Chunyan Zhang1,*
1SEGI University,Postal Code 250000
Email: 576707728@qq.com
Abstract
With the rapid growth and pervasive influence of social media platforms in China, consumer behavior has undergone significant transformation, particularly in sectors such as organic food, where public perception and purchase decisions are increasingly shaped by online discussions. Platforms such as Weibo, Douyin, WeChat, and Xiaohongshu have become critical spaces for individuals to share personal experiences, reviews, and opinions, which in turn influence broader consumer attitudes towards organic food brands.Grounded in the Stimulus–Organism–Response (S–O–R) framework and the Elaboration Likelihood Model (ELM), this study conceptualizes social media language features—emotional valence and functional keyword frequency—as stimuli, brand perception and social identity as organismic states, and purchase intention as the response. Using a mixed-method design, the study combines text mining and sentiment analysis of social media posts with a survey of 500 organic food consumers in China. Text preprocessing employs Chinese segmentation, stopword removal, and sentiment dictionaries (BosonNLP, HowNet, NTUSD), while the survey uses validated scales for brand perception, purchase intention, social identity, and information credibility. Structural equation modeling (SEM) is applied to test direct, mediating, and moderating effects.The results show that positive emotional language and high frequency of health- and sustainability-related keywords significantly enhance brand perception, whereas negative emotional language weakens it. Brand perception mediates the relationship between social media language and purchase intention. Social identity strengthens the impact of social media language on brand perception, and information credibility moderates the effect of keyword frequency. The findings extend S–O–R and ELM into the digital organic food context, highlight the importance of Chinese cultural social identity in online consumption, and provide actionable guidance for content strategy, targeting, and credibility management in social media marketing.
Keywords: social media language; S–O–R model; Elaboration Likelihood Model; organic food; brand perception; social identity; purchase intention
1. Introduction
1.1 Background
In recent years, China’s organic food market has experienced remarkable growth, primarily driven by a shift in consumer preferences toward healthier and more sustainable lifestyles. As Chinese consumers become increasingly aware of the environmental and health impacts of their food choices, organic products are gaining popularity as a preferred alternative to conventional, mass-produced food. Organic products are often viewed as cleaner, more nutritious, and less harmful to the environment, which has led to a surge in their market share.
However, the relationship between consumers and organic food brands is not shaped solely by the inherent qualities of the products—such as nutritional value, safety, and environmental benefits. Consumer perceptions and purchase intentions are influenced by a complex array of external factors, including media portrayals, brand messaging, and, importantly, social media interactions (Kaplan & Haenlein, 2010). Social media platforms like Weibo, Douyin, WeChat, and Xiaohongshu are now integral to contemporary Chinese consumer culture, serving as central hubs for information dissemination, social interaction, and public discourse.
On these platforms, users frequently share their experiences, opinions, and reviews of products, including organic food. This dynamic environment exposes consumers to both brand-generated content and peer-generated content, which often carries significant weight in shaping consumer perceptions (Cheung & Thadani, 2012). Through hashtags, comments, likes, shares, and interactions with influencers and brands, consumers form impressions that may directly influence their decision to purchase. As a result, the language used in social media discussions—both by brands and consumers—plays a critical role in shaping brand perception and purchase intention.
1.2 Research Questions and Objectives
This study investigates the critical role that social media language plays in shaping consumers’ perceptions and purchasing behavior toward organic food brands in China. Specifically, it addresses three core research questions:
Which features of social media language (e.g., emotional valence, interaction frequency, and functional keywords) have the most significant impact on consumers’ brand perception of organic food?
How does social media language influence purchase intention via brand perception as a mediating mechanism?
How do cultural and social identity factors within the Chinese context shape the relationship between social media language, brand perception, and purchase intention, and how does this differ from patterns reported in other markets?
To answer these questions, the study:
Develops a theoretical model grounded in the Stimulus–Organism–Response (S–O–R) framework and the Elaboration Likelihood Model (ELM);
Operationalizes social media language features using text mining and sentiment analysis;
Measures brand perception, social identity, information credibility, and purchase intention using validated scales;
Employs structural equation modeling (SEM) to test direct, mediating, and moderating relationships.
The overall objective is to provide a comprehensive, theory-driven understanding of how social media language influences brand perception and purchase intention in the Chinese organic food market and to generate actionable implications for social media marketing strategies.
1.3 Structure of the Paper
The remainder of the paper is structured as follows. Section 2 presents a literature review on social media and consumer behavior, organic food consumer perception, and the role of language in brand perception and purchase intention. Section 3 introduces the theoretical framework, drawing on the S–O–R model and ELM. Section 4 develops the research hypotheses. Section 5 details the methodology, including data sources, text mining procedures, measurement scales, and data analysis techniques. Section 6 reports the SEM-based results, including reliability and validity, model fit, and tests of mediation and moderation. Section 7 discusses theoretical and practical implications, as well as limitations and directions for future research. Section 8 concludes the paper, and an Appendix provides a textual description of the SEM model diagram.
