This is an open-access article distributed under the terms of the
This study examines the digital representation of Non-Suicidal Self-Injury (NSSI) on TikTok, with particular attention to the emergence of online communities and the communicative strategies users employ to share content while evading platform moderation. As TikTok becomes increasingly influential among adolescents and young adults, understanding how sensitive mental health topics like NSSI circulate on the platform is critical for developing effective digital health interventions. We conducted a qualitative content analysis of 400 posts referencing NSSI, collected using a mixed-method approach: 25.5% using TikTok's official API and 74.5% via the “For You” feed of a simulated account designed to mirror organic user experience. Posts were selected based on visual indicators (e.g., scars, tools), textual cues (e.g., hashtags, metaphors), and thematic references to emotional distress, recovery, or relapse. The analysis focused on user profile characteristics, linguistic strategies, and audiovisual aesthetics. Findings reveal a loosely structured yet emotionally resonant digital community characterized by subcultural codes, such as euphemisms, ambiguous hashtags, and stylized imagery. Despite content moderation policies, most accounts remained active and visible, with minimal enforcement of warnings or restrictions. While some posts portray NSSI as a coping strategy or seek to normalize the behavior, others subtly encourage recovery or offer indirect support. However, explicit messaging that discourages self-harm is notably rare. These dynamics suggest that TikTok unintentionally enables both the concealment and dissemination of self-harm-related content, functioning as a space for affective connection but also as a vector for potential normalization of harmful behaviors. The study underscores the need for targeted, ethically grounded prevention strategies that address not only the psychological functions of these communities but also the algorithmic infrastructures that sustain their visibility. These findings contribute to ongoing debates about digital mental health, platform responsibility, and the design of safer online environments.
Non-Suicidal Self-Injury [NSSI], defined as the deliberate, self-inflicted damage of body tissue without suicidal intent and not for purposes that are socially accepted, is a significant health concern among adolescents and young adults (
TikTok presents a distinctive appeal linked to its participatory and algorithm-driven environment, which encourages rapid content sharing, replication, and engagement. Over 80% of creators are under 25 (
There is a growing body of literature documenting the complex relationship between social media use and youth mental health. Some studies have reported negative outcomes linked to excessive screen time and harmful content exposure, such as depressive symptoms, anxiety, and problematic Internet use (
Recent findings underscore how the algorithm TikTok uses, specifically its “For You Page” (FYP), can intensify both positive and negative experiences. It offers unique opportunities not available in other online communities, fostering self-discovery and visibility, but also overexposes users to emotionally charged or triggering content (
Studies show that adolescents turn to online communities to connect with people with similar interests (
Although some creators may use TikTok to promote resilience, share recovery journeys, or encourage professional help-seeking, others risk blurring the line between raising awareness and glamorizing NSSI. Most adolescents maintain positive expectations of social media, even when they have already experienced negative consequences (
This study focuses on TikTok to explore the representation of Non-Suicidal Self-Injury [NSSI] within this unique online social environment. TikTok was selected as a case study because of its massive youth user base ( RQ1) What are the characteristics of TikTok profiles that post NSSI content? RQ2) What strategies do users employ to present NSSI content and build a community around it? RQ3) How do NSSI content creators circumvent moderation systems in TikTok?
It must be noted that TikTok's official policy prohibits content that promotes, glorifies, or normalizes Non-Suicidal Self-Injury [NSSI]. According to its Safety & Privacy Guidelines (
A qualitative content analysis methodology was employed to explore how Non-Suicidal Self-Injury (NSSI) is communicated and represented on TikTok. The analysis followed a multi-step process inspired by thematic analysis and digital ethnography. First, an open coding phase was conducted to identify recurring patterns in visual, textual, and symbolic elements. These initial codes were then grouped into broader thematic categories through axial coding. Finally, interpretive analysis was applied to understand the socio-cultural meanings and community dynamics embedded in the content. While descriptive statistics (e.g., percentages) were used to illustrate the frequency of certain features, the primary aim was to uncover the communicative logic and affective structures underpinning NSSI-related content on the platform.
Given the nature of the platform, each TikTok post was treated as a unit of analysis, including the audiovisual content (e.g., music, voice-over, visual elements) and the accompanying textual content (e.g., captions, hashtags).
A mixed sampling strategy was employed to reflect both user-directed (search-based) and platform-directed (algorithmic) pathways for discovering content related to NSSI. Data collection took place between January and April 2024.
