Approval for the study was from the LFF ethics committee, the organisation hosting the chatline. As this work fitted criteria for a service evaluation and used anonymised data, no further approvals were required (http://www.hra-decisiontools.org.uk/research/index.html).

We used mixed methods, quantifying chat-line users and their characteristics in terms of age, sex, nature of stated problem and police involvement, and using a modified grounded theory approach (Glaser & Strauss, 1967; Glaser 1998) to analyse chat transcripts.

In 2002, the Lucy Faithful Foundation (LFF) established the Stop it Now UK and Ireland telephone helpline ‘to make sure adults know what they can do to keep children safe’ – specifically from sexual abuse. In late September 2020, the Foundation first implemented a live-chat/texting feature. Stop It Now services in the USA and Belgium (Flanders) have also introduced live-chat, but not evaluated it so far.

The live-chat service is provided by responders with relevant health, psychology, social work or criminal justice background, specific helpline training and additional training for live-chat responding. Its introduction was not widely advertised but explicitly offered on The Foundation’s website. At time of this research, as a pilot project, it was available for three four-hour periods per week.

Every new contact with a Stop It Now responder is given a unique personal number and each contact a unique contact number, using a Salesforce operating system proforma. Initially, people who contact the service are asked for a made-up first name and a true county of residence, the latter to facilitate advice on availability of services. For added anonymity, at this stage everyone is also asked for an ‘age bracket’ (see Online Supplement 2) rather than specify age or date of birth. In the event of further contact, the three datapoints of unique LFF number, chosen name and county are used to check the contactor identity. Callers might make new calls using different identities, but responders always ask about prior contact and try to link files, and review cases together periodically. There have been rare occasions when the story of a purportedly new caller suggests otherwise. All those designated as repeat callers/chatters in our study have had records linked as described.

The Foundation provides other forms of support and interventions, not all anonymously, but this study is confined to the anonymous live-chat and phoneline services.

Records of all new ‘chatters’ between 28^th September and 31^st December 2020 and all new phone-callers for the same period 24 months previously formed the comparison cohorts. The comparison caller period was chosen to ensure that none in the comparison group was also using the chatline and because call types may be season related. In view of 2020 pandemic-related effects, we also checked patterns of conventional phone call type by sex, age, call type and reported police involvement of phoneline users between 2018 and 2020. The total number of new calls for 2020 was higher (1,130) than in 2018 (987), but all other parameters similar: mean age of callers (2018: 41.65 [SD = 14.27]; 2020: 42.72 [SD = 14.20]; 60% of each cohort were men. Problem type distribution was similar, with around 54% in both concerned about their own behaviours, and 78% (2018) and 72% (2020) reporting police involvement.

Data on number of contacts, sex, age, reason for/nature of contact and police involvement were extracted into a database from the records according to a data extraction sheet (see Online Supplement 2). All calls and chats are anonymous, but we further anonymised by applying a research number to replace the Salesforce number.

Full transcripts of the live chats were obtained and all given names and geographic location data removed. For these purposes, single occasion chatters were those who had made a first single contact on one day or the repeat contact had been on that day to restore a dropped line. Definition of multiple time chatters was that at least two chats had occurred on at least two separate days and the repeated contact was not just about simple practicalities such as confirmation of receipt of advice or an appointment. Given that we adopted qualitative methods here, there was no formal comparison between single and multiple occasion chatters; we analysed their data separately.

In the 95 days between the live-chat launch on September 28^th and census date of 31^st December 2020, staff made 120 hours available. All these hours were used. 143 different ‘clients’ used live-chat at least once, most (105, 73%) using this alone; 38 also used the phoneline. Eighty-eight (84%) of sole live-chat users did so once only, nine on two occasions at least 24 hours apart and seven on three or more separate days. A further four connected for multiple chats within one day, but not otherwise during the study period. The largest number of chats by one service user on separate days during the period was 17.

Two-thirds of people using live-chat were men (overall: 95, 66% men; 44, 31% women; 4 not known [NK]; proportions identical for exclusive live-chat users). Mean [M] age was 35.41 years (SD = 13.38, range 14–65 years). No-one over 65 used live-chat; the line is offered for adults only but, three under-18s made contact.