2. Literature Review
2.1 Social Media and Consumer Behavior
Over the past decade, social media has evolved into a powerful force that significantly shapes consumer behavior. It offers consumers a unique platform for sharing opinions, discussing personal experiences, and discovering products, facilitated by both formal advertising and peer-to-peer interactions (Kaplan & Haenlein, 2010). Unlike traditional one-way mass communication, social media enables interactive, many-to-many communication, where consumers can engage directly with brands and other users.
In the context of organic food, consumer behavior is driven by a combination of factors, including health consciousness, environmental concern, and the desire for transparency. Social media platforms provide access to product reviews, ethical assessments of companies, and personal experiences of other consumers. Exposure to peer-generated content can significantly influence consumer attitudes toward organic food (Cheung & Thadani, 2012; Hutter et al., 2013). Electronic word-of-mouth (eWOM) in social media has been shown to affect brand awareness, perceived credibility, and purchase intention (Chu & Kim, 2011; Erkan & Evans, 2016).
2.2 Organic Food Consumer Perception
Consumer perceptions of organic food are shaped by health, environmental, ethical, and quality considerations. Organic products are often associated with superior health benefits, environmental sustainability, and better taste or quality, making them particularly attractive to health-conscious and environmentally aware consumers (Hughner et al., 2007; Vojnovic et al., 2014). At the same time, the price premium of organic products can act as a barrier (Padel & Foster, 2005; Rana & Paul, 2017), and trust in certifications, labels, and brands becomes critical (Chen, 2007; Prakash et al., 2023).
In China, the organic food market is still evolving, with consumer awareness and perceptions becoming more sophisticated (Thøgersen et al., 2016; Zhou, 2013). Trust and perceived authenticity are central. Consumers look for credible signals related to product origin, safety, and environmental claims (Smith & Paladino, 2010). Social media plays a particularly important role in this process, allowing consumers to gain information, verify claims, and observe others’ experiences.
2.3 Social Media Language, Brand Perception, and Purchase Intention
Language is a primary vehicle through which brands communicate with consumers and shape their perceptions. The words, tone, and emotional content of brand communications can evoke affective responses that influence brand attitudes and behaviors (Kotler & Keller, 2016). On social media, this effect is amplified because language is embedded in interactive formats, including comments, replies, mentions, and sharing behavior.
Positive language—emphasizing health, safety, environmental responsibility, and quality—tends to enhance brand image, foster trust, and strengthen brand attitudes. In contrast, negative language—highlighting dissatisfaction, risk, or distrust—can damage brand reputation (Berger & Milkman, 2012; Stieglitz & Dang-Xuan, 2013). Emotionally charged language, whether positive or negative, is more likely to attract attention and be shared, thereby amplifying its impact.
Brand perception is a multidimensional construct that includes brand image, brand trust, and brand attitude (Keller, 2003). It plays a central role in linking marketing communications to behavioral outcomes. Prior studies show that positive brand perception enhances purchase intention, while negative perceptions reduce it (Hutter et al., 2013; Rana & Paul, 2017). In the organic food context, perceptions related to health, environmental responsibility, and authenticity are particularly influential (Chen, 2007; Hughner et al., 2007).
2.4 Social Identity, Cultural Context, and Online Influence
Social identity theory suggests that individuals derive part of their self-concept from membership in social groups and seek to maintain a positive social identity (Tajfel & Turner, 1979). In the Chinese context, collectivist values and sensitivity to social norms heighten the influence of social proof and group-based cues in consumption (de Mooij & Hofstede, 2011; Cialdini, 2007).
On social media, social identity manifests through group memberships (e.g., health-conscious communities), shared values, and community-oriented language. Consumers with strong social identity related to organic or sustainable lifestyles may be more responsive to congruent social media language and peer endorsements, leading to stronger effects on brand perception and purchase intention. Understanding this moderating role is particularly important in China, where conformity to group norms and reliance on peer validation are prominent in consumer decision-making.
3. Theoretical Framework
3.1 Stimulus–Organism–Response (S–O–R) Model
The S–O–R model, originally proposed in environmental psychology, posits that external stimuli (S) influence internal organismic states (O), which in turn lead to behavioral responses (R) (Mehrabian & Russell, 1974). In this study:
Stimulus (S): Social media language characteristics, including emotional valence (positive vs. negative emotional content) and functional keyword frequency (e.g., “health,” “eco-friendly,” “organic,” “sustainable”).
