Offline Factors Influencing the Online Safety of Adolescents with Family Vulnerabilities

Abstract

Online safety guidance is frequently delivered as a specialist technology issue without considering adolescents’ home lives, offline vulnerabilities, or wellbeing. Yet, while the digital world offers connection, autonomy, and entertainment, vulnerable teens also encounter more violent content, sexual exploitation, and content concerning body image, self-harm or suicide than their non-vulnerable peers. Many struggle with social inclusion or less engaged and credible caregiver e-safety support, which may contribute to their negative experiences online. To improve their online safety and resilience, caregivers and educators might consider offline factors that can mediate exposure to online harms. This study compared the experiences of 213 adolescents with family vulnerabilities to 213 age- and gender-matched non-vulnerable adolescents. The contribution of (a) e-safety education, (b) close friendships, (c) a trusted adult at school, and (d) life-affecting worry was considered. No differences were found for exposure to, or engagement with, e-safety education. However, despite having received e-safety education, those with family vulnerabilities were more at risk of encountering severe online harms. This was mediated by life-affecting worry and parental e-safety guidance. These findings provide unique insights into the impact of family vulnerabilities on adolescents’ exposure to online harms and suggest a more holistic intervention framework for caregivers.

1. Introduction

Despite adults’ efforts to restrict children’s Internet use, almost all UK children are active online (Ofcom 2024). Moreover, despite platforms’ minimum age requirements and the introduction of the Online Safety Act 2023, 75% of children aged 8–17 have their own social media accounts (Ofcom 2024). Although it is a source of social connectivity, entertainment, learning, and autonomy for young people, the Internet also has the potential to facilitate exposure to risks and harms.

Livingstone et al. (as cited in Livingstone and Stoilova 2021) defined these as the ‘four Cs’ to conceptualise the high-risk online experiences that children may encounter. These include seeing harmful content, such as pro-anorexia or pro-suicide content, or violent or graphic images; being targeted by harmful contact, such as cyberbullying, sending nudes, and sexting; witnessing or being involved in conduct risks, such as gambling or visiting adult pages; and contract—also referred to by others as cyberscams—such as being tricked into buying fake goods, or hacking.

Although these are potential harms for all young Internet users, some appear to be at increased risk. These include adolescents who experience offline challenges—such as family vulnerabilities, communication difficulties, physical disabilities, Special Educational Needs (SENs), or mental health difficulties. These individuals are significantly more likely to encounter all types of online harm (Brett et al. 2024; El Asam and Katz 2018; El Asam et al. 2022; Katz and El Asam 2020; Livingstone and Smith 2014; Whittle et al. 2013). Whilst all vulnerable groups warrant attention, those experiencing family vulnerabilities—such as adolescents in social care, young carers, and young people experiencing high levels of family stress—are often understudied.

1.1. Family Vulnerabilities

The presence of a supportive and stable family bond can be protective against various online harms, including cyberbullying (Brett et al. 2024), grooming (Whittle et al. 2013), and radicalisation (Lösel et al. 2018). However, for children with family vulnerabilities, these bonds may be less stable and characterised by negativity.

In 2020, there were approximately 80,850 Looked After Children in the UK (Department for Education 2021), which increased to 83,840 by 2024 (Department for Education 2024). These children frequently describe relationships with others that are characterised by distrust and instability, leading to greater social isolation (Hong et al. 2021). School attendance may be reduced (Children’s Commissioner for England 2018), with as many as 11% of Looked After Children experiencing high placement instability, moving three or more times (Department for Education 2024; Salazar 2013). Furthermore, teachers may pre-label these children as difficult or problematic for being in the social care system (Vacca and Kramer-Vida 2012), perpetuating stigma and ostracising them further. It is therefore unsurprising that many children in care report being bullied by peers just for being in care (Dansey et al. 2019; Rogers 2017).

Moreover, strained relationships are also reported by young carers (Dharampal and Ani 2020): Young carers are described as children under the age of 18, who provide care or support to a family member on a regular basis and hold a level of responsibility that is usually placed on an adult (Becker 2000). Although exact figures are unknown, recent reports suggest that approximately 7% of 11–15-year-olds in the UK provide ‘high’ levels of care to a family member and 3% provide a ‘very high’ level (Joseph et al. 2019). These children report having less time and opportunity to maintain friendships and a need for peers who recognise their responsibilities (Becker and Becker 2008).

