Through the lens of Fairclough’s (1992) critical discourse analytic approach, this paper demonstrates how power is conveyed through word choices and how social media practices shape who is heard, influence the performance of identities, and determine which forms of mental health knowledge are legitimised or marginalised. This paper builds on calls for further research on the meaning-making process of health and illness in digital spaces (Conrad & Barker, 2010), with a focus on mental health. The potential of health communication continues to grow with the increasing development and advancement of social media platforms (Huo et al., 2019; Stellefson et al., 2020). The digital era, especially the social media age, offers multiple tools and opportunities that can be effectively leveraged for health communication initiatives. As such, ongoing research is necessary to understand the potential and impact of health communication on these platforms, for instance, how these platforms function and what makes communication goals and tactics practical. The World Health Organisation (2019) notes that one in four people is affected by mental illness at some point in their lifetime globally. The National Study of Mental Health and Wellbeing show that 15% of Australians aged 16-85 years suffered from high to very high levels of mental distress in 2020-21 (ABS, 2021).

Research suggests that online mental health communities benefit mental health well-being. Positive effects include gaining access to mental health information, receiving emotional support, and fostering a sense of community (Kim et al., 2023; J. Liu & Wang, 2021; Saha & Sharma, 2020). In contrast, other studies have shown that participating in online communities can negatively affect mental well-being. Adverse effects include increased levels of psychological distress, cyberbullying and decreased life satisfaction (Zhou et al., 2023). While there is an increasing body of knowledge on online communities for mental health promotion and education (Abouei et al., 2025; S. Liu et al., 2020; Nattembo, 2022; Tuan et al., 2024), this study adds to current knowledge by drawing on empirical evidence from posts and the experiences of users and moderators to examine mental health meaning-making on interactive website forums.

Critical health communication approaches examine how meanings and representations of health are associated with power and inequality linked to culture, resources, and social structures, and how power shapes cultural constructions of health and responses to illness (Zoller, 2014, p. 270). This “involves deconstructing dominant, taken-for-granted assumptions about health, often with the hope of introducing possibilities for alternative, more inclusive meaning systems,” (Zoller & Kline, 2008, p. 271). Critical mental health communication can provide invaluable qualitative insights into how illness is understood, experienced, and treated. Although a growing body of research exists about mental health communication on social media, few studies have utilised a critical perspective. Existing studies have primarily examined mental health narratives. For example, Michlig et al. (2022) conducted discourse analysis on mental health in a Somali-American community, identifying two key discourses: biomedical and community-driven mental health knowledge. The latter fuses psychiatric terms and employs stigma and social control to limit disclosure. Another study (Sukhera et al., 2022) used critical discourse analysis to examine mental health stigma in medical education, revealing an emancipatory discourse of disclosure that normalised help-seeking, which conflicted with a performance discourse, suggesting that public disclosure remains challenging in private contexts due to a medical culture that rewards perfectionism and heroism. However, limited research has examined mental health meaning-making on website forums, particularly the meanings of mental health and the power relations that shape the production, circulation, negotiation, consumption, and reproduction of mental health content. This study addresses this gap by taking a multilayered approach to analyse social media posts and user and moderator interviews.

Moderation of online mental health communities can improve the quality of online discourse by increasing user engagement, encouraging users to discuss negative emotions more candidly, and minimising disruptive behaviour, thereby enhancing the efficacy and safety of online mental health conversations (Wadden et al., 2021). Some research has focused on content moderation practices (Russell, 2024; Young, 2022), noting that, despite advancements in algorithmic tools, human moderators remain critical, and that there is also a need for more inclusive user-centred design (Haime & Biddle, 2025). Human content moderators analyse online comments and posts shared by users using the community guidelines defined by the social media platforms, which include policies on hate speech, targeted violence and bullying (Roberts, 2016). Other research has examined moderators’ experiences to improve their working conditions (Roberts et al., 2023; Spence et al., 2023; Steiger et al., 2021). Rapid growth in social media use during the pandemic necessitated a significant expansion of content moderation teams, creating challenges for training, communication, and maintaining sustainable management practices (Roberts et al., 2023). Moreover, moderators face significant challenges, including exposure to traumatic content, which can lead to lasting psychological and emotional distress (Steiger et al., 2021).

