The Siren's call: How social media influencers are using identity leadership to shape diagnostic label identification and self‐care intentions

Abstract

As social media becomes a central source of mental health information, there is growing concern that influencers can lead people to interpret everyday experiences as symptoms of mental ill health. Across two experimental studies that utilized distinct media formats, we examined how, by creating, representing, advancing and embedding a shared sense of ‘us’ (i.e. engaging in identity leadership), influencers' ADHD‐related content shapes perceptions of self, symptoms and behavioural intentions. In Study 1 (N = 289 men) authentic video stimuli demonstrating high‐ versus low‐identity leadership were extracted from TikTok, whilst in Study 2 (N = 259 men) tightly controlled text posts were used. Across both studies, identity leadership was associated with greater identification as ADHDers, help‐seeking attitudes and self‐care intentions. Planned comparisons in Study 1 provided some evidence that high‐identity‐leadership content increased participants' identification as ADHDers. In Study 2, identity leadership was particularly important for mental health professional influencers (compared to lay influencers) in promoting self‐care intentions. These findings suggest that influencers can use identity leadership to meaningfully influence both people's understanding and response to mental health content.

INTRODUCTION

Access to mental health care is critical for people experiencing, or at risk of, mental illness (Saxena et al., 2006). Despite 1 in 8 people worldwide experiencing mental health conditions, 85% of them fail to seek professional support, with common barriers including cost, long wait times and limited access to services (Rosenberg et al., 2022; Vacher et al., 2023; World Health Organization, 2025). Consequently, many turn to online platforms for mental health support, seeking either peer connection or professional‐level advice that is free and available on demand (Naslund et al., 2016). Social media, in particular, has become a dominant source of mental health content (Akhther & Sopory, 2022). On TikTok alone, videos with hashtag #mentalhealth had reached 25.3 billion views worldwide by 2022 (Basch et al., 2022).

Although there are many benefits of this type of peer connection and support, recent research highlights two critical concerns. First, social media content can pathologize everyday experiences by presenting them as symptoms of mental illness, often without appropriate clinical context, which may encourage inaccurate self‐diagnosis (Yeung et al., 2022). This has been linked to problematic outcomes, including use of unprescribed medications that may exacerbate mental health difficulties or delay appropriate care (Monteith et al., 2024). Second, beyond describing symptoms, these platforms can also promote social identification with diagnostic labels, which can lead to poorer well‐being (Cruwys & Gunaseelan, 2016).

This raises important questions about how influencers shape both (a) how viewers interpret their experiences (e.g. seeing common behaviours as signs of mental health conditions) and (b) what those viewers choose to do in response (e.g. adopting suggested self‐care behaviours or seeking further help from professionals). The present research explores this using the social identity approach to leadership (Haslam et al., 2020) with a specific focus on ADHD. Specifically, we examine whether influencers' use of identity leadership strategies increases social identification with diagnostic labels and influences their behavioural intentions to engage in self‐care and help‐seeking (defined as an individual's willingness to engage with professional mental health services; Hammer et al., 2018).

We begin by outlining the emerging literature on mental health information on social media, followed by a social identity analysis of how social influence processes shape perception and behaviour—generally and in social media contexts. We then present two experimental studies investigating how real influencer content on social media uses identity leadership strategies to shape viewers' social identification with attention‐deficit hyperactivity disorder (ADHD) and their intentions to engage in self‐care behaviour.

Mental health‐related content on social media

Social media has rapidly become a primary source of health information. For example, Marar et al. (2019) found that 85% of hospital patients and their companions reported using social media for this purpose. One issue with this is that, as well as providing free access to useful health information, social media also exposes people to a plethora of misinformation. For example, an analysis of popular ADHD‐related TikTok videos found that 52% were misleading and only 21% were clinically accurate (Yeung et al., 2022). Much of this misinformation comes from non‐healthcare providers who present everyday behaviours, such as procrastination or difficulty focusing, as definitive signs of ADHD. This kind of framing may lead viewers to misinterpret ordinary non‐clinical experiences as symptoms of a mental health condition or overestimate symptom severity (Karasavva et al., 2025). Given that the top 100 ADHD videos have accumulated nearly half a billion views, the scale of this impact is immense. One factor that can make this content particularly influential is the way that it is used to create communities of viewers who begin to see themselves in these descriptions of symptoms (Corzine & Roy, 2024). Indeed, in this way, online spaces can become more than just sources of information: places where people develop social identities centred around shared experiences, and which may thereby contribute to an increasing risk of self‐diagnosis. Given the prominence and influence of ADHD content, understanding how it shapes perception and behaviour requires examining the social psychological processes—particularly those related to social identity—that underpin its appeal and persuasive power.

The social identity approach

The social identity approach provides a theoretical framework for understanding how membership in online mental health communities can shape the way that individuals perceive themselves and act. This perspective, which incorporates both social identity theory (Tajfel et al., 1979) and self‐categorization theory (Turner et al., 1987), suggests that people's sense of self derives from both their personal identity (‘who I am’; e.g. ‘I have ADHD’) and their social identities (‘who we are’; e.g. ‘we ADHDers’). The social identity approach further proposes that, when individuals define themselves in terms of their social identities (e.g. as ‘we ADHDers’), this shapes their perceptions, emotions and behaviours. This influence has been documented across various domains in both beneficial and detrimental ways. For example, greater social identification as a member of an exercise group, a therapy group and a school have been shown to positively influence people's exercise participation, therapeutic outcomes and academic achievement, respectively (Anderson et al., 2009; C. Haslam et al., 2016; Reynolds et al., 2020; Stevens et al., 2022). These positive effects can be underpinned by a range of psychological processes. For example, identifying more strongly as a member of a particular group increases people's motivation to align their behaviours with representative group members (i.e. in‐group norms; Turner et al., 1987). If group norms are positive (e.g. encouraging regular exercise or preparation for exams) then this desire to enact those norms will generally have positive downstream implications for people's own behaviour. However, social identification has also been shown to undermine well‐being and impede illness recovery when people identify with a stigmatized identity—for example, as someone with a mental illness or substance user (Cruwys & Gunaseelan, 2016; Dingle et al., 2019).