A final group of young people characterised as having a family vulnerability are those who report high levels of concern or stress about their family life. The source of this may differ; in 2018, the Children’s Commissioner’s Office estimated that there were over 2.1 million children in England living in families with complex needs, such as domestic abuse, parental substance abuse, and mental health problems (Children’s Commissioner for England 2018). Additionally, the COVID-19 pandemic saw many of these exacerbated (Calvano et al. 2022), alongside large numbers of families being driven into poverty (The Legatum Institute 2020). These children are at an increased risk of subsequent victimisation (Duncan 1999).

Alongside the aforementioned strains on relationships, children with family vulnerabilities are also at an increased risk of experiencing additional challenges. These include poor mental health (Cree 2003; Dubois-Comtois et al. 2021; Engler et al. 2022; Lewis et al. 2023), SENs (Stanley 2012), and language and communication barriers (Hounsell 2013). Each of these renders them more at risk of online harm than those without vulnerabilities. Moreover, El Asam et al. (2022) note that psychological distress can increase the risk of encountering online harm among vulnerable adolescents.

1.2. Relationships and Resilience

Interpersonal relationships are essential for promoting resilience, with supportive networks encouraging positive outcomes for those with vulnerabilities (Bronfenbrenner 2005). In particular, Hammond et al. (2024) demonstrate that digital resilience is rooted in a socio-ecological framework, which assumes that resilience—defined as the ability to manage and recover from negative online experiences—is embedded within individual, family, community and societal factors. Each of these networks is vital in supporting children and young people, yet those with family vulnerabilities already experience problems or a breakdown in their interpersonal relationships, as previously discussed. They may spend more time on the Internet and social media to seek refuge or cope with their loneliness (Cauberghe et al. 2021; O’Day and Heimberg 2021). The charity Barnardos suggests that the increased time spent online since the COVID-19 pandemic is increasing children’s risk of exploitation online (Barnardos 2020).

In addition to spending increased time online, due to weaker social bonds, those living with family vulnerabilities may find that their online experiences are not adequately addressed by a basic online safety approach. These children require a more targeted and intensive form of support that not all services are equipped to offer (Badillo-Urquiola et al. 2024). Offline factors, such as engaged supportive parenting, e-safety support, and relationships with peers, can play a role in protecting children against online harms for those with SENs (El Asam et al. 2023). It is likely that these factors, and the presence of trusted adults at school and positive relationships within the family, will also lessen the harm experienced by those with family vulnerabilities.

1.3. Severe Online Harms

Whilst it is established that vulnerable children are at risk of encountering a ‘basket’ of online harms (El Asam and Katz 2018; El Asam et al. 2022; Katz and El Asam 2020; Livingstone and Smith 2014), not all of these harms carry the same impact risk. For instance, children encountering certain harms (e.g., grooming or viewing pro-anorexia or pro-suicide content) may be at a more immediate and sustained risk than other harms (e.g., buying fake goods); we argue that attention should be paid to the experiences of vulnerable children regarding some of the most severe online harms (SOHs), to better understand their experiences and any antecedents.

Adults do not always view SOHs in the same way as adolescents. Jiang et al. (2021) investigated perceptions of severity of different online harms. Although perceptions differed among the eight countries studied, certain content was consistently reported to be in the top half of ranks, including self-harm and suicide promotion, violent content, and child nudity. These are all harms that young people may experience when using the Internet and social media. Similarly, Boyd and Hargittai (2013) found that parents were most frequently concerned about their children meeting strangers online: in 2020, approximately 17% of children aged 10–15 spoke to strangers online, and 5% of all children met up with strangers in person (Office for National Statistics 2021). To our knowledge, existing research has not specifically explored SOHs, which this paper endeavours to address.