Just over half (55, 52%) of people who used live-chat alone were concerned about their own behaviour, all but six of these identifying as men (3 women, 3 NK). Twenty-nine (28%) were concerned about the behaviour of someone close to them, e.g. partners, ex-partners, parents, all but 8 of these identifying as women. A miscellaneous group of 20 chats included professionals seeking advice.

Table 1 shows details of the age band distribution of the pure live-chat users, pure phoneline users and the mixed group. These confirmed that the age difference between service type users was mainly accounted for by about twice the proportion of under-30s preferring live-chat to the phoneline. Under 7% of chatters were over 60 and none over 65; just over 10% of phoneline users were over 60, with nearly 3% aged 70+.

Overall, those exclusively using live-chat were significantly younger. Subgroup checks confirmed that the difference in age distribution between those concerned about their own behaviours and those people concerned about ‘significant others’ were each, separately, in line with this (see Table 1).

Men and women were about as likely to use live-chat as the voice service (chatline only 68 men, 67%; phoneline only 596 men, 62%; both 27, 71%; u = 49675.5, p = .837). Table 2 shows that the distribution of main type of concern was similar between pure live-chat and pure phoneline users. Distributions were more varied among people using both but, as some of these groups were so small, they were excluded from analyses.

Examination of any caller-reported police contact showed that pure live-chat users concerned about others (excepting professionals, survivors, and information requesters) were less likely to report police involvement than pure phoneline users (Table 3a).

Among the 55 chatline and 509 conventional phone users concerned about their own behaviour, significantly fewer pure live-chat than pure phone users reported any police involvement prior to the first chat (Table 3b).

The age band distributions by reported police involvement of those concerned about their own behaviour are shown graphically in Figure 1. This confirms a significant difference in police involvement by median age, with the age discrepancies particularly evident at the age extremes. None of the 16 users over 69 were without prior police involvement.

A similar examination of age and contact method within the cohort concerned about their own behaviours is shown in Figure 2, confirming the median age of live-chat users was significantly lower than that of phoneline users, and that 14% more live-chat than phonelines users were under 30.

Figures 1 and 2 confirm that having no police involvement and being a chatline user are both associated with being under 30 (two-sample Kolmogorov-Smirnov test: D = 0.13, p = .34) whilst police involvement and phoneline use is more common amongst users over 30 (two-sample Kolmogorov-Smirnov test: D = 0.05, p = .55). We therefore explored the relationship between all three variables together. Figures 3a and 3b illustrate the relationships between contact medium (live-chat/phoneline), age, and prior police involvement.

Critically, age distributions between chatline and phoneline users are comparable in both subgroups of police involvement, as highlighted by the similarity of the chatline and phoneline curves in the cumulative frequencies of the age distributions. In the police involved subgroup, both chatline and phoneline users have a median age band of 30-39, in contrast to the median age band of 19-29 for the chatline and phoneline users in the no police involvement subgroup. Figure 3 thus confirms user age as the more dominant factor, younger age accounting for both lower likelihood of police involvement and higher likelihood of live-chat than conventional phone use.

It is perhaps unsurprising that younger people, growing up with chat technology, are more likely to choose it for assistance and this is consistent with a systematic review of research into health promotion chatlines (Brody et al., 2020). Furthermore, detailed examination of people relying solely on live-chat in our study suggested such difficulty in accessing any other help, that there was otherwise nothing between them and harming others. Matteson et al. (2011), albeit researching emotionally neutral use of live chat within library services, found that reasons for preferring live-chat were ‘convenience’ coupled with familiarity with and rejection of other services. The latter is perhaps relevant here as those concerned about their own behaviours described themselves as avoidant of other communication options.

The finding of lower likelihood of police contact in the chat group, as indicated by both ‘bystanders’ and probable/actual perpetrators may simply follow from the fact that younger men had had less time to be arrested. Further, the analyses tend to support this, but the finding is important anyway. It indicates that the chatline may be opening a resource for people not yet under external pressure to seek help, or who find even voice contact difficult and who might otherwise not seek help. Live-chat availability may, thus, create opportunities for primary prevention of harms as well as earlier interruption of harmful pathways.

