewer “following” behavior, especially how live streaming content attracts viewers to maintain continuous attention15,16. By employing a reverse thinking approach, this study breaks from the traditional “following” perspective and shifts attention to the exploration of NCFI, i.e., why viewers interrupt their attention to TLS. This innovative perspective provides new research directions for the TLS and digital marketing fields5, filling the research gap in the literature on “churn” or “non-continuous following” behavior. The study focuses on the formation mechanisms of NCFI, revealing how psychological contract breach, viewer-live streamer social distance, and alternative attractiveness drive viewer attrition, offering significant theoretical and practical value. In particular, existing literature has paid limited attention to viewers’ emotional and trust dynamics during live streaming2,16. At the same time, this study deepens the theoretical exploration of this area by focusing on the negative effects of a psychological contract breach.

Management implications

Practical implications for TLS enterprises

TLS enterprises should prioritize institutional and resource allocation strategies to enhance viewer retention and loyalty. Firstly, trust management and contract fulfillment mechanisms should be improved. TLS enterprises should establish institutionalized management processes to ensure viewer trust in the platform. This includes developing standardized trust management systems, such as transparent communication mechanisms and multi-tiered feedback channels, to promptly address viewer complaints and concerns, thereby minimizing the negative emotions caused by psychological contract breaches36. Furthermore, integrating contract fulfillment into the performance evaluation system for streamers and clear behavioral guidelines and standards can encourage streamers to honor their commitments during TLS, avoiding trust erosion and viewer disloyalty due to unfulfilled promises.

Secondly, enhances the perceived controllability of content. TLS enterprises should leverage technology and resources to improve viewer engagement and their sense of control over live-streaming content. For instance, they can develop highly interactive and customizable live-streaming templates and provide streamers with intelligent tools and resources to enrich the viewer’s sense of control. Specific measures include introducing AI-powered interactive technologies, real-time data analytics tools, and personalized content recommendation modules to help streamers dynamically adjust live-streaming content based on viewer preferences, thereby significantly boosting viewer satisfaction and engagement.

Lastly, strengthens the emotional connection between streamers and viewers. Research shows that when the viewer-live streamer social distance is considerable, viewers’ emotional involvement tends to be lower64, increasing the likelihood of viewers’ NCFI in TLS. Therefore, TLS enterprises should implement systematic incentive measures and operational models to help streamers establish stronger emotional bonds with their viewers. On the one hand, TLS enterprises can set up incentive mechanisms to encourage streamers to increase the frequency and depth of interactions during TLS, such as responding to viewer questions in real-time or engaging in comment exchanges, thereby narrowing the psychological distance between streamers and viewers. On the other hand, TLS enterprises can regularly organize online and offline interactive activities, such as fan meetups or themed community events, to foster deeper emotional connections between streamers and their viewers, enhancing viewer loyalty and a sense of belonging.

Practical implications for TLS streamers

As content creators who directly engage with viewers, TLS streamers should optimize their performance at the operational and behavioral levels to enhance viewer engagement and emotional investment.

Firstly, emotional interaction should be strengthened to reduce viewer-live streamer social distance. Streamers should actively engage with viewers by interacting with them in real time and promptly responding to comments and questions to foster participation. Sharing personal stories or expressing emotional resonance can further build trust and a sense of closeness with the viewer. This approach effectively reduces social distance and deepens emotional connections.

Secondly, maintain transparency and credibility of content. Streamers should avoid false advertising or misleading viewers, ensuring the authenticity and consistency of their content. For example, they can disclose detailed information about live-streamed products or destinations in advance, avoiding exaggerated claims to build viewer trust. Regularly updating live-streaming content to ensure accuracy and freshness enhances viewer satisfaction and loyalty.

Lastly, innovative content formats increase appeal. Streamers must continuously adapt live-streaming formats and enrich content based on viewer preferences to sustain their interest8. For instance, integrating virtual reality technology can create immersive tourism experiences while incorporating cultural stories or historical backgrounds that add depth to the content. Additionally, interactive features such as live polls or other viewer-driven activities can increase engagement and focus, reducing the likelihood of attention diversion due to competing attractions. By implementing these strategies, TLS streamers can effectively foster deeper connections with their viewer and sustain their attention.