This study drew on the Faircloughian critical discourse analytical approach, which postulates that relations between discourse elements, such as texts, processes of production and consumption, and social and historical contexts, are interdependent and mutually constitutive (Fairclough, 1992). This analytical approach was valuable in analysing the mental health meaning-making process on social media across three dialectical levels. Firstly, discourse as text (micro/descriptive level) entails analysing the formal properties of the texts. Textual meanings are never fixed; they vary and often slip into conflicting counter-meanings. This level focused on analysing vocabulary/word choices, as well as text structure. Secondly, discourse as a discursive practice (meso/interpretation level) examines how text producers draw on available discourses as they create texts and how text recipients attach meaning to them. This level examined how identities and power relations were negotiated. Finally, discourse as a social practice (macro/explanatory level) articulates the connections among discourses, social and historical perspectives, and institutional practices to explain the interaction between the processes of production and consumption and their social effects. Analysis at this level focused on exploring the ideological and broader societal contexts that frame the discourses.

Online accounts offer naturally occurring data (Markle et al., 2015), providing unique opportunities to systematically examine people’s narratives over time and explore how narratives change, thereby affording the possibility of identifying patterns in the representation of meaning. Data were collected between April 2019 and July 2019, and between January 2020 and June 2020. The first period was selected because it was when the study began, and the second was selected to enrich the data set. Studies indicate that the COVID-19 pandemic exacerbated mental distress (Li et al. 2020; Van Rheenen et al. 2020) due to the drastic measures taken to contain the spread of the disease, such as the closure of schools, university campuses, shops, restaurants, and companies, and the implementation of physical distancing. Other studies noted that heightened mental distress increased the need to access mental health-related information online (Marshall, Dunstan & Bartik 2020). Posts (original posts and comments) were drawn from depression and anxiety-related website discussion forums that target people in Australia. Website forums were purposively selected because publicly available data were available. Data were collected from SANE, Headspace, and Beyond Blue websites forums. A purposive sample of relevant pages was selected. These threads were selected because they are managed by some of Australia's most reputable mental health organisations. A purposive sampling strategy was used to select information-rich threads aligned with the study's focus. Threads were included if they had substantive engagement (comments), explicit discussion of lived experience, or broader discursive patterns around mental health. The unit of analysis was the thread, which included original posts and comments, enabling examination of individual expressions and interactional dynamics, such as how meanings are negotiated or contested. Inclusion criteria included threads that referenced anxiety or depression, demonstrated active participation, and were publicly accessible. Threads were excluded if they had minimal engagement or were not related to mental health discourse, even when they matched the search terms.

Nunkoosing (2005) suggests that participant narratives are important because they enable individuals to reflect on some aspect of their lived experiences. Interviews provided a nuanced understanding of the diverse perspectives on mental health on social media platforms. Twelve people participated in the interviews, comprising seven users and four moderators of mental health-related social media platforms. The sample size was deemed sufficient to answer the research questions after data saturation was reached. Morse (2000) notes that sample size should depend on factors such as the topic, study scope, and research design. The ethnically and culturally diverse sample ensured a range of perspectives in the data collection and analysis.

Respondents were recruited through mental health organisations and online communities. Among the seven users, three were studying for postgraduate degrees, and four were Respondents were recruited through mental health organisations and online communities. Among the seven users, three were studying for postgraduate degrees, and four were Respondents were recruited through mental health organisations and online communities. Among the seven users, three were studying for postgraduate degrees, and four were working, primarily young urban individuals aged 20 to 30+. Participants were identified using convenience sampling. All participants engaged with various social media platforms. Interview participant characteristics are provided in Appendix 3. Their experiences prompted reflections on their information sharing, identity, relationships, and online conventions. Interviews were conducted after analysis of posts. Four moderators from different organisations with varying experience in moderating online mental health resources, ranging from 3 to 30 years, were interviewed. One was a commercial moderator managing various platforms; two moderated website forums, and another managed a layperson’s Facebook page. By interviewing moderators, this paper examined moderation practices to understand their role in shaping mental health constructions on social media and how users navigated these contexts. Interviews followed a semi-structured approach, using the guides in Appendix 2. Leavy (2014) notes that semi-structured interviews can produce knowledge by allowing the researcher to direct the conversation to the issues most pertinent to the research questions. Interviews were conducted via Zoom, email, and face-to-face and took 45 minutes. Interviews were conducted in English and transcribed verbatim to ensure critical data were not lost.

The University of Adelaide Ethics Review Committee in Australia granted ethical approval (H-2019-149) prior to the study commencing. Ethics approval was granted for interviews and the analysis of publicly available user-generated posts on the website forums. Although the posts were publicly available, safeguards were considered to reduce re-identification and the traceability of quotations. Identifying details such as usernames were removed or generalised, and contextual markers that could enable reverse searches were slightly edited. When reporting findings, care was taken with direct posts, as they could be traced via search engines. Posts quoted in the findings were paraphrased or slightly edited to retain meaning while minimising risk of source identification. Written and verbal consent were obtained from all respondents before the commencement of interviews. The consent process included permission to audio-record interviews, with participants informed recordings were for transcription and analysis. Participation was voluntary, and participants were informed that they could decline responding to any question or withdraw at any time. Participants also consented to the use of anonymised responses in publications and presentations.