Organism (O): Internal psychological states of the consumer, represented by brand perception (brand image, trust, and attitude) and social identity within the Chinese cultural context.
Response (R): Purchase intention toward organic food brands.
By treating social media language as the stimulus, brand perception and social identity as organismic states, and purchase intention as the response, this study connects text analytics with consumer psychology in a unified framework.
3.2 Elaboration Likelihood Model (ELM)
The Elaboration Likelihood Model (ELM) explains how individuals process persuasive information through either a central or peripheral route (Petty & Cacioppo, 1986).
Central Route: Involves deliberate, thoughtful processing of message content and arguments. In this study, information credibility (e.g., perceived accuracy, expertise, and trustworthiness of social media content) reflects the central route.
Peripheral Route: Involves reliance on heuristic cues such as emotion, source attractiveness, or popularity. Here, emotional tone and interaction frequency (likes, comments, shares) capture peripheral cues.
Combining S–O–R and ELM allows the study to distinguish between cognitive and affective pathways through which social media language shapes brand perception and purchase intention.
4. Hypotheses Development
4.1 Emotional Valence and Brand Perception
Emotional language can elicit affective responses that shape consumer attitudes toward brands (Kotler & Keller, 2016; Berger & Milkman, 2012). Positive emotional expressions (e.g., satisfaction, trust, enthusiasm) are likely to enhance perceived brand quality and trust, while negative emotions (e.g., dissatisfaction, skepticism) undermine them.
H1: Positive emotional language in social media has a significant positive effect on brand perception.
H2: Negative emotional language in social media has a significant negative effect on brand perception.
4.2 Functional Keyword Frequency and Brand Perception
Functional keywords related to health, environmental protection, and product safety signal attributes that organic food consumers care about (Hughner et al., 2007; Rana & Paul, 2017). Frequent mention of these keywords strengthens associations between the brand and these desired attributes, thereby enhancing brand perception.
H3: A high frequency of functional keywords (e.g., “health,” “eco-friendly,” “organic”) in social media has a significant positive effect on brand perception.
4.3 Brand Perception as a Mediator
Brand perception is a key organismic state linking external communication to behavioral outcomes (Keller, 2003). Social media language may not affect purchase intention directly but does so through shaping how consumers perceive the brand’s image, trustworthiness, and overall attractiveness.
H4: Brand perception plays a significant mediating role in the relationship between social media language (emotional valence and functional keywords) and purchase intention.
4.4 Social Identity as a Moderator
Consumers with strong social identity related to organic or sustainable lifestyles are more likely to process congruent messages favorably (Tajfel & Turner, 1979; de Mooij & Hofstede, 2011). In collectivist cultures, social identity can amplify the effect of social media language on brand perception.
H5: Social identity significantly moderates the relationship between social media language and brand perception; specifically, the effect of language on brand perception is stronger among consumers with high social identity.
4.5 Information Credibility as a Moderator
According to ELM, credibility is a key central-route cue that shapes how consumers evaluate persuasive messages (Metzger, 2007; Petty & Cacioppo, 1986). When content is perceived as credible, consumers are more likely to integrate functional claims into their brand evaluations.
H6: Information credibility significantly moderates the relationship between functional keyword frequency and brand perception; specifically, the effect of keyword frequency on brand perception is stronger when information credibility is high.
5. Methodology
5.1 Data Sources and Sampling
Data for this study were collected using a mixed-method approach.
Social media data. Posts related to organic food were collected from major Chinese social media platforms, including Weibo, Douyin, WeChat public accounts, and Xiaohongshu. Organic food–related keywords such as “organic,” “green food,” “no additives,” “health,” and “sustainability” were used to identify relevant content. Web scraping tools were employed to collect user-generated posts and comments over a specified time period. After removing duplicates, advertisements without textual content, and posts with fewer than ten characters, more than 1,000 valid posts and comments remained for analysis.
Survey data. To complement the text mining results, a structured survey was administered to 500 consumers who regularly purchase organic food. Participants were recruited from first-tier, second-tier, and lower-tier cities in China, ensuring diversity in regional and demographic backgrounds. The survey captured respondents’ perceptions of organic food brands, social identity, information credibility, and purchase intention, as well as their social media usage patterns.
Table 1. Demographic Information of Survey Respondents
| Demographic Characteristic | Category | Frequency | Percentage (%) |
| Gender | Male | 250 | 50 |
| Female | 250 | 50 | |
| Age | 18–24 years | 120 | 24 |
| 25–34 years | 150 | 30 | |
| 35–44 years | 100 | 20 | |
| 45 years and above | 130 | 26 | |
| Region | First-tier cities | 200 | 40 |
| Second-tier cities | 150 | 30 | |
| Third-tier and below | 150 | 30 | |
| Education Level | Bachelor’s degree or higher | 350 | 70 |
| Below associate degree | 150 | 30 |
5.2 Text Mining and Sentiment Analysis
Text mining and sentiment analysis were applied to the social media corpus to operationalize social media language variables.