In these contexts, the influence of social identity is considerable, shaping not just behaviour but also perception. For example, Levine and Reicher (1996) found that female sports science students judged injuries differently depending on the social identity that was salient for them: appearance‐related injuries were rated as more serious when their gender identity was primed, whereas disabling injuries were rated as more serious when their identity as sports science students was salient. Similarly, St. Claire et al. (2008) showed that participants perceived medical symptoms as more severe when those symptoms were relevant to their salient illness identity. Here, social identity provides the interpretative lens through which individuals appraise ambiguous bodily sensations. Perception is thus not a passive reflection of biological states, but an active process where health experiences gain meaning and significance based on their implications for the salient social category.

In online mental health communities, this dynamic has the potential to take on particular significance. Here, social identity theorizing would suggest that when influencers promote certain experiences as symptoms of mental health conditions, they can create new social identities around these diagnostic labels. Some research suggests that this influence could, in turn, lead to negative outcomes (Cruwys & Gunaseelan, 2016). In part, this is because as viewers begin to identify with these diagnostic categories, they may reinterpret their own experiences through the lens of this new identity. Moreover, whilst individuals may be experiencing genuine distress, the likelihood of receiving an accurate diagnosis and appropriate care could be compromised if they adopt a diagnostic label that does not align with clinical criteria (Saunders, 2023).

Identity leadership

Given the powerful role that social identity plays in shaping how people interpret their experiences, it is important to understand who influences these identities and how. In this regard, a considerable body of research has shown that this influence can be, and often is, exerted by leaders—specifically, those who engage in social identity leadership (Haslam et al., 2020). According to the identity leadership framework, leaders will be more influential to the extent that they create and advance a sense of shared social identity with potential followers. More specifically, leaders are predicted to be more influential as a function of four features that together comprise identity leadership and which provided the theoretical framework for the experimental stimuli in the present study: (a) their identity prototypicality—being seen as a model group member (e.g. an influencer openly disclosing their own mental health struggles could signal ‘I'm one of us’), (b) their identity advancement—being seen to promote the group's interests (e.g. using their platform to engage with policymakers or destigmatize mental health issues in ways that benefit their followers), (c) their identity entrepreneurship—being seen to create a sense of shared identity (e.g. using inclusive language like ‘we ADHDers’ or sharing memes that reinforce a common experience); and (d) their identity impresarioship—being seen to create opportunities for group members to live out their identity (e.g. hosting virtual meetups, challenges or campaigns that allow people to participate and express themselves as part of the group; see also Steffens, Haslam, Reicher, et al., 2014).

In line with these propositions, research has shown that leaders in a range of contexts are more effective to the extent that they engage in identity leadership. For example, in organizational contexts, identity leadership predicts employees' organizational citizenship, innovation and job satisfaction (Bracht et al., 2023; van Dick et al., 2018). In political contexts, it predicts supporters' political commitment and engaged followership (Haslam et al., 2023). In sport contexts, it increases team members' participation in team sessions and well‐being (e.g. Fransen et al., 2020; Stevens et al., 2020; see Stevens et al., 2021 for a review).

Yet whilst the effect of identity leadership has been well‐documented in these traditional leadership settings (e.g. politics, organizations, sport), its role in social media remains understudied. In a recent meta‐analysis, Han and Balabanis (2024) found that research examining social media influence has focused on message and leader attributes, such as credibility and perceived expertise, with little attention to social identity. This gap is important because social identity is central to how such attributes are interpreted, especially in online spaces where people form communities around shared identities and experiences (Bingley et al., 2025).

Indeed, to our knowledge, only one study to date has explored the role of identity leadership in social media contexts. In this, Uysal et al. (2022) showed that identity leadership promotes collective action. Investigating participation in a protest in Turkey, they found that 70% of protestors initially learned about the event through social media and that it was in an online context that they formed a new group identity as ‘protestors’. Certain social accounts were also observed to actively encourage participation and to be successful in doing so (see also Haslam et al., 2023). Supporting these claims, Uysal and colleagues measured protesters' social identification with the protest group, their participation frequency, and how much they perceived the account as demonstrating identity leadership. Results showed that perceived identity leadership positively predicted participation and that this was mediated by social identification with the protest group. However, whilst this study speaks to the role of identity leadership in influencing behaviour via social media, its observational design prevents causal conclusions about identity leadership's effectiveness in such contexts.