Data were analysed using the analytical framework outlined in Appendix 1 in NVivo version 12. Critical discourse analysis enabled the examination of taken-for-granted meanings in the posts and interviews. Interviews and initial analysis proceeded simultaneously to enable Data were analysed using the analytical framework outlined in Appendix 1 in NVivo version 12. Critical discourse analysis enabled the examination of taken-for-granted meanings in the posts and interviews. Interviews and initial analysis proceeded simultaneously to enable ongoing assessment of thematic sufficiency and to identify the point at which additional interviews no longer provided substantially new insights. Saturation was reached when subsequent interviews yielded diminishing returns in terms of new themes or conceptual developments. Saturation was evaluated through code saturation (the point at which no new codes emerged) and meaning saturation (the point at which no new dimensions, nuances, or insights appeared within existing themes). Although the sample size aligns with established qualitative research practices for in-depth thematic analysis, it is recognised that saturation depends on context and is influenced by factors such as sample heterogeneity, scope of inquiry, and methodological approach. In this study, the unique participant characteristics and the targeted nature of the research questions allowed saturation to be reached within 11 interviews. Nonetheless, the decision to cease data collection was guided not by a predetermined number but by systematic, iterative evaluation of analytic completeness.

Data analysis involved an iterative multi-stage coding process. In the first stage, preliminary codes were developed inductively from close readings of interview transcripts and posts from website forums. Interview data provided reflective, retrospective accounts of participants’ experiences, while forum posts offered naturally occurring, interactional discourse. In the second stage, preliminary codes were systematically compared and grouped into broader themes through constant comparison. In the final stage, themes were interpreted using Fairclough’s three-dimensional critical discourse analysis. Textual analysis examined linguistic features; analysis of discursive practice examined processes of production, distribution, and consumption; and analysis of social practice situated these findings within broader discourses of mental health.

Freshwater & Cahill (2012) suggest that researchers should continually reflect on their approach and rationale for the data to increase transparency in the research process. In addition to self-reflection, meetings were held between senior researchers with backgrounds in media and communications and the author (the lead analyst), who has professional experience and postgraduate training in health communication research and is a certified mental health first aid provider, to discuss and refine interpretations of the data, to help mitigate research bias.

The paper employs Fairclough’s (1992) three-dimensional model of critical discourse analysis to investigate the construction of mental health discourse on social media. This approach helped reveal how language reflected realities about mental health, personal agency, and institutional power. The findings focus on the textual features; multimodal aspects are excluded as they were analysed using a different analytical framework. The study presents the three dominant themes derived from social media posts and interviews with users and moderators, providing a historical account of the production, circulation, and consumption of mental health information on social media. Themes are exemplified through sample social media posts and interview excerpts, presented using Fairclough’s critical discourse analysis approach. The findings show that users and moderators of mental health social media pages and website forums drew upon different experiences and knowledge to construct mental health. Two broad themes were found: support structures and risks and triggers.

The study examines how mental health is constructed on social media, drawing on social media posts and the experiences of moderators and users. This study shows how mental health discussions on social media are dynamic and contested, with users, moderators, and technology co-creating meaning. Using Fairclough’s three-dimensional model revealed that beneath everyday posts and moderation activities lie complex ideological processes that both uphold and challenge dominant views of mental health. Across the two main themes—risks, triggers, and support structures—the analysis highlighted how digital communication both empowers and controls individuals within a broader culture of surveillance and self-regulation.