The first dataset was retrieved using the TikTok API, accessed via the TikApi library (
To complement search-based results and account for TikTok's algorithmic recommendation dynamics, a simulated user account (@aiko3135) was created. This account was designed to replicate the behavioral patterns of an adolescent user with an emerging interest in NSSI-related content. The simulated account was designed to reflect the behavior of adolescents and young adults, who represent the majority of TikTok's user base. Although demographic data of content creators was not always explicit, profile indicators (e.g., language, imagery, self-descriptions) suggest that most users in the sample belong to this age group.
The simulation protocol involved passive scrolling, “liking” and saving NSSI-related videos over several sessions to train the recommendation algorithm. This approach aligns with prior research that leverages synthetic accounts to study how social media algorithms surface sensitive content (
The simulated account was active for a total of 90 days, during which it interacted with NSSI-related content to train the algorithm. This process resulted in the identification of approximately 1,200 posts, from which 298 were selected based on the inclusion criteria. Combined with the API-based sample, the final dataset consisted of 400 posts.
Posts were included if they contained explicit or implicit references to NSSI, operationalized as visual indicators (e.g., scars, fresh wounds, tools associated with self-injury), textual cues in captions or hashtags, thematic references (e.g., emotional pain, self-harm, recovery, relapse). Duplicates, off-topic content, or videos lacking sufficient contextual information were excluded (see
Inclusion and exclusion criteria.
| Inclusion criteria | Exclusion criteria |
|---|---|
| Posts published on TikTok | Content from other platforms (e.g., Instagram, X.com, YouTube) |
| Videos in English or Spanish | Videos in other languages not understood by research team |
| Posts with explicit or implicit references to non-suicidal self-injury (NSSI), including: Visual indicators, textual references in captions, hashtags, or overlays, and thematic references to emotional distress or coping through self-injury. | Posts unrelated to NSSI or containing exclusively suicidal content without self-harm references. |
| Original, edited, or repurposed videos created by users, containing visual, textual, or symbolic representations related to NSSI. | Content unrelated to the topic (e.g., lifestyle, humor, beauty tutorials). |
| Public individual user accounts. | Institutional, commercial, or private accounts. |
| Posts accessible at the time of collection and sufficiently contextualized for analysis. | Deleted, suspended, private, or context-deficient posts (e.g., no caption, visuals, or relevant hashtags). |
Following this criteria, the final sample consisted of 400 posts, of which 25.5% were collected via the API and 74.5% via the simulated account's “For You” feed.
A structured codebook of 103 variables was developed to guide the qualitative content analysis. These variables were derived from a comprehensive review of prior research on NSSI communication, mental health in social media, and media representation of self-harm ( Visual content: presence of scars, blood, or tools. Textual markers: use of recovery vs. pro-NSSI language. Narrative tone: expressions of despair, romanticization, normalization, or encouragement of self-injury. Affective cues: expressions of emotional pain, isolation, or solidarity.
Coding was performed by two trained researchers following a two-phase intercoder training protocol to ensure consistency and minimize subjectivity (
The combined use of data mining and algorithmic exposure allowed for the identification of digital communities and content circulation patterns that reflect how TikTok users engage with NSSI-related material. As Airoldi (
Each post was saved as a video file and accompanied by a metadata sheet including the caption, hashtags, and user profile information available. Screenshots were captured when relevant to facilitate visual analysis. Transcripts were created for videos with spoken or written text, and all materials were stored in a secure, encrypted institutional server. This transformation of data into analyzable formats allowed for coding and thematic interpretation.
All information and metadata analyzed were obtained from public TikTok profiles and did not include any personally identifiable information. The data were securely stored on the servers of the leading author institution and encrypted to ensure that only the researchers involved in the data analysis had access. This approach aligns with the ethical guidelines for conducting research on publicly accessible online content while safeguarding user privacy (
This section presents the findings from the qualitative content analysis, organized into three main areas: profile characteristics (RQ1), textual analysis, and audiovisual analysis of the identified content. Across these dimensions, we highlight how TikTok users create and share NSSI content (RQ2) while concealing it from platform moderation (RQ3).
Accounts were identified as related to NSSI through a process combining keyword detection (e.g., references to “self-harm” or associated metaphors in bios or captions), qualitative indicators (e.g., use of metaphors such as “beans”, “barcodes”, or red color codes), and the presence of visual content displaying either fresh or healed wounds. While the academic literature uses NSSI as a descriptor for this type of content, users themselves seem to employ the term “self-harm” [SH]. Therefore, we adopted the term SH when describing user-generated content as a synonym for NSSI.
Based on the classification, 53% of the analyzed posts originated from profiles explicitly centered on SH-related content, whereas 44.75% came from profiles that were not framed as such but regularly posted SH material.