1.4. Aims

The extraordinary pace of transformational change brought about by technology demands nuanced and informed approaches to support safety and resilience among vulnerable children online. This can be a challenge for adults who act to keep children safe. Children’s services can struggle to integrate monitoring of case data, updating referral systems, and staff training (El Asam et al. 2021). Training for professionals—where it exists—can struggle to adapt fast enough, with practitioners who work with young people describing a ‘generational gap’ (Quayle et al. 2023). In addition, more than two thirds of psychiatrists report not feeling sufficiently prepared to assess digital risk, despite half of these treating patients exposed to these risks (Aref-Adib et al. 2020).

With the aim of providing insights for frontline practice, this study considers offline factors that could contribute to building online resilience in adolescents experiencing family vulnerabilities, including engaged supportive parenting, credible online safety advice, good quality offline friendships, and having a trusted adult at school. It considers the role of constant worry about life at home (life-affecting worry). Responses were collected from 12 schools in England in December 2020. Data from 426 adolescents was analysed, with 213 having a family vulnerability, and 213 age- and gender-matched controls reporting no vulnerabilities. Participants were asked about their experiences of e-safety education, alongside their interpersonal relationships, life-affecting worry, and SOHs encountered.

It was hypothesised that adolescents with family vulnerabilities would receive lower rates of e-safety education than non-vulnerable adolescents, but it was unclear if and how the sources of e-safety education would differ. It was further hypothesised that those with family vulnerabilities would be more at risk of encountering SOHs, alongside reporting poorer relationships with friends and family and less trust in adults at school; these interpersonal relationships would mediate the association between vulnerability and SOHs. Finally, it was hypothesised that adolescents with family vulnerabilities would report higher levels of life-affecting worry compared to their non-vulnerable peers.

2. Methods

This study utilised data from ‘The Cybersurvey 2020’, which is an annual survey used to explore young people’s online experiences. Each year focuses on a different theme, with the 2020 survey exploring digital lives during the COVID-19 pandemic.

2.1. Participants

Data was collected in the Winter Term (December 2020) in the backdrop of the COVID-19 pandemic, when much of England was experiencing a compulsory lockdown and many young people were attending school virtually. An opportunity sampling technique was adopted, with schools and colleges being invited to participate. Data was collected from a total of 2033 young people from 12 primary schools, secondary schools, and colleges across England; this paper focuses on a subsection of these respondents.

Data was included from a total of 426 adolescents: 213 self-identified as having a family vulnerability (living in or having lived in social care, being a carer for a family member, or being worried about family life at home), and a sample of 213 age- and gender-matched controls, who were randomly selected from those who reported not having any vulnerabilities. To achieve this, the IBM SPSS Statistics 27 ‘Select Random Sample’ function was used, allowing for randomisation in the participants selected from an age- and gender-matched pool. Notably, many of those with family vulnerabilities also reported experiencing other vulnerabilities: 47.9% experienced emotional and mental health vulnerabilities (n = 102); 21.6% experienced COVID-related vulnerabilities, feeling that the pandemic had a substantial impact on them or their family (n = 46); 18.3% experienced communication vulnerabilities (n = 39); 16.9% experienced physical vulnerabilities (n = 36); and 15.0% had SENs (n = 32). Those with family vulnerabilities will hereafter be referred to as ‘FV’, whilst the non-vulnerable adolescents as ‘NVAs’.

The average age of participants was 13.35 years (SD = 1.76); 176 participants identified as males, 222 identified as females, and 28 declined to specify their gender.

2.2. Materials

Details on the construction of the Cybersurvey have been reported elsewhere (El Asam and Katz 2018). Schools provided students with a link to the Cybersurvey to complete during their virtual and in-person classes: the survey comprises 34 questions exploring young people’s social support and friendships, time spent online and using mobile phones, and their encounters with online harms and e-safety education. The variables relevant to this study are outlined below. The individual items related to each variable are outlined in Appendix A, alongside an indication of any reverse coding.

2.3. E-Safety Education

2.3.1. Who Provides E-Safety Education?

Participants were asked to identify who—if anyone—taught them how to stay safe online. Six possible sources were provided, and participants were asked to provide a dichotomous ‘yes/no’ response for each. These included school or college, websites and online videos, friends, parents/carers, they worked it out themselves, or they had not been taught how to stay safe online.