The theme of risks and triggers highlights how users frame distress by using language related to threats and triggers, indicating a shift from simply expressing emotions to actively managing them. Moderators’ discussions around safety and trust reinforce this perspective, viewing online mental-health communities as caring spaces shaped by moral and institutional guidelines. The discursive-practice analysis showed that moderators play a strong gatekeeping role in shaping the circulation of mental health discourse through a complex balancing act that prioritises safety while supporting users emotionally. This aligns with Gillespie’s (2018) argument that platform moderation serves as a form of custodial authority, where both human and algorithmic entities collaboratively shape the norms of acceptable vulnerability. Although moderation practices aim to protect users, they also limit genuine expression by making some emotions more acceptable than others. This change reflects a broader trend in mental health, transforming it into a sphere of controlled vulnerability where openness is expected to stay within socially and technologically defined limits. Fairclough (1992) argues that discourses become powerful not only through linguistic choices but also embedding in ideologies. The theme of risks and triggers in this study demonstrates a broader riskification ideology in which mental health is understood through a risk lens, necessitating continuous monitoring, management, and intervention. Social media environments strengthen this ideology by emphasising moderation, highlighting risk narratives. This shows a worldview where vulnerability is both expected and monitored. User self-awareness, expressions, and moderator enforcement create a space where vulnerability becomes manageable. This aligns with Foucault’s (1977) notion of biopolitical regulation, in which institutions influence behaviour through normalisation rather than direct control. Consequently, the findings support critical discourse analysis postulates by showing how digital spaces amplify ideologies, integrating them into everyday language and emotional expression (Fairclough, 1992).

The theme of support structures shows how discourses of biomedicalisation and empowerment coexist in tension. Users expressed agency through peer support and self-care, but these actions often align with ideals of personal responsibility and resilience. Moderators navigate the tension between human empathy and automated governance, simultaneously implementing and challenging institutional authority. The analysis exposes the tension between care and control that underlies content moderation practices on online platforms (Haime & Biddle, 2025). The same systems designed to protect individuals can enact disciplinary power through diagnosis, risk assessment, and institutionalisation. This analysis highlights the ambivalent nature of online mental health discussions, which are at once liberating and monitored. Digital platforms allow individuals with lived experience to share alternative narratives that challenge institutional authority, but they are also moderated by algorithms and humans who reinforce normative ideas about what is acceptable to express (Zhang et al., 2024). The findings demonstrate that automated systems are not neutral background tools but active discursive actors that shape which narratives appear permissible. This study shows that platform affordances in digital mental health interactions act as discursive conditions that shape how meaning is created. Algorithms, moderation protocols, and platform structures serve as non-human contributors to discourse, expanding the scope of CDA from linguistic analysis to include technological mediation. This analysis extends CDA into the realm of ongoing social media discourse broadly and algorithmically mediated discourse specifically, supporting contemporary calls by KhosraviNik (2022) for CDA to incorporate analysis of technological affordances.

This study's findings have important implications for digital mental-health communication, policy development, and moderation practices. Understanding that social media serves both as a support space and a surveillance tool calls for a multidimensional approach to creating safer, more equitable online environments. Platform moderation and policy frameworks need to evolve from solely risk-averse approaches to models based on relational safety. This shift emphasises the importance of emotional honesty, enabling users to express their distress without fear of punishment or algorithmic suppression. Moderator training should prioritise trauma-informed communication and cultural awareness instead of mere procedural adherence. Policies and automated systems should be transparent and developed with input from users with lived experience, through participatory methods such as community review panels or co-created guidelines, promoting democracy and trust in online mental health spaces. Campaigns and initiatives should explicitly consider how cultural norms around masculinity, productivity, and positivity affect emotional expression. Interventions encouraging vulnerability and diverse wellness perspectives can challenge the dichotomy of ‘normal’ versus ‘abnormal’ mental health. Investing in peer-led support networks that acknowledge social determinants can complement professional services, empowering users without leaving them unsupported.

While this research offers valuable insights into how mental health is discussed on social media, several limitations should be recognised. The analysis primarily relied on publicly available posts and interview snippets with users and moderators. Although these sources provided rich linguistic and thematic details, they do not fully capture private, ephemeral, or direct-message interactions, where most vulnerable mental health disclosures occur. Consequently, the results may reflect more moderated, self-aware self-presentation rather than the full spectrum of users’ emotional expressions. Moreover, the focus was on a specific digital environment, which restricts how broadly the findings can be applied across different cultural or platform contexts. Mental health meaning-making is likely to vary across regions, languages, and platform features. The study also did not incorporate triangulation with clinical or policy insights, which could have helped better understand how institutional discourses interact with everyday online communication. For future research, it is essential to compare platforms and cultures to see how platform design, moderation practices, and local societal norms influence mental health discussions differently. Longitudinal studies would allow tracking how discourse patterns evolve over time, particularly in response to platform policy changes, public health crises, or shifts in social perspectives. Additional research could include participant-generated materials, such as diaries or personal narratives, to deepen understanding of how users manage their identities across public and private digital spaces.

Overall, the findings confirm that digital mental health communication is not just therapeutic but also ideological. Social media platforms reflect societal discourses on risk, normality, and responsibility while providing spaces for resistance and redefinition. Future research and policy should address these tensions by encouraging digital environments that prioritise structural understanding over procedural safety and individualised self-regulation.