This study offers TLS enterprises and streamers practical insights on addressing viewers’ NCFI in TLS. TLS enterprises should establish robust trust management and contract fulfillment mechanisms, enhance viewer perceived controllability, and deepen the emotional connection between streamers and viewers. Streamers should emphasize strengthening emotional interaction, ensuring the transparency and credibility of live-streaming content, and continuously innovating content formats to boost viewer engagement and emotional investment. Implementing these strategies allows the TLS industry to foster long-term viewer attention, enhance user retention, and promote sustainable development. These insights provide valuable management experience for existing enterprises and offer guidance for the growth of emerging platforms and creators.

Limitations and future research

This study explores the impact of psychological contract breach, viewer-live streamer social distance, and alternative attractiveness on viewers’ NCFI in TLS. While it provides important insights into the TLS field, several limitations remain.

First, this study relies on online data collection, which may not fully capture the subtle differences in viewer behavior and emotional responses. Due to the risks of memory bias and self-reporting bias in online surveys, participants’ responses may not entirely reflect their genuine emotions and behavioral reactions. To improve the reliability and validity of future research, a combination of experimental designs and longitudinal tracking data could be employed53. Multiple data collection methods, such as real-time behavioral tracking and sentiment analysis, could further refine our understanding of dynamic viewer responses, yielding more accurate and in-depth findings.

Additionally, this study did not fully consider the impact of cultural differences on viewer behavior. As the research sample primarily focuses on specific groups or regions, future studies should consider how cultural backgrounds influence perceptions and reactions to psychological contract breach, viewer-live streamer social distance, and alternative attractiveness. Viewers’ expectations, emotional involvement, and NCFI toward TLS content may differ in cultural contexts77. For example, in individualistic cultures, viewers may emphasize the streamer’s personal charm, personality, and interactive style, tending to establish a connection with the streamer and valuing the streamer’s style and engagement. In contrast, in collectivist cultures, viewers may focus more on the social value and consistency of the content, prioritizing its authenticity and reliability over the streamer’s appeal. Therefore, future research could further explore how cultural differences affect the interaction between these key factors and provide more accurate theoretical support and practical guidance for TLS platforms in a globalized context.

Finally, this study does not distinguish between different types of TLS. However, various types (natural landscape live streaming and cultural history live streaming) may influence viewers’ emotional investment and NCFI through their unique content styles and presentation formats. For example, natural landscape live streaming might attract viewers through stunning visual experiences and immersive commentary. In contrast, cultural history live streaming may enhance viewers’ interest through in-depth cultural explanations or engaging storytelling. These live streaming types may vary viewer-live streamer social distances and alternative attractiveness, leading to varied impacts on viewer behavior. Future research should further explore the differences between types of TLS, analyzing their effects on viewer interaction, emotional investment, and behavioral responses. This will enrich existing theoretical frameworks and provide more targeted guidance and recommendations for the TLS industry.

Conclusion

Based on the PPM theory, this study explores the formation mechanism of viewers’ NCFI in TLS. The findings reveal that factors such as psychological contract breach, viewer-live streamer social distance, and alternative attractiveness significantly affect viewers’ NCFI in TLS, with relational psychological contract breach having the most prominent impact. Customer complaining behavior mediates the relationship between psychological contract breach and viewers’ NCFI in TLS, while perceived controllability positively moderates the relationship between psychological contract breach and customer complaining behavior. The contributions of this study are threefold: First, it constructs and expands the NCFI model in TLS, enriching the application of PPM theory; second, it refines the dimensions of alternative attractiveness, providing a clear framework for future research; and third, it innovatively examines viewer behavior from the perspective of NCFI, filling the research gap on viewer attrition. At the same time, this study provides practical recommendations for TLS enterprises and streamers, including strengthening trust management, enhancing content controllability, and deepening emotional connections between streamers and viewers, thereby effectively improving viewer engagement and loyalty and promoting the sustainable development of the TLS industry. However, this study has certain limitations regarding sample size and data collection. Future research could combine various data methods to explore the impact of cultural differences and different types of TLS on NCFI, further improving the theoretical framework and providing more comprehensive guidance for the TLS industry.