2.3.2. Following Advice

Following this question, participants were asked if they followed the e-safety education that they had received (one item). Responses were scored on a 4-point Likert scale, from ‘never’ (1) to ‘always’ (4).

2.3.3. Parental Understanding

Participants were asked if they believed that their parents or carers understood enough about online safety (one item). Responses were scored on a 4-point Likert scale, from ‘never’ (1) to ‘most of the time’ (4). Higher scores indicated poorer parental understanding.

2.4. Life-Affecting Worries

One item was used to measure life-affecting worry, with participants being asked to indicate if, and to what extent, their worries affect their everyday life. Participants rated this on a 4-point Likert scale from ‘never’ (1) to ‘most of the time’ (4).

2.5. Friendships

Eight items were used to measure belonging at school, exploring perceived quality of friendships, ability to share their problems with others, loneliness, and the impact of the pandemic on friendships. Items were scored on a 4-point Likert scale from ‘strongly disagree’ (1) to ‘strongly agree’ (4). The mean score was calculated to give a composite score for friendships, with lower scores indicating poorer perceived quality of friendships. A good level of internal consistency was found for this measure (α = 0.81).

2.6. Parental E-Safety Guidance

An exploratory factor analysis was conducted on items surrounding parental involvement and support within the family, and a three-factor solution was identified. The resulting measures were ‘parental e-safety guidance’, ‘emotional support from family’, and ‘personal competency’. The latter was not explored further in this paper, as it did not correlate with the outcome variable.

Parental e-safety guidance accounted for 32.61% of the variance [KMO = 0.836, χ² (66) = 6396.34, p < 0.001] and included six items covering active behaviours by parents to regulate online safety (e.g., limiting time online, checking age restrictions, parental controls, and learning to stay safe). Items were scored on a 4-point Likert scale from ‘never’ (1) to ‘most of the time’ (4): mean scores were calculated to provide a single score for the six items, and higher scores indicated greater parental e-safety guidance. A good internal consistency was calculated as α = 0.84. Notably, this variable differs from the aforementioned ‘who provides e-safety education’, with a greater focus on specific engagement behaviours from parents.

2.7. Emotional Support from Family

Emotional support from family accounted for an additional 14.56% of the variance (47.17% cumulative) and included two items covering the perceived ability to rely on parents and family members when experiencing problems online. These items were also scored on a 4-point Likert scale from ‘never’ (1) to ‘most of the time’ (4): mean scores were calculated to provide a single score for the two items, and higher scores indicated greater perceived emotional support from family. An acceptable internal consistency was identified (α = 0.72).

2.8. Trust in School Adults

Perceived relationships with adults at school were considered an important measure separate from those with peers at school. Only one item in the Cybersurvey 2020 measured relationships with adults outside the family and focused on the trust that participants felt towards staff. This was scored on a 4-point Likert scale from ‘strongly disagree’ (1) to ‘strongly agree’ (4).

2.9. Severe Online Harms

Focus was given to the risks perceived as the most severe (Boyd and Hargittai 2013; Jiang et al. 2021). Seven items were used to assess these, which included encountering the following risks: content that promotes eating disorders (pro-anorexia content); content that promotes self-harm; content that promotes suicide; content that displays violent videos and/or photos; feeling pressured into sexual activity with someone they met online (grooming); sharing nude photos of themselves; or meeting up with strangers met online. All items were given a dichotomous response for having encountered them (2) or not (1), and a total score was calculated. Higher scores indicated experiencing SOHs, with a total possible score of 14. A good internal consistency was identified (α = 0.80).

2.10. Analyses

Mann–Whitney U tests were conducted to explore differences between the two groups regarding their sources of e-safety advice, whether they perceived their parents understood enough about online safety, and if they followed the advice given. Non-parametric tests were favoured due to the violation of several assumptions, which are outlined in Appendix B. Eta-squared was utilised as the measure of effect size for the Mann-Whitney U tests and was interpreted as follows: 0.009–0.05 is small, 0.06–0.13 is moderate, and >0.14 is large (Richardson 2011).