Bringing these threads together, we propose that part of the appeal and influence of information about mental health on social media arises from the ways that influencers leverage identity leadership techniques. Critically, this involves building a shared identity with audiences that might cohere around existing group membership, such as one's age group or cultural background. There is also anecdotal evidence that diagnostic labels themselves often form the basis for these shared identities. To give some sense of how this might occur, consider a TikTok video discussing ‘symptoms of undiagnosed ADHD’, in which the influencer suggests that behaviours like rewatching TV shows and craving sweet things could be signs of ADHD. Such (mis‐)information might lead a person without ADHD to believe that they have the condition on the basis of these, arguably common, experiences. Indeed, from an identity leadership perspective, one can see this as a kind of identity entrepreneurship around ADHD (Reicher et al., 2005). And if viewers internalize this identity, they may then reinterpret their experiences through this lens and be more likely to adopt behaviours promoted by these identity leaders (i.e. influencers).

The present research

The goal of the present research was to examine how social media influencers affect viewers' diagnostic label identification and behavioural intentions through identity leadership processes. Whilst previous research has demonstrated identity leadership's effectiveness in offline settings (e.g. Haslam et al., 2020; Hou et al., 2021; Stevens et al., 2020) and its potential in social media contexts (Uysal et al., 2022), its role in shaping health‐related perceptions and behaviours (e.g. self‐care behaviours) via social media remains unexplored. To pursue this goal, across two experimental studies, we investigated whether influencers who exhibit identity leadership—such as communicating shared identity, advocating for the group's interests and fostering a sense of belonging—can shape how viewers perceive ADHD‐related content. Specifically, we examined two key outcomes: (1) social identification as an ‘ADHDer’ (a popular social media term for a person with ADHD) and (2) intention to engage in promoted self‐care behaviours (e.g. exercise, diet). Additionally, help‐seeking attitudes were included as a more distal, exploratory outcome to examine potential spillover effects of identity‐based ADHD influencers in the broader clinical context. To test these effects, the first study used real TikTok videos, whilst the second employed controlled text social media posts to eliminate confounds inherent in video stimulus.

STUDY 1

Study 1 had a three‐condition design in which participants watched one of three authentic TikTok videos sourced from social media. In two of the conditions, the videos were ADHD‐related and promoted self‐care behaviours. The influencer displayed either high or low identity leadership depending on the condition. The third video contained clips of cute animals (a control condition). We had two key hypotheses. First, that participants' intention to engage in self‐care behaviours would be greater following the high identity leadership video than the low identity leadership and control videos (H1). Second, that participants who watched the high‐identity leadership video would more strongly identify as an ‘ADHDer’ than participants who watched the low identity leadership or control videos (H2).

Method

Participants and design

A total of 303 people were recruited via Prolific (www.prolific.com), an online platform specifically designed for academic research that facilitates access to diverse participant pools. Prolific's pre‐screener function was used to display the study advertisement only to men aged 18–30 years from English‐speaking countries: the United States (n = 175), the United Kingdom (n = 85) and Australia (n = 43). The eligibility was restricted to young men from these countries to increase apparent similarity between participants and the influencers (e.g. in gender, cultural references, accents, etc.) in the video stimuli. We reasoned that this baseline level of similarity may be a precondition for influencers to gain identity‐based influence. After 14 respondents were excluded for failing a manipulation check, the final sample comprised 289 men aged 18 to 30 (M = 25.09, SD = 3.44).

Power analysis

An a priori power analysis conducted in G*Power (Faul et al., 2007) for a one‐way fixed‐effects ANOVA with three groups, using a small effect size (f = .20) and power of .80, indicated a minimum required sample of 246. A small effect was anticipated given the brief exposure (a single video), which differs from the extended and repeated engagement typical of real‐world social media use. To account for potential withdrawals, the recruitment goal was set at 300, and we exceeded the power target with a final sample of 289. A sensitivity power analysis using G*Power indicated that with the final sample of 289 participants, the design had 80% power to detect an effect of f = 0.18 in a one‐way between‐subjects ANOVA.

Materials and procedure

The study was administered online via Qualtrics. After providing informed consent, participants were asked to confirm their gender and age before being randomly assigned to one of the three experimental conditions.

Experimental manipulation

The high and low identity leadership conditions featured real influencers discussing the same ADHD‐related symptoms (procrastination, disorganized spaces) and self‐management strategies (exercise, high‐protein diet). Short segments from TikTok videos were compiled to create stimuli that were matched on length, content and visual format¹—allowing us to isolate differences in identity leadership (e.g. guided by Robertson, Cruwys, Stevens, & Platow, 2025).

For the high identity leadership condition, the video featured a younger male influencer (approximately 30 years old). This influencer demonstrated identity prototypicality by aligning with the typical age cohort of ADHD communities on social media, choosing an ADHD‐relevant account name, and describing himself as having ADHD (e.g. ‘ADHD lesson I wish someone told me after getting diagnosed’). He emphasized building community and creating content for others with the condition (showing identity advancement and impresarioship) and defined what it means to be a person with ADHD in ways that showed identity entrepreneurship. The low identity leadership condition featured an older man (approximately 60 years old) who presented himself as a health professional. He used more clinical or technical language when discussing ADHD symptoms (e.g. ‘chronic procrastination’) and management strategies (e.g. ‘aerobic exercise’) and did not convey a sense of shared identity or provide any group‐based framing. The control condition featured a neutral, non‐health‐related video featuring a range of cute animals (e.g. cats and dogs). This served as a baseline to control for the effects of generalized exposure to positive social media content.

Following the video, participants completed the following measures in order:

Video engagement

Two measures assessed participants' engagement with the video content. Video enjoyment was measured using a 4‐item scale adapted from Cruwys et al. (2022; e.g. ‘How much did you enjoy watching this video?’, where 0 = not at all, 10 = very much; α = .86). Video sharing intention was measured using a single‐item slider scale (‘How likely are you to recommend this video to others? where 0 = not at all likely, 100 = highly likely’).