Although TikTok's community guidelines restrict SH-related content, 97.5% of the analyzed accounts were still active at the time of data collection, suggesting that much of this content successfully circumvented moderation protocols and control filters.
Regarding the identity of users who posted the analyzed content, 53.75% employed anonymous or fictional profile images, while 43.5% used personal photos. Based on images and available profile data, 44.5% of users presented as female, 11.5% as male, and 41.5% were unidentifiable in terms of gender expression or omitted gendered references entirely. The language distribution across profiles was as follows: 52% in English, 42.5% in Spanish, and 2.75% in other languages.
Users strategically employed indirect, coded, or euphemistic language. None of the analyzed posts carried TikTok-issued sensitive content notifications. This may be related to the indirect or coded language that users employ. For instance, 82.2% of the post titles made no mention of mental illness or disorder, and explicit autobiographical references were relatively rare, with only 25.8% referencing their own experience in the title and 28.4% doing so in the post text. Instead, users often adopt generic or impersonal language, such as second or third-person phrasing. References to SH were frequently metaphorical or euphemistic (e.g., “tiger stripes” to refer to scars), which may serve to avoid detection and signal group membership. Only 4.2% referred to wounds as real in the title, while 95.8% did not mention wounds at all.
Emojis played a symbolic role in framing the content, adding an emotional and aesthetic layer. Emojis were included in 19.3% of the content analyzed in the title. Symbols that could be interpreted as “love” (hearts, kisses) were the most common (8%), followed by emojis associated with happiness (4%), animals (3%), and sadness (less than 1%). There even seems to be a positive reinterpretation not only by using these emojis but also by the accompanying narrative and hashtags. For example, one post featured stars drawn around scars with hashtags such as #drewstarsaroundmyscars and #malikandlikay, framing the content as affectionate or gentle rather than alarming.
Hashtag strategies play a critical role in how NSSI-related content circulates on TikTok. Specific coded expressions or algospeak, such as “tiger stripes”, appear to serve dual functions: they enhance the discoverability of content for in-group users while circumventing moderation mechanisms that could lead to content removal. Additionally, 55.5% of the analyzed posts included direct prompts for engagement (e.g., “follow me”), suggesting a strong emphasis on community-building and social connectivity. This type of content often goes beyond algorithmic visibility or network logic, reflecting deeper social dynamics. For example, some posts combine insider terminology with positive emotional cues (such as depictions of users expressing joy or relief upon recognizing others with visible signs of self-injury) thereby reinforcing mutual identification and emotional solidarity within the self-harm community.
The vast majority of content successfully bypassed TikTok's moderation protocols. Notably, none of the analyzed posts displayed platform-generated sensitivity warnings, despite the presence of SH-related themes.
Most of the analyzed videos (85.3%) were original content recorded by the users themselves, predominantly featuring real people. The remaining 14.8% used clips from fictional sources, such as TV shows or films, and 20% repurposed video fragments from other users.
Aesthetic or fictionalized alternatives were sometimes used to reference SH practices to avoid sharing real content, including clips from video games or anime (13.3% of videos), or stylized content with a cute (4.5%) aesthetic characterized by themes involving teddy bears, kittens, hearts, and pastel colors. Other minoritarian tactics employed to communicate on TikTok while evading controls are “Asian-inspired” themes (2.3%) and the use of memes (1.2%).
Text overlays were a dominant feature (90.2%). Of these, 33.1% used red-colored text, a practice often identified by users themselves with hashtags like #redtexttoidentify. These texts were written from a first-person perspective in 63.4% of the videos, and 74% explicitly described alleged personal experiences of the poster.
Despite the sensitive nature of the content, very few videos contained graphic imagery. Only 0.8% showed blood, but scars were more common, as 18% of videos showed healed wounds. Regarding written mentions of wounds in the videos, 59.2% referred to self-harm in some manner. Tools such as blades, lighters, or pencil sharpeners appeared in 13.5% of the videos, and in 3.1% of the videos, they were romanticized using filters, music, or symbolic captions. In 12.5% of the videos, there was a positive attitude toward wounds. Conversely, a smaller number of posts (8%) featured recovery-oriented visuals, such as screenshots from apps and graphs tracking days without SH. Captures are usually taken from apps originally used for addictive behaviors.
In summary, visual and symbolic strategies show a balance between self-expression, peer engagement, and strategic concealment, often using platform-native tools (e.g., aesthetics, hashtags, text overlays) to create a subcultural language that signals belonging while evading censorship.
An emergent theme identified during data analysis was the framing of self-harm (SH) as a coping mechanism or tool for emotional regulation. Several posts explicitly described SH as helpful during moments of emotional distress, with some users defending past episodes of self-injury as functional or necessary in the context in which they occurred. These narratives were often met with sympathetic responses from other users, who contributed their own experiences in the comments, expressing empathy or identification with the emotional state described in the original post.