Regressions with parallel mediation were conducted using Hayes’ Process Macro (v4.2). The predictor variable was a dichotomous measure of vulnerabilities (family vulnerabilities vs. no vulnerabilities), and the outcome variable was the exposure to SOHs. The mediating variables were participants’ life-affecting worrying, friendships, parental e-safety guidance, emotional support from family, and trust in school adults. Age and gender were controlled. The standardised regression coefficient (β) was utilised as the measure of effect size for the regression analyses, and was interpreted as 0.10–0.29 is small, 0.30–0.49 is moderate, and >0.50 is large (Nieminen 2022).

Following the regression analyses, further tests were conducted post hoc to explore the experiences of those with family vulnerabilities and life-affecting worries: Mann–Whitney U tests were used to explore differences in experiences of friendship, trust in adults at school, and emotional support from family for those with life-affecting worry.

3. Results

Encounters with at least one SOH were reported by 72.65% of adolescents with FV compared to 50.0% of NVAs. Correlations for each variable are outlined in Appendix C.

3.1. E-Safety Education

Analyses were conducted to compare the differences between the sources of online safety education reported by adolescents with FV and NVAs. It was found there were no differences between these two groups in learning e-safety from school or college (p = 0.38), websites or online sources (p = 0.35), parents or carers (p = 0.33), themselves (p = 0.06), or not having learned this at all (p = 0.48). There was, however, a significant difference between those with FV and NVAs in the extent to which they had learned about e-safety from friends, U = 20,448.00, z = −2.34, p = 0.02, η² = 0.01: adolescents with FV (mean rank = 224.00) were significantly more likely to report having learned e-safety from friends than NVAs (mean rank = 203.00), but the effect size for this difference was small.

To assess engagement with e-safety education, respondents were asked if they followed the e-safety education advice given, and whether they felt that their parents or carers understood enough about online safety. Mann–Whitney U tests found that there were no differences between the two groups in the extent to which they said they followed the e-safety advice (p = 0.08, η² = 0.008); most adolescents in both groups reported following the advice frequently (FV, $\overline{\mathrm{x}}$ = 3.41; NVA, $\overline{\mathrm{x}}$ = 3.52). However, those with FV were significantly more likely to report feeling that their parents did not understand enough about online safety, U = 16,605.00, z = −3.97, p < 0.001, η² = 0.04. Those with FV felt that their parents were less knowledgeable about online safety (mean rank = 229.00) than NVAs (mean rank = 184.22), but the effect size was small.

3.2. Severe Online Harms

A multiple regression analysis with parallel mediation was carried out to predict the effect of having a family vulnerability on SOHs. Figure 1 displays these findings in the form of a path analysis. A significant effect was found, R² = 0.30, F(8, 218) = 11.74, p < 0.001, whereby FV was associated with increased exposure to SOHs (β = 0.50, t = 3.97, p < 0.001, 95% CI [0.25, 0.75]), with a large effect size found. FV predicted greater levels of life-affecting worry (β = 0.87, t = 7.27, p < 0.001, 95% CI [0.63, 1.10]) and poorer perceived friendships (β = −0.68, t = −5.46, p < 0.001, 95% CI [−0.92, −0.43]), both with large effect sizes. FV also predicted poorer emotional support from family with a moderate effect size (β = −0.43, t = −3.35, p = 0.001, 95% CI [−0.69, −0.18]) and lower rates of parental e-safety guidance with a small effect size (β = −0.27, t = −2.07, p = 0.039, 95% CI [−0.53, −0.01]). There was no effect of FV on trust in adults at school (p = 0.28). The mediation effect was partial: FV had both a direct and indirect effect on exposure to SOHs through life-affecting worry and parental e-safety guidance only. Adolescents who experienced high levels of life-affecting worry reported greater exposure to SOHs, with a small effect size (β = 0.19, t = 2.75, p = 0.006, 95% CI [0.05, 0.32]), whilst those whose parents provided greater levels of e-safety guidance experienced lower exposure to SOHs, with a moderate effect size (β = −0.30, t = −4.74, p < 0.001, 95% CI [−0.43, −0.18]). Friendships (p = 0.14), emotional support from family (p = 0.10), and trust in school adults (p = 0.68) did not have an effect on exposure to SOHs.