(1) Text preprocessing
Segmentation: Chinese posts were segmented using Jieba segmentation to split text into individual words.
Stopword removal: A custom stopword list (including punctuation, numbers, function words, and common Chinese stopwords) was applied to filter out noise.
Part-of-speech (POS) tagging: POS tagging was used, and only nouns, verbs, and adjectives were retained for subsequent analysis, as these carry most semantic and affective information.
(2) Sentiment dictionary selection
A combined sentiment lexicon was constructed using three widely used resources: BosonNLP, HowNet, and NTUSD. BosonNLP and HowNet provided broad coverage of general Chinese sentiment words, while NTUSD was particularly useful for capturing internet slang and informal expressions common on social media. Following Mohammad and Turney (2013) and Pang and Lee (2008), words in the posts were matched to the sentiment lexicon and labeled as positive or negative.
(3) Variable operationalization
Positive Emotion (PosEmo): For each post, the ratio of positive sentiment words to total words was computed.
Negative Emotion (NegEmo): Similarly, the ratio of negative sentiment words to total words was calculated.
Functional Keyword Frequency (KeyFreq): A domain-specific keyword list (e.g., “organic,” “health,” “green food,” “sustainability,” “no additives,” “environmental protection”) was used to count the occurrence of functional terms in each post.
Aggregated indices were created at the post level and then linked to brand-level or consumer-level evaluations as appropriate.
5.3 Measurement Scales
All latent constructs in the survey were measured using validated scales adapted from prior literature. Items were translated into Chinese and back-translated into English to ensure conceptual equivalence. Unless otherwise noted, all items were rated on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree).
Brand Perception (BP). Brand perception, encompassing brand image, trust, and attitude, was measured using items adapted from Keller (2003). Example items include:
“This organic food brand has a high-quality image.”
“I trust this organic food brand.”
“Overall, I have a favorable opinion of this organic food brand.”
Purchase Intention (PI). Purchase intention was measured using items adapted from Dodds et al. (1991), such as:
“I am willing to buy products from this brand.”
“I would consider this brand as my first choice for organic food.”
Social Identity (SI). Social identity was adapted from social identity theory (Tajfel & Turner, 1979), focusing on group-based identification with organic/healthy consumption communities. Example items include:
“I feel a strong sense of belonging to people who care about organic or healthy food.”
“Being part of a group that values organic food is important to me.”
Information Credibility (IC). Perceived credibility of social media information about organic food was measured using items adapted from Metzger (2007), such as:
“The information about this brand on social media is trustworthy.”
“Social media posts about this brand provide accurate information.”
Control variables (e.g., age, gender, education level, frequency of social media use) were also included in the survey for robustness checks.
5.4 Data Analysis Procedure
Data analysis proceeded in several steps:
Descriptive statistics: SPSS was used to compute descriptive statistics for demographics, social media usage patterns, and item responses.
Preliminary regression analysis: As an initial step, multiple regression models were used to examine the direct relationships among social media language features, brand perception, and purchase intention. A simplified equation for brand perception is:
BPi=β0+β1PosEmoi+β2NegEmoi+β3KeyFreqi+β4Interactioni+εi
where Interaction represents engagement metrics such as likes and comments.
Measurement model assessment: Using AMOS 26.0, confirmatory factor analysis (CFA) was conducted to evaluate the reliability and validity of the latent constructs (BP, PI, SI, IC).
Structural Equation Modeling (SEM): SEM was used to test the hypothesized relationships (H1–H6), including direct effects, mediation (H4), and moderation (H5, H6).
Mediation analysis: Bootstrap procedures with 5,000 resamples were used to estimate indirect effects and their confidence intervals.
Moderation analysis: Multi-group SEM was conducted by splitting respondents into high vs. low social identity and high vs. low information credibility groups (median split). Model comparison using χ² difference tests assessed moderation.
6. Data Analysis and Results
6.1 Descriptive Findings and Social Media Language Features
The social media text mining analysis revealed distinct patterns in language features associated with organic food discussions. Positive emotional language—words associated with health, sustainability, freshness, and high quality—was more prevalent than negative emotional language and generated higher levels of engagement (likes, shares, comments). Posts that emphasized benefits such as “natural,” “healthy,” “green,” and “eco-friendly” attracted more interaction, indicating consumer interest in these themes (Berger & Milkman, 2012).
Negative emotional posts, particularly those discussing high prices, questionable authenticity, or disappointing product experiences, tended to receive lower engagement, suggesting that consumers may avoid or disengage from pessimistic content, even though such posts may damage brand evaluations.