Study urges safer social media controls and supportive spaces against self-harm

Social media policies designed to reduce exposure to self-harm and suicide content may also remove some of the online spaces that users rely on for recovery, connection and crisis support, according to a new study led by researchers at Swansea University’s National Centre for Suicide Prevention and Self-Harm Research. The findings arrive amid growing international pressure to restrict social media access for children and young people, including age-based bans and stronger platform controls. Researchers say the evidence available to guide these policies remains limited, and that poorly designed restrictions could produce unintended consequences.

The study is described as the first to examine the views of people with lived experience of self-harm on the safety features and content policies used by social media platforms. Researchers combined responses from a survey of more than 5,000 people with detailed interviews involving 17 participants. Most interviewees had a personal history of self-harm, and all had viewed self-harm-related material online. The research explored how participants experienced content restrictions and safety messages introduced by platforms in 2019, as well as their views on moderation tools, warnings and user controls.

The results suggest that restrictions can improve perceived safety while simultaneously changing how people access support. Half of survey respondents said that additional restrictions made social media safer. However, 34.9 per cent said the changes affected what they saw, while 9.9 per cent said they changed what they posted. These figures indicate that platform policies do not simply remove harmful material; they also alter users’ behaviour and the information environments in which conversations about self-harm take place.

A substantial majority—79.4 per cent—said restrictions could make people feel bad or isolated. During interviews, participants described how moderation systems sometimes censored recovery stories, discussions about coping and images showing self-harm scars. Such material may be detected by automated content-moderation systems because of visual or linguistic signals associated with self-harm, even when the post’s purpose is educational, supportive or focused on recovery. Participants said that the removal of these posts could suppress constructive narratives and make it harder for people to find communities where they feel understood.

Content warnings produced a more complex response. More than 90 per cent of respondents said they would be likely to click on a post carrying a generic warning, compared with only 40 per cent who said they would view a post marked with a self-harm-specific warning. At the same time, participants reported that specific warnings helped them make informed choices about whether to continue. The findings suggest that warnings can function as an autonomy tool when they provide meaningful information without automatically preventing access.

The survey also found that 21.1 per cent of respondents had experienced a post being censored or removed because self-harm scars were visible. Ninety per cent of those who reported this experience described it as harmful. Similar concerns emerged around the removal of recovery stories. From a technical perspective, these outcomes highlight the difficulty of using automated moderation to distinguish between content that promotes harmful behaviour and content that documents survival, recovery or support. A system that prioritises removal based on keywords or images may reduce exposure to triggering material, but it can also produce false positives that silence people seeking help.

Personal control emerged as one of the strongest themes in the research. Some 87.8 per cent of respondents wanted more control over the content they see. Researchers say this could include simple, intuitive settings that allow users to filter topics, adjust warning preferences or decide whether potentially distressing material appears in their feeds. Such tools would shift some responsibility from broad, platform-wide censorship toward user agency, although the study suggests that controls must be easy to understand and practical to use.

Participants also expressed support for stronger protections for younger users. Rather than relying exclusively on blanket bans, respondents suggested that platforms could provide age-appropriate versions with more limited features, tailored safety settings and stricter defaults. The researchers argue that this approach could reduce exposure to harmful material while preserving access to legitimate communication and support. They also warn that bans may push young people toward less visible online spaces, where harmful content may be harder for families, researchers and services to identify.

Dr Amanda Marchant, the study’s first author, said restrictions could reduce harmful content but might also unintentionally limit recovery narratives and community support. She called for continuous evaluation of online safety measures and for platforms to use social media more effectively for signposting, stigma reduction and access to supportive communities. Professor Ann John, senior author and director of the National Centre for Suicide Prevention and Self-Harm Research, said evidence surrounding social media bans for under-16s remains uncertain. The researchers recommend that new policies and interventions be co-designed with young people and people with lived experience, then tested and evaluated before being introduced at scale.

The study’s conclusions point toward a more targeted model of online safety: one that combines risk reduction with transparency, user choice and evidence-based design. Researchers say safety features should be assessed not only by how much harmful content they remove, but also by whether they preserve access to recovery information and supportive relationships. As governments and technology companies debate tighter regulation, the findings suggest that the most effective systems may be those that reduce harm without making vulnerable users invisible.

Web References: https://popdatasci.swan.ac.uk/centres-of-excellence/ncsr/ ; https://www.jmir.org/2026/1/e73343/ ; https://www.swansea.ac.uk/staff/a.john/

References: Journal of Medical Internet Research, DOI: 10.2196/73343