3.3. Life-Affecting Worries and Perceived Relationships

Following the findings from the regression model, analyses were conducted to examine FV in greater depth, with a focus on those experiencing life-affecting worry. It was found that adolescents with life-affecting worry were more likely to report poorer perceived friendships, with a moderate effect size (U = 3227.00, z = −4.64, p < 0.001, η2 = 0.12); these participants also reported feeling less trust in adults at school (U = 4249.00, z = −2.05, p = 0.04, η2 = 0.03) and lower emotional support from family (U = 3914.00, z = −2.47, p = 0.01, η2 = 0.03), with small effect sizes. There were no significant differences in their parents’ e-safety guidance (p = 0.07).

4. Discussion

This study aimed to understand how offline factors may contribute to online harm exposure for adolescents with family vulnerabilities. Attention was given to providers of e-safety education, parental e-safety guidance, good quality offline relationships (family, friends, and trusted adults at school), and life-affecting worry. Initial prevalence rates suggest that exposure to at least one SOH is normative for those with and without family vulnerabilities (75% and 50%, respectively). Although this is consistent with recent findings by Internet Matters (2025), who reported that 67% of children experience online harms; this study explores the most severe harms experienced by the most vulnerable young people. The high prevalence reported is cause for concern.

4.1. E-Safety Education

No differences were found between those with FV and NVAs in whether they had received e-safety education from parents or carers, schools, or online resources, with most respondents having received some guidance. This may be indicative of increasing awareness of online harms and a societal push towards increasing e-safety education; in 2012, Ofsted introduced requirements for schools to educate students on e-safety, with the Department for Education providing non-statutory guidance for schools in 2019 and 2023 (Department for Education 2023).

Those with FV reported higher reliance on friends for e-safety education than NVAs, although effect sizes were small. This reliance on friends is surprising given that children in care and young carers report disrupted social lives and can struggle to be accepted by peers (Brett et al. 2024; Dearden and Becker 2004; Dharampal and Ani 2020). Nonetheless, the quality and content of this e-safety advice was not reported in the Cybersurvey, suggesting that this guidance may not always be of a good quality; friends may encourage the use of ‘hidden apps’ or offer workarounds to avert caregiver restrictions. Future research would benefit from exploring what types of ‘e-safety education’ young people are providing to their peers, and whether this is consistent with the messages provided by schools and parents. Similarly, whilst young people may report an awareness of online risks and e-safety measures, they can be poor at accurately articulating how they can avoid these (Macaulay et al. 2020). This could question the credibility of peers as sources of e-safety education, unless given safety education on how to help each other. Likewise, Livingstone et al. (2014) found that young people may be confident in identifying some online harms, but they may lack knowledge around other harms, including some of the SOHs identified in this paper (e.g., pro-anorexia content). This further questions whether children can provide sufficient e-safety education to their peers. To our knowledge, previous research has not explored peer-based education in relation to e-safety education and the efficacy of this method, and future research would benefit from exploring this as a supplement to existing education programmes.

This reliance on friends for e-safety education may be partially explained by the findings of this analysis: those with family vulnerabilities reported feeling that their parents or carers did not ‘understand enough about’ online safety, which may undermine their credibility as a source of advice.

4.2. Exposure to Severe Online Harms

Adolescents with family vulnerabilities were substantially more at risk of SOH exposure than their counterparts without vulnerabilities, which is consistent with previous research (El Asam and Katz 2018; Katz and El Asam 2020). These young people frequently experience difficult and unstable offline lives, which may lead them to seek refuge or escapism in online sources.