Perceived identity leadership

To evaluate the effectiveness of the manipulation, participants completed the Identity Leadership Inventory (ILI; Steffens, Haslam, Reicher, et al., 2014). This scale has excellent reliability (α = .93 in Steffens, Haslam, Reicher, et al., 2014; α = .95 in the current study). Participants rated their perceptions of the influencer's identity leadership (e.g. ‘This influencer stands up for people with ADHD’) on a scale from 1 (not at all) to 7 (completely). This variable was not measured in the control condition as it was not applicable.

Identification as an ADHDer

This was measured using Leach et al.'s (2008) scale (e.g. ‘I feel a bond with a person with ADHD’; 1 = strongly disagree, 7 = strongly agree; α = .86 in Leach et al. (2008); present α = .89).

Self‐care intentions

The intention to engage in self‐care behaviours was measured with an adapted eight‐item version of the Physical Care subscale of the Mindful Self‐Care Scale (α = .70 in Cook‐Cottone & Guyker, 2018). Participants rated their behavioural intentions to exercise and diet (the self‐management strategies presented in the videos) for the following week (e.g. ‘I will exercise at least 30 to 60 minutes’). Responses were made on 5‐point scales (α = .77) from 1 (never) to 5 (regularly).

Help‐seeking behaviour attitudes

Attitudes towards seeking professional help for mental health concerns were measured using the Mental Help Seeking Attitudes Scale (α = .94 in Hammer et al., 2018; present α = .93). Participants responded to the statement, ‘If I had a mental health concern, seeking help from a mental health professional would be…’ on a 7‐point scale with 9 items from − 3 (e.g. useless) to + 3 (e.g. useful).

Manipulation check

To assess whether participants had paid attention to the video stimulus, they were asked as follows: ‘What did you see in the video?’ and selected from three options corresponding to the experimental conditions.

Results

Descriptive statistics and correlations among key variables are presented in Table 1. There were no missing data. All variables demonstrated acceptable skewness and kurtosis (between −3 and +3; Kline, 2023). Levene's tests confirmed homogeneity of variance for the primary dependent variables. Whilst 12 univariate outliers were identified across the ILI (n = 8), identification as an ADHDer (n = 1), self‐care intentions (n = 1) and help‐seeking attitudes (n = 2), their removal did not substantially alter the results. All cases were therefore retained to preserve statistical power.

TABLE 1.

Correlations among focal variables in Study 1.

Variable	M (SD)	1	2	3	4	5	6
1. Condition	–	–	–	–	–	–	–
2. Identity leadership	4.92 (1.18)	.19**	–	–	–	–	–
3. Identification as an ADHDer	4.19 (1.07)	.15*	.35***	–	–	–	–
4. Intention to self‐care	2.84 (0.72)	.03	.17*	.06	–	–	–
5. Help‐seeking attitude	5.27 (1.24)	‐ .02	.26***	.10	.13	–	–
6. Video enjoyment	5.32 (2.58)	.13	.68***	.49***	.18*	.20**	–
7. Video sharing intention	49.45 (31.25)	.11	.56***	.49***	.19**	.22**	.80***

Note: Condition was coded as 0 = low identity leadership and 1 = high identity leadership, without control condition; *p < .05, two‐tailed; **p < .01, two‐tailed; ***p < .001, two‐tailed N = 289, except for Identity Leadership (n = 190), which was only measured in the experimental conditions.

Main analyses

Manipulation check

An independent samples t‐test confirmed the effectiveness of the identity leadership manipulation. Participants in the high identity leadership condition (M = 5.15, SD = 1.02) reported that the presenter in the video had significantly higher levels of identity leadership than the presenter in the low identity leadership condition (M = 4.71, SD = 1.28), t(188) = 2.61, p = .001, d = 0.38.

Neither the one‐way ANOVA (F (2, 286) = 0.15, p = .859, 𝜂² = .001) nor planned comparisons between specific conditions revealed a significant effect of condition on self‐care intention. This hypothesis was not supported.

A one‐way ANOVA revealed no significant main effect of condition on identification as an ADHDer (F (2, 286) = 2.25, p = .107, 𝜂² = .02). However, to test our a priori prediction that high identity leadership would be more effective than low identity leadership and the control condition (H2), we conducted planned pairwise comparisons. These comparisons indicated that participants in the high identity leadership condition reported significantly stronger identification (M = 4.38, SD = 1.05) than those in the low‐identity leadership condition (M = 4.07, SD = 1.01), t(286) = 2.02, p = .045, d = 0.29, but did not differ from the control condition (M = 4.13, SD = 1.14), t(286) = − .42, p = .109 (see Figure 1), providing partial support for H2.

FIGURE 1.

Levels of identification as an ADHDer as a function of experimental condition. Error bars represent standard error.

Exploratory analyses

Exploratory correlation analyses showed that perceived identity leadership was strongly associated with video enjoyment (r = .68, p < .001) and sharing intention (r = .56, p < .001), indicating that content engagement was closely tied to identity leadership perceptions. For all mediation analyses in this article, the predictor variable was the experimental condition, which was theorized to directly target the mediator.

Help‐seeking attitudes

To assess potential spillover effects, a one‐way ANOVA revealed no significant effect of condition on help‐seeking attitudes (F (2, 286) = 1.14, p = .322, 𝜂² = .01).