Quantitatively, 10.9% of the posts contained explicit statements that SH helped the user feel better, while an additional 10.9% referenced either mood disturbances or the physical pain associated with self-harming. A smaller but significant portion of users (3.9%) used the platform to seek advice or opinions from others regarding SH, sometimes posing direct questions to content creators or the broader community. This pattern suggests that beyond its expressive function, NSSI content on TikTok may serve as a peer-to-peer support space where users seek validation, normalization, or shared coping strategies.
Regarding mentions of ending self-harming behavior, 12.5% of the analyzed posts discussed desistance from the practice, typically framed as an aspiration or ongoing process. However, only 2.5% of the posts actively encouraged avoiding SH. Most content did not clearly endorse or condemn the behavior, but the messages tended to convey emotional pain instead. In fact, 89.3% of the titles contained no positive framing of SH, and 92% did not express any negative evaluations either.
Requests for help were rare, as only 1.3% of posts explicitly mentioned seeking support. However, social references were frequent (for example, 25.3% mentioned friends or family in some way). Some users directly addressed TikTok's moderation in captions like “block don't report”, “no flop”, or “TikTok no me banees” (“TikTok don't ban me”), suggesting awareness of the policies and trying to discourage others from reporting their sensitive content, requesting that those who are uncomfortable with their content block them instead of reporting them. Other posts illustrated the tension between visibility and stigma. For example, a video entitled ‘POV: you've got SH scars in summer' showed the anticipated judgement of others for wearing fashionable jeans with ripped holes showing the scars of self-harm, but ended with a message of empowerment: ‘wear what you want to wear’.
Summary of communicative strategies in NSSI-related TikTok posts.
| Category | Strategy | Example/Frequency |
|---|---|---|
| Textual strategies | Euphemisms | e.g., “tiger stripes” for scars |
| Hashtags | e.g., #beanstwt, #shscars, #redtexttoidentify | |
| Indirect language | 82.2% did not mention mental illness explicitly | |
| Autobiographical references | 25.8% in title, 28.4% in post text | |
| First-person perspective | 63.4% of videos | |
| Visual strategies | Scars | 18% of videos showed healed wounds |
| Blood | 0.8% of videos | |
| Tools | 13.5% showed blades, lighters, etc. | |
| Red text overlays | 33.1% of videos | |
| Cute aesthetic | 4.5% of videos | |
| Anime/Game clips | 13.3% of videos | |
| Engagement strategies | Calls to follow | 55.5% of posts |
| Comments for empathy | Users shared experiences and support | |
| Avoiding moderation | Captions like “TikTok don't ban me” | |
| Recovery visuals | 8% of posts showed sobriety apps |
This table categorizes the main communicative strategies identified in the qualitative analysis of 400 TikTok posts related to Non-Suicidal Self-Injury (NSSI). It includes textual, visual, and engagement strategies used by users to express emotional distress, foster community, and evade platform moderation. Frequencies are based on the coded sample and illustrate the prevalence of specific tactics such as euphemisms, red text overlays, and calls for interaction.
This study examined TikTok posts related to Non-Suicidal Self-Injury (NSSI), focusing on whether a digital community forms self-harm practices and the communicative strategies used by content creators not only to engage with others and also avoid platform moderation. The results suggest that despite platform guidelines (
A considerable portion of posts (73%) featured real individuals, though explicit images of self-harm were rare. Instead, creators utilized symbolic or euphemistic codes, such as terms like “tiger stripes” or visual motifs with a “cute aesthetic” to signal distress in ways recognizable to those familiar with the subculture. These findings align with prior research documenting platform-specific forms of communication that allow users to maintain visibility while circumventing content removal (
Regarding the characteristics of TikTok profiles posting NSSI content (RQ1), our findings indicate that some users actively construct a digital identity centered on self-harm. These users often reference NSSI in their bios and describe their accounts as spaces for emotional expression or sharing personal struggles. This behavior aligns with prior research showing that NSSI can serve as a coping mechanism for emotional regulation (
Based on the results, 53% of the analyzed posts originated from profiles explicitly centered on SH-related content, often referencing emotional relief. This pattern reflects psychological models that conceptualize NSSI as a maladaptive yet functional strategy for short-term emotional regulation (
Regarding RQ2 (What strategies do users employ to present NSSI content and build community?), our data reveal a consistent use of coded language and shared symbols to foster group recognition and cohesion. Hashtags like #tigerstripes, #beanstwt, or #shscars act as in-group markers that signal community affiliation. These practices help establish shared norms and facilitate affective bonding, consistent with theories of online subcultures (
Interestingly, RQ2 and RQ3 (How do NSSI content creators circumvent moderation systems in TikTok?) have at least one shared answer. Users appear highly aware of moderation practices and tailor their content accordingly, avoiding explicit language or images. The use of a distinct communicative style, often referred to as algospeak (