This may also be partially explained by the types of harms considered ‘severe’ in this paper. Exposure to pro-anorexia, pro-suicide, and pro-self-harm content was studied, and adolescents with family vulnerabilities are disproportionately more likely to suffer from poor mental health (Cree 2003; Dubois-Comtois et al. 2021; Engler et al. 2022; Lewis et al. 2023), which itself is a risk factor for online harm exposure. Searching for content that promotes any of these behaviours—even once—can lead to increased later exposure, whether or not the young person wants to see it. Likewise, these online communities can provide a sense of belonging and social support to these vulnerable visitors (Rodgers et al. 2012). In addition, this study explored grooming, sending nude photos, and meeting strangers in real life. In interviews with young victims of online grooming and sexual abuse, Whittle et al. (2014) found that many of the victims experienced difficult family lives, which led to increased risk-taking online and lower resilience in their responses.

Alongside a direct effect on exposure to SOHs, FV predicted poorer perceived relationships with family and friends, supporting the findings of previous research (Brett et al. 2024; Dharampal and Ani 2020; Hong et al. 2021; Vacca and Kramer-Vida 2012). These adolescents report relationships characterised by negativity, conflict, and instability (Chikhradze et al. 2017; Linares et al. 2010; Mazzone et al. 2019; Rose and Cohen 2010): adolescents in social care may experience placement changes that limit their ability to form lasting relationships, whilst young carers seek friends in similar circumstances (Becker and Becker 2008). Furthermore, the pandemic represented a turbulent and isolating time for the general population, and these adolescents may have felt this to a greater extent: Those in social care would have missed out on face-to-face family visits, whilst young carers may have felt the pressures of caring for a family member without any respite. For these young people—and those reporting family worries—anxieties may have been amplified and felt inescapable. Likewise, domestic violence rates soared (Office for National Statistics 2020). This may have put pressure on family dynamics and friendships, which could have impacted how these relationships were perceived.

Interestingly, there was no effect of FV on trust in adults at school, contrary to what was expected: neither group (those with FV or NVAs) reported substantially high trust in adults at school, which could reflect the missed face-to-face school time during the pandemic. Typically, those with family vulnerabilities report poorer relationships with school staff due to instability in placements, missed classroom time, and a lack of understanding of their experiences (Rose and Cohen 2010; Vacca and Kramer-Vida 2012), but this was a common experience for all children during this period.

Interpersonal relationships did not mediate the effect of vulnerability on SOHs; nonetheless, this does highlight an additional challenge faced by these individuals. Strong support networks—both online and offline—are beneficial for adolescents’ wellbeing and sense of belonging (Mertika et al. 2020; Webster et al. 2021), which can consequently influence social media use. O’Day and Heimberg (2021) note that young people experiencing loneliness may seek support from online communities to compensate for the absence of these in their personal lives. This underscores the need for targeted digital safety interventions that recognise the unique challenges faced by adolescents with family vulnerabilities.

Whilst relationships with family members—including parents—did not mediate the effect of FV on SOHs, parental e-safety guidance did. Adolescents with FV reported lower rates of parental e-safety guidance, which thus increased the risk of encountering SOHs, highlighting that caregivers play a vital role in supporting young people online. In the case of family vulnerabilities, caregivers may have less of a capacity to actively monitor and educate their children about online harms, or they may lack sufficient knowledge on how to do so. To our knowledge, this has not been studied in existing research, and future research may investigate the perceptions of caregivers providing e-safety guidance, specifically for those who have vulnerable children in their care. Future research should also endeavour to understand whether there are differences between the various family vulnerability subgroups: It is likely that caregivers within each subgroup will have different perceptions of e-safety, alongside different needs when implementing online safety measures. This would allow for a targeted approach towards supporting caregivers in educating their children. Overall, parental e-safety guidance remains an important factor to consider when supporting young people online.

Finally, life-affecting worries mediated the relationship between family vulnerabilities and SOHs: young carers, those living in social care, and those with family concerns experienced higher levels of life-affecting worry, which in turn increased SOH exposure. In a study of 16 European countries, Milosevic et al. (2022) demonstrated that greater influence on children’s sense of wellbeing emanates from socio-economic factors, rather than from social media experiences alone. Children with family vulnerabilities are more likely to be experiencing socio-economic hardship than their counterparts: this may explain the increased life-affecting worries identified in this group and underscores their vulnerability across multiple aspects of their lives.