Mediation effects of identity leadership

To test whether experimental condition (high vs. low) affected outcomes indirectly via perceived identity leadership, a series of bootstrapped mediation models were run using 5000 samples. Across all three models in Study 1, the experimental condition accounted for 3% of the variance in the mediator, perceived identity leadership; all reported coefficients are standardized (β). The control condition was not included in this analysis as identity leadership was not measured in this condition. The purpose of these analyses was to determine whether the observed effects of the experimental condition participants were assigned to on identification, self‐care intentions and help‐seeking attitudes were driven by the perceived identity leadership of the influencer. That is, they enabled us to test whether identity leadership was a mechanism of social influence in this context.

For ADHD identification, being in the high identity leadership condition predicted greater perceived identity leadership (β = .37, p = .010), which predicted ADHD identification (β = .33, p < .001; see Figure 2). The overall model accounted for 13% of the variance in ADHD identification (R ² = .13). The indirect effect (IE) was significant, IE = .12, SE = .05, 95% CI [.03, .24]. This suggests that exposure to high identity leadership content increased perceived identity leadership, which, in turn, was positively associated with stronger identification as an ADHDer.

FIGURE 2.

Mediation model: effect of exposure to high‐ versus low‐identity leadership videos on identification as an ADHDer. All numbers are standardized regression coefficients, *p < .05, **p < .01, ***p < .001.

For self‐care intentions, experimental condition predicted identity leadership (β = .37, p = .010), which in turn predicted self‐care intentions (β = .17, p = .023; see Figure 3). The overall model accounted for 2% of the variance in self‐care intention (R ² = .02). There was also again a significant indirect effect, IE = .06, SE = .04, 95% CI [.004, .15]. This suggests that exposure to high identity leadership content promoted participants' self‐care intentions through identity leadership.

FIGURE 3.

Mediation model: effect of high versus low identity leadership videos on self‐care intentions. All numbers are standardized regression coefficients, *p < .05, **p < .01.

Similarly, for help‐seeking attitudes, experimental condition predicted identity leadership (β = .37, p = .010), which in turn predicted help‐seeking attitudes (β = .27, p < .001; see Figure 4). The overall model accounted for 7% of the variance in help‐seeking (R ² = .07). And again, there was a significant indirect effect, IE = .10, SE = .05, 95% CI [.02, .21]. This suggests that exposure to high‐identity leadership content shaped participants' help‐seeking attitudes through identity leadership.

FIGURE 4.

Mediation model: effect of high versus low identity leadership videos on help‐seeking attitudes. All numbers are standardized regression coefficients, **p < .01, ***p < .001.

Discussion

Study 1 examined whether identity leadership influenced diagnostic label identification and self‐care intentions in an online context. By utilizing real influencer content sourced directly from social media platforms, we aimed to observe these psychological dynamics in a naturalistic setting. Regarding our primary hypotheses, H1 was not supported, as the identity leadership condition did not directly increase self‐care intentions. For H2, although the overall effect of identity leadership condition on social identification as an ADHDer did not reach statistical significance, the finding that participants in the high‐identity leadership condition reported significantly stronger identification than those in the low condition provided partial support for our hypothesis. It is worth noting, too, that average identification scores were above the scale midpoint in all three conditions (including the control), suggesting that a substantial proportion of young men already self‐define in this way.

We also found that perceived identity leadership consistently mediated the relationship between experimental condition and all three outcomes: identification as an ADHDer, self‐care intentions and help‐seeking attitudes. These results suggest that identity leadership is an effective strategy that influencers can use to shape both diagnostic label identification and intention to engage in self‐care behaviour online. Such results resonate with previous research which has confirmed the capacity for identity leadership to foster new shared identities and drive associated behavioural outcomes in digital environments (Uysal et al., 2022). Whilst this study explored responses to real‐world social media by using videos sourced directly from online platforms, its internal validity was constrained because we could not perfectly match all features of these naturalistic videos across experimental conditions (e.g. the speaker's age, tone, language, production quality). Study 2 sought to address this limitation.

STUDY 2

Study 2 had a 2 × 2 experimental design and used text social media posts that afforded higher levels of control than in Study 1. In particular, all content was held constant across conditions except for features directly related to the manipulated variables: identity leadership (high vs. low) and mental health professional status (present vs. absent). Prior research indicates that professional self‐disclosure of mental health symptoms can have both positive and negative consequences—enhancing perceived empathy whilst potentially compromising perceived professionalism (Robertson, Cruwys, Quayle, et al., 2025). Study 2 extended this by examining how professional status interacts with identity‐based disclosures on social media. This also helped disentangle potential confounds from Study 1, where the low identity leadership condition featured an influencer who used clinical terminology—raising the possibility that perceived professionalism may have contributed to the observed effects. Consistent with Study 1, we hypothesized that posts featuring high levels of identity leadership would increase (a) people's intentions to engage in self‐care behaviours (H1) and (b) their identification as an ‘ADHDer’ (H2), relative to posts featuring low levels of identity leadership. We also hypothesized that posts which referred to the influencer's status as a mental health professional would increase participants' intentions to engage in self‐care behaviours compared to those that did not (H3).

METHOD

Participants and design

We recruited participants from the same demographic group as Study 1 (N = 400; men aged 18–30 from English‐speaking countries: the United States, n = 291; the United Kingdom, n = 76; and Australia, n = 33). However, 130 failed the manipulation check and 11 failed the attention check, resulting in a final sample of 259 (M _age = 25.19, SD = 3.24).² Only one participant has missing data for self‐care intention (N = 258). This missing value was handled in analyses using listwise deletion.