Other research has suggested that discussing child sexual abuse with potential perpetrators could have a positive impact on preventing abuse. This work, however, has focussed on treatment programmes, in all but one study of outcomes, face-to-face treatment (Seto et al., 2024). McKibbin et al. (2017), from semi-structured interviews with 14 young people who had perpetrated harmful sexual behaviour and six treatment-providing workers, found three key opportunities for prevention – the first focussing on helping potential perpetrators manage their use of pornography. Accessing indecent images of children was the commonest concern of those contacting the live-chat service we studied, and there was encouraging anecdotal evidence of live-chat user acceptance of advice on this. The other two opportunities for prevention were attention to own, earlier victim experiences and provision of sexual education. There was certainly some reference to both of these needs even in single-occasion life-chat transcripts.

The abusing or potentially abusing live-chat men studied in detail were in emotional turmoil, sometimes with self-harm or suicide as clear risks. This fitted with the hopelessness and extreme labels they gave themselves. That said, there seems to be an important difference between them and people contacting a live-chat service primarily for suicide-related-behaviour concerns. In a large study of self-reported outcomes after live-chat for suicide prevention, many felt helped and just under half less suicidal (Gould et al., 2021), but shame was not mentioned; in our live-chat group concerned about their own sexual behaviour, shame was common, strongly expressed and may partly account for the low return rates. It is known that, while guilt may have value for resolution, shame is associated with avoidance (Tangney et al., 2011). The transcripts in our study suggested little capacity for self-forgiveness but avoidance at several levels. Even within the relative transience of live-chat work some attention to this may be possible, perhaps enabling a path towards an appropriate sense of guilt, acknowledging wrongdoing together with intent not to repeat it, and working accordingly (Griffin et al., 2016).

We found just three studies of outcome relating to live-chat with more similarities to our study, with young people finding it difficult to access help otherwise and including victims of sexual abuse. Users of a live-chat service for people who had been sexually assaulted were more likely than not to be satisfied with it, satisfaction being associated with perception of responder knowledge and skills (Finn et al., 2011). A more recent study considered how to optimise matching chatter characteristics and responses (Sindahl & van Dolen, 2020). Service users who completed questionnaires reported that longer sessions with high responder density of material were more beneficial than session frequency. That said, these were only chatter impressions of benefit. Even within a confidential service this might be researched further, for example by randomising chatters to more open conversational models and more closed advice giving and including number of subsequent calls, quality of exchanges and take up of other services as outcomes. For prevention of child sexual offending, analyses would also have to take account of potential confounding factors such as open versus arrest driven choice to chat, concerns about own behaviour versus concerns about whether or not a specific act constitutes a crime.

A limitation of this, and all other services for anonymous users, is that we could not follow through to data on actual prevention. As a first study of a new live-chat service for people concerned about sexual abuse of children, however, added prevention value is implied by the reach to more people without police involvement. Another limitation of studying anonymised services is that the truth of claims cannot be checked, but consistency in descriptions of concerns across contacts by multi-occasion users suggests little conscious dissimulation.

Patterns of service use may change as the service becomes more widely known. Generalisability may also be limited in that such service use differs between cultures; in this context, it is interesting that for this UK based helpline service the largest group of contacts is of people concerned about their own behaviour, whereas in a USA-based sister organisation the largest groups was of people concerned about others (Grant et al., 2019). That said, the youth factor in live-chat use seems a constant across all studies of chat-based services (Kuerbis et al., 2017). No data were available on ethnicity. Although it is recognised that prevention of child sexual abuse is a concern across all communities, there may be particular barriers in this context (e.g. Sawrikar, 2020). This would be an important area for further study.

A strength in the qualitative analyses was that we had verbatim conversation data, a limitation was, therefore, that it was not completely free speech; we compensated for direct questions by relying only on free text sections for coding.

Our research adds to already published literature as the first study of live-chat service for prevention of child sexual abuse. The service shows promise as all available time was taken, mostly by people who explicitly preferred it to other communication modes. In attracting a younger clientele, less likely to be police involved than conventional phoneline users, it opens new preventive prospects. Qualitative analyses of verbatim transcripts suggested that it may also be preferred by a particularly vulnerable group, otherwise feeling too ashamed to talk. Among men contacting about their own abusive behaviours, few returned for further chats. Men who did were characterised by sense of painful inner struggle with limited vision of resolution in desistence and feeling relief, but there was some evidence of psychological engagement about their difficulties. This suggests that future research should explore strategies to increase engagement.