4.3. The Role of Life-Affecting Worries and Perceived Relationships

The role of excessive and constant worry has emerged as a vital consideration, which is less studied in the context of e-safety. With the previous analyses identifying that adolescents with family vulnerabilities report higher rates of life-affecting worry, this study further explored how life-affecting worry can further impact relationships and belonging within this subgroup. The results revealed that those with life-affecting worries reported poor-quality relationships with family and friends and less trust in adults at school. This is consistent with previous studies, which found that poorer perceived interpersonal relationships are associated with increased anxiety symptoms (Zheng et al. 2023). Additionally, this effect may be bidirectional, with poorer perceived relationships exacerbating worries, whilst the worries may lead to withdrawal from others. This highlights an additional risk for these young people.

4.4. Limitations and Recommendations for Future Research

Despite offering novel insight into the experiences of adolescents with family vulnerabilities relative to online harm exposure, there are limitations to this study. First and foremost, this study explores the experiences of young people living in social care, young carers, and those with worries surrounding their family life; although this offers a representation of various family vulnerabilities, the experiences will differ substantially between them. For example, those living in social care will have been removed from the care of their parent(s) and placed into the care of another, whilst young carers and those with worries will frequently reside with their families. Likewise, there will be differences between those in the social care system, with those living with a foster family having drastically different experiences from those living in a residential home (Brett et al. 2024); the distinction between types of social care was not measured in the Cybersurvey. Future research would benefit from distinguishing between the different types of family vulnerability, alongside focusing on different forms of social care on online harm exposure.

Secondly, the Cybersurvey 2020 was conducted during the COVID-19 pandemic. This offers a unique glimpse of an unprecedented situation, and it may call into question the temporal validity of some responses. The number of participants reporting life-affecting worries was considerably higher than the Cybersurvey 2019 (Katz and El Asam 2019), which is likely to be a reflection of the context in which the survey was conducted. As mentioned, the pandemic led to increased family difficulties (Calvano et al. 2022), alongside decreases in adolescents’ wellbeing (Jones et al. 2021). In addition, Widnall et al. (2022) reported that adolescents with low school and peer connectedness were at greater risk of suffering anxiety and depression between lockdowns and returning to school. This may be reflected in the results reported, with those experiencing life-affecting worries having lower quality interpersonal relationships. It would be beneficial for future research to draw comparisons between the results found in this study and later versions of the Cybersurvey.

Relatedly, the Cybersurvey only explored relationships with peers and family members; having an adult you trust at school was measured, but this is not necessarily an indication of relationships with teachers, nor other important school staff. Future research should explore a wider range of interpersonal relationships, including those with non-academic school staff, social workers, and other important groups.

4.5. Implications

This study has implications for a more holistic approach to keeping young people safe online. This approach emphasises the important roles played by emotional support, having a trusted adult at school and supportive friendship groups, alongside appropriate online safety advice from a range of credible sources.

However, as more than one in four (27%) households with children have parents without the critical skills for understanding and managing digital risk (Yates et al. 2024), both adults and children in vulnerable family situations are likely to need more tailored support. Parental e-safety guidance was found to be a significant mediator for the association of FV and SOHs.

The main assumption underlying the field of online safety has been that online risks can be mitigated via educational interventions in isolation, yet this was not corroborated in these analyses. The broader context within which digital resilience is derived is described by Hammond et al. (2022) as a socio-ecological framework, all domains of which contribute to building this online resilience. However, Estellés and Doyle (2025) found that the dominant discourse in the literature about online safety remains on education, with schools as the locus of delivery. Finkelhor et al. (2021) propose that the teaching of online safety could be integrated into well-established and evidence-based programmes currently addressing related offline harms rather than taught as an isolated subject. Sağlam et al. (2023) recommend that what is taught should be influenced not only by expert advice but also by listening to children’s voices.

In 2021, the UN Committee on the Rights of the Child adopted General Comment 25, clarifying that children’s rights also apply in the digital world and stressing the importance of access and protection for young people online. Concerns about online harms tend to lead to controlling or restrictive approaches, limiting agency, particularly for those in care.

To conclude, we propose that by understanding who is more at risk—alongside which offline factors may exacerbate harmful experiences—it might be possible to reduce their exposure to severe online harms.