This study employed a 2 × 2 experimental design that involved independently manipulating identity leadership and mental health professional status through exposure to written scripts that were presented in the format of a social media post (see Figure 5). In high identity leadership conditions, the influencer explicitly self‐identified as having ADHD, demonstrating identity prototypicality (e.g. ‘As someone with ADHD, I know this struggle firsthand’). The script further incorporated language reflecting identity entrepreneurship (e.g. ‘Many people don't realise that these aren't just bad habits. They are part of how our ADHD brain works’), identity advancement (e.g. ‘That's why I've been sharing everything I've learned about managing ADHD for people like us’) and identity impresarioship (e.g. ‘That's why I started an ADHD community where we check in weekly, share strategies, and go through ADHD research together’). The influencer's username in this condition included ‘ADHDwith’ to reinforce their in‐group identity. The low identity leadership conditions conveyed the same ADHD‐related content, but the influencer did not self‐identify as part of the ADHD group or use language aligned with identity leadership. Instead, the language they used to describe ADHD symptoms and management strategies was more general and detached (e.g. ‘Individuals with ADHD often find it easy to get distracted by random thoughts or interruptions’). To address limitations from Study 1, all visual cues and non‐focal information about the influencers (including age) were removed. Whilst this method sacrificed multimodal richness, these complementary designs ensure that each study addresses the limitations of the other and that the observed effects are not merely artefacts of a single medium.

FIGURE 5.

Post featuring influencer's identity leadership and professional status.

The manipulation of mental health professional status was achieved by framing the influencer as a mental health professional, explicitly stating their profession and presenting their advice as grounded in research evidence. To reinforce this manipulation, the influencer's username included the title ‘Dr.’ in relevant conditions.

Power analysis

A priori power analysis using G*Power was conducted for a fixed‐effects ANOVA designed to test main effects and their interaction in a 2 × 2 experimental design with four conditions (Faul et al., 2007). A conservative small effect size (f = .15) with a power of .80 was selected, corresponding to the effect size found for the test of H2 in Study 1. The G*Power results indicated that a sample size of 351 would be necessary to detect this small effect. To account for potential withdrawals, we recruited 400 participants. However, due to a greater‐than‐expected failure rate on the manipulation checks, our sample fell short of this target. Our achieved power was .67 to detect an effect size of the same magnitude as found in Study 1 for the test of H2. A sensitivity power analysis using G*Power 3.1 indicated that with the final sample of 259 participants, the design had 80% power to detect an effect of f = 0.17 in a 2 × 2 between‐subjects ANOVA.

Measured variables

The variables were the same as those used in Study 1, except for the manipulation checks, which we changed to align with the new experimental manipulation. Participants were asked: ‘Did the influencer say he has ADHD?’ and ‘Did the influencer say he is a mental health professional?’ Participants who answered either question incorrectly were excluded.

Results

Main analyses

Influence check

An independent samples t‐test confirmed the effectiveness of the identity leadership manipulation. Participants in the high identity leadership condition reported significantly higher levels of perceived identity leadership (M = 5.32, SD = 1.24) than those in the low identity leadership condition (M = 4.63, SD = 1.37); t (257) = 4.26, p < .001, d = 0.53.

Self‐care intention

Contrary to H1 and H3, a 2 × 2 ANOVA revealed no main effects of identity leadership, F(1, 254) = 0.06, p = .813 or professional status, F(1, 254) = 0.44, p = .506. The interaction between identity leadership and professional status was significant, F(1, 254) = 6.19, p = .014, 𝜂² = .024. Interestingly, pairwise mean comparisons showed that when identity leadership was low, posts authored by lay influencers (with no professional status) led to significantly greater self‐care intentions (M = 3.22, SD = 0.98) than those authored by professionals (M = 2.90, SD = 0.79), t(254) = 2.24, p = .026, d = 0.40. Or, put another way, identity leadership predicted self‐care intentions in the direction consistent with H1, but only when the influencer described himself as a mental health professional (see Figure 6).

FIGURE 6.

Interaction between identity leadership and mental health professional status on self‐care intention. Error bars represent standard error.

ADHD identification

Factorial ANOVA showed no significant main effects of identity leadership (F(1, 255) = 3.18, p = .076) or mental health professional status (F(1, 255) = 2.69, p = .102) and no significant interaction (F(1, 255) = 0.15, p = .697). Whilst the effect of identity leadership did not reach statistical significance (p = .076), the observed effect size (d = 0.22) was consistent in magnitude with the findings from Study 1 (d = 0.29).

Exploratory analyses

Mediation analyses replicated the patterns from Study 1, using 5000 bootstrapped samples. Correlations among focal variables are presented in Table 2. Across all three models in Study 2, the experimental condition accounted for 7% of the variance in the mediator. As in Study 1, all reported coefficients are standardized (β).

TABLE 2.

Correlations among focal variables in Study 2.

Variable	M (SD)	1	2	3	4	5
1. Condition	–	–	–	–	–	–
2. Identity Leadership	4.98 (1.35)	.26***	–	–	–
3. Identification as an ADHDer	4.28 (1.08)	.11	.45***	–	–	–
4. Intention to self‐care	3.07 (0.82)	−.01	.33***	.28***	–	–
5. Help‐seeking attitude	5.68 (1.17)	−.04	.29***	.30***	.34***	–

Note: Condition was coded as 0 = low‐identity leadership and 1 = high‐identity leadership; ***p < .001 two‐tailed. N = 259.