The symbolic and indirect nature presentation of content suggests a deliberate strategy to bypass content moderation and aligns with prior observations of users creatively reframing mental health content on TikTok (
On the other hand, our data suggest that the algorithm reinforces exposure to NSSI-related content once a user begins interacting with it. After the research account engaged with SH-related posts by liking or saving them, its feed began to show more similar content. This pattern supports previous research indicating that recommendation systems can unintentionally promote harmful content through engagement loops (
These findings contribute to the growing literature highlighting the double-edged nature of online peer-to-peer mental health content. On the one hand, NSSI communities may provide a safe space for emotional expression and peer connection. On the other, they may also normalize maladaptive coping strategies or create echo chambers that discourage proffesional help-seeking (
Despite TikTok's official Well-Being Guides and partnerships with mental health organizations, our findings echo earlier concerns about the limited impact of institutional accounts (
Regarding the limitations of this study, the use of a simulated account introduces certain biases, as algorithmic exposure may not fully reflect the experience of real users. The geolocation of the account in Spain likely influenced content language and cultural framing, limiting the generalizability of findings to broader TikTok communities. Furthermore, the performative nature of online identity complicates the interpretation of intent behind posts, and without user interviews, conclusions about motivations remain tentative. This study is limited by its reliance on a simulated account, which may not fully replicate the experience of real users. The lack of engagement with actual users further restricts our ability to assess the nuances behind content creation and sharing or infer the intent of creators. While a simulated account provides valuable observational insights, it cannot capture the subjective experiences or emotional states of those posting NSSI-related material.
Despite these limitations, the study offers important insights for digital mental health interventions and platform governance. The findings highlight the need for moderation systems that can recognize and respond to subcultural codes, as well as for public health messaging that aligns with the aesthetic and communicative preferences of youth communities online. Future research should move beyond content-analysis to center lived experiences and investigate how these online interactions influence offline well-being, particularly among young people at heightened risk of self-injury. Additionally, the influence of these communities on help-seeking behavior and long-term mental health outcomes should be examined, as well as platform-level interventions that balance freedom of expression with user safety.
This study investigated how self-harm is represented on TikTok, revealing a loosely formed yet affectively resonant digital community in which users construct identities rooted in distress and coping. Leveraging specific hashtags, subcultural aesthetics, and semi-coded language like metaphors or red-tinted text, creators can generate group belonging while evading moderation. References to recovery tools (i. e., apps to track sober behavior) further indicate a user base simultaneously aware of risks and invested in navigating them. Algorithmic exposure patterns suggest that engaging with NSSI content leads to increased visibility, further revealing the dual role of TikTok as both a site of peer support and a space where maladaptive behaviors can be subtly reinforced. Recommendation systems and moderation policies must adapt to the coded vernacular of NSSI communities.
Overall, while this digital community may provide validation, emotional release, and a sense of belonging for some users, it also normalizes self-injurious behaviors and raises complex questions for prevention efforts. Effective interventions must be grounded in an understanding of platform-specific cultures and the affective needs these communities fulfill, particularly among young people. Addressing this challenge requires interdisciplinary collaboration across youth psychology, communication and media studies, platform governance, and online-community research to develop empathetic, evidence-based digital-mental-health strategies. Future work should explore how digital platforms can respond more effectively and how public health strategies can become responsive to platform-specific cultures and communicative practices.
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
This project was approved by the Ethics Committee of the "Universidad Rey Juan Carlos". The social media data was accessed and analyzed in accordance with the platforms terms of use and all relevant institutional/national regulations.
EM-P: Funding acquisition, Formal analysis, Methodology, Writing – original draft, Resources, Supervision, Conceptualization, Validation, Investigation, Project administration, Writing – review & editing, Data curation. MB-R: Formal analysis, Writing – original draft, Investigation, Data curation, Methodology, Writing – review & editing. SF: Writing – original draft, Formal analysis, Methodology, Data curation, Writing – review & editing.
The author(s) declare that financial support was received for the research and/or publication of this article. This article has been funded by the project: Media Representation of Self-Harm in Minors in the Media and Social Networks (PID2021-124550OB-I00) of the Ministry of Science and Innovation (Spain).
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The author(s) declare that no Generative AI was used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.