ADHD identification

Despite the non‐significant total effect (β = .21, p = .093) and direct effect (β = − .02, p = .839), there was a significant indirect effect, IE = .23, SE = .06, 95% CI [.11, .37]. This indicates that participants in the high‐identity leadership condition perceived the post to show greater identity leadership, which in turn predicted stronger identification as an ADHDer. The full model accounted for 20% of the variance in ADHD identification (R ² = .20; see Figure 7).

FIGURE 7.

Mediation model: effect of high‐ versus low‐identity leadership videos on identification as an ADHDer. All numbers are standardized regression coefficients, ***p < .001.

Self‐care intentions

Despite a non‐significant total effect (β = − .02, p = .902) and direct effect (β = − .20, p = .109), there was again a significant indirect effect, IE = .18, SE = .05, 95% CI [.09, .29]. This suggests that high‐identity leadership increased perceived identity leadership, which in turn led to higher intentions to engage in self‐care behaviours. The full model accounted for 12% of the variance in self‐care intentions (R ² = .12; see Figure 8).

FIGURE 8.

Mediation model: effect of high‐ versus low‐identity leadership videos on self‐care intentions. All numbers are standardized regression coefficients, ***p < .001.

Help‐seeking attitudes

Factorial ANOVA showed no significant main effects of identity leadership (F(1, 255) = .14, p = .706) or mental health professional status (F(1, 255) = .00, p = .952) and no significant interaction (F(1, 255) = 2.37, p = .125). For mediation effects, despite a non‐significant total effect (β = −.08, p = .539), the direct effect of condition was significant and negative (β = −.24, p = .049; see Figure 9). There was also a significant positive indirect effect, IE = .17, SE = .05, 95% CI [.07, .28]. This pattern suggests two opposing pathways: whilst exposure to high‐identity leadership content directly reduced help‐seeking attitudes, it simultaneously increased perceived identity leadership, which predicted greater help‐seeking attitudes. Such result constitutes a suppression effect and may explain the non‐significant total effect of the condition (MacKinnon et al., 2000). The full model accounted for 10% of the variance in help‐seeking attitudes (R ² = .10; see Figure 9).

FIGURE 9.

Mediation model: effect of high‐ versus low‐identity leadership videos on help‐seeking attitudes. All numbers are standardized regression coefficients, *p < .05, ***p < .001.

Discussion

Study 2 sought to replicate Study 1 and address its limitations. We did this by increasing experimental control over the manipulation of identity leadership and thereby the internal validity of our conclusions, whilst also seeking to retain ecological validity by presenting the stimuli as social media posts. Given the deliberately minimal manipulation, small effect sizes were expected and, unfortunately, Study 2 had less statistical power than planned. Whilst the identity leadership manipulation did not directly influence identification as an ADHDer or help‐seeking attitudes, the mediation analyses were consistent with Study 1 in showing an indirect effect on all three outcomes: identification as an ADHDer, self‐care intentions and help‐seeking attitudes. Importantly too, Study 2 found an interaction between identity leadership and professional status, suggesting that identity leadership principles are especially important when mental health professionals aim to promote self‐care intentions. In line with findings reported by Robertson, Cruwys, Quayle, et al. (2025), for mental health professionals, identity leadership is a key mechanism through which disclosures of lived experience (e.g. having ADHD) can lead to positive outcomes.

GENERAL DISCUSSION

The present research aimed to examine the growing power of social media influencers—often lacking clinical credentials—to shape how audiences interpret mental health symptoms. By utilizing stimuli that mirrored authentic social media environments, we extended previous identity leadership research (e.g. Steffens, Haslam, Kerschreiter, et al., 2014; Stevens et al., 2020; van Dick et al., 2018, 2021) by focusing on its role in online domains, specifically investigating its impact on viewers' diagnostic identification (i.e. identifying as an ADHDer) and self‐care intentions. This is vital as platforms like TikTok increasingly serve as unregulated sources for mental health self‐perception (Marar et al., 2019; Yeung et al., 2022).

Regarding our primary hypotheses, H1 (direct effect on intentions) was not supported as a main effect in either study. However, Study 2 revealed an interaction effect: identity leadership was vital for mental health professionals, whose messaging was effective only when employing identity‐based framing. We propose that this reflects source‐dependent differences in perceived authenticity. For professionals, identity leadership acts as a bridge that mitigates the social distance of clinical status, making them seem more trustworthy and ‘human’ (Robertson, Cruwys, Quayle, et al., 2025). This speaks to Han and Balabanis's (2024) meta‐analytic finding that relational factors, specifically perceived trustworthiness and parasocial interaction, exert a stronger influence on behavioural intentions than expertise alone. Conversely, lay influencers may not require such narratives, as audiences already perceive them to be authentic in‐group members. H2 (diagnostic identification) received partial support in Study 1: high identity leadership increased ADHD identification compared to the low identity leadership condition, but not the control. In Study 2, although the effect of identity leadership on identification with ADHD was non‐significant, the effect size and direction were similar to those observed in Study 1.

Although identity leadership's impact varied across media formats, mediation analyses replicated a robust indirect effect of the experimental condition via identity leadership perceptions across all primary outcomes. This suggests that perceived identity leadership operates as a stable mechanism influencing identification, self‐care intentions and help‐seeking attitudes, resonating with previous literature on identity leadership's capacity to mobilize group behaviours—from organizational performance to health‐related actions (Haslam et al., 2018; Steffens, Haslam, Kerschreiter, et al., 2014; van Dick et al., 2018).

Exploratory analyses in Study 2 indicated that there was a potential suppression effect for the relationship between identity leadership condition and help‐seeking attitudes. That is, the relationship between the condition and help‐seeking attitudes consisted of direct and indirect paths that were both significant but in opposite directions. Given that identification is a prerequisite for acting upon group expectations (Livingstone et al., 2011), collective language (e.g. ‘we’) could trigger psychological reactance for non‐ADHDers, potentially running counter to the positive indirect effect of perceived identity leadership on help‐seeking intentions.

Strengths and limitations

Some key strengths and limitations of this research relate to the (deliberate) trade‐off between internal and external validity across the two studies. Study 1 prioritized ecological validity using authentic videos, recognizing that this naturalistic approach introduced potential confounds with age, tone, production choices etc. This was directly addressed in Study 2, which utilized text‐based stimuli to isolate the effects of identity leadership (i.e. after removing all visual cues and non‐focal information about the influencers, including age). Whilst this method sacrificed multimodal richness, these complementary designs ensure that each study addresses the limitations of the other and that the observed effects are not merely artefacts of a single medium.

Furthermore, whilst focusing on young men provides insights into a cohort typically characterized by low engagement with self‐care (Sagar‐Ouriaghli et al., 2019), these findings may not generalize to women and other gender identities. Social media algorithms often employ gender‐based classification to curate content (European Parliament, 2023), with the consequence that different genders are exposed to divergent narratives and may respond differently to identity‐based framing. Finally, the high attrition in Study 2 due to manipulation check failures reflects the practical reality of audience attention scarcity in high‐information digital environments.

Implications and future directions

Previous research highlights identity leadership's power to increase social identification in offline domains like workplaces, sports teams and politics (Steffens, Haslam, Kerschreiter, et al., 2014; Stevens et al., 2020; van Dick et al., 2018, 2021). Extending this, we found that even a 2‐min video can engender diagnostic identification online. It is possible that influencers instinctively emulate these principles—communicating shared identity and fostering belonging—to lead audiences towards a mental health condition without clinical evidence. Whilst this can promote positive self‐care, overidentification with diagnostic labels may also lead to negative outcomes (Cruwys & Gunaseelan, 2016).

This influence is amplified by digital dynamics; correlational analyses indicate participants are more likely to enjoy and share content from influencers engaging in identity leadership. In this, the influence here is akin to that of the Sirens in The Odyssey (Homer, 2016)—luring followers towards connection, community and care, whilst also drawing them deeper into self‐diagnosis and stigmatized identities. These forces illustrate social identification as both a ‘cure’ (Jetten et al., 2012) and a ‘curse’ (Kellezi & Reicher, 2012). Whilst identity promotes resilience (Haslam et al., 2018; Jetten et al., 2017), identification with a stigmatized label can reinforce potentially maladaptive behaviours (Cruwys & Gunaseelan, 2016) and lead individuals to experience their symptoms as more severe (Levine & Reicher, 1996; Saunders, 2023). This could create a reinforcing cycle where individuals actively seek information that further entrenches their diagnostic identity.

Consequently, mental health messaging must harness collective identity whilst preserving clinical integrity—fostering belonging without pathologizing everyday experiences. Our findings suggest professionals cannot rely on expertise alone; they must engage in identity leadership to craft messages reflecting a shared sense of ‘us’. The Sirens are already singing—but unlike Odysseus, professionals must lead not by silencing, but by offering a steadier voice that counters illusion with clarity. Moving forward, research should build upon these empirical foundations by shifting from single‐exposure designs to examining the cumulative impact of sustained influence. This could utilize discrete manipulations that isolate the unique contribution of each identity leadership dimension. This would also reflect the ways in which social media engagement gains momentum through repeated identity cues over time, leveraging both its positive and negative effects.

CONCLUSION

That is the Island of the Sirens. Circe warned me to steer clear of it, for the Sirens are beautiful but deadly.

This project examined the capacity for identity leadership in social media to increase diagnostic label identification and self‐care intentions in the context of ADHD‐related content. Across two experimental studies, we found that influencers who were perceived to show identity leadership tended to foster greater self‐care intentions, mental health help‐seeking and increased identification as an ADHDer. Study 2 additionally found that identity leadership might be even more important for mental health professionals on social media.

Together, these results highlight the power of identity leadership to shape how people interpret their health‐related experiences and choose to act upon them. Our finding aligns with concerns about the growing frequency and appeal of unverified self‐diagnoses whilst also demonstrating the potential for influencers to encourage engagement in positive self‐care behaviours. As identity‐driven messaging continues to shape the digital mental health world, understanding and guiding its influence is imperative. Circe's advice to fill our ears with beeswax will not work.

AUTHOR CONTRIBUTIONS

Yuang Cheng: Conceptualization; methodology; software; validation; formal analysis; investigation; writing – original draft; writing – review and editing; funding acquisition; visualization; data curation; project administration. Tegan Cruwys: Conceptualization; methodology; supervision; writing – review and editing; resources. Alysia M. Robertson: Visualization; writing – review and editing; supervision. Mark Stevens: Writing – review and editing; supervision. S. Alexander Haslam: Writing – review and editing; supervision.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

DATA AVAILABILITY STATEMENT

Relevant data will be made available on open‐source platform once paper is published.