How often do users of specific apps encounter harms?
Last updated . Reviewed by Ravi Iyer, USC Neely Center for Ethical Leadership and Decision Making.
Understanding the rates of harms experienced by users of a specific application can help improve safety in several ways, including:
- It can help youth and parents make more informed decisions about which apps to use, balancing risks and benefits.
- It can help platforms, policymakers, and regulators make more informed decisions around what is going well and where efforts to reduce risk should be focused.
This page summarizes data from every survey in our index that provides data on harm rates at the level of individual applications: seven studies across the US, UK, and Australia, covering ages 9–17 — USC Neely / Stanford CDH, Pew, Thorn, Common Sense, the Youth Endowment Fund, Ofcom, and Australia’s eSafety Commissioner. It is organized as key findings, then limitations and areas for future work, then the studies table and the individual study sections.
Key findings
These seven studies use different populations, question wordings, recall windows and denominators. For this reason, we focus on comparisons of within-study ordering across studies; this reveals the following findings.
1. Problematic use: Sleep disruption common across popular apps, with highest rates on TikTok
Three independent US samples find TikTok to have the highest rates of reported sleep disruption. For example:
- Common Sense (girls ages 11–15): “Girls were also more likely to say that TikTok use (versus other platforms) interferes with their sleep” (Report p.12 (printed p.6)). 41% say TikTok gets in the way of sleep at least weekly, compared to 31% for Snapchat, 29% for Instagram, and 28% for YouTube.
- Pew (ages 13–17): “TikTok users are more likely to report that it negatively impacts their sleep than those on Snapchat or Instagram” (Report p.8 (printed p.7)). 37% say TikTok has hurt how much sleep they get, compared to 26% for Snapchat and 24% for Instagram.
- USC/Stanford (ages 13–17): 42.2% of TikTok users say their use of the app gets in the way of sleep, followed by 27.8% for Instagram, 27.7% for YouTube, and 20% for Snapchat.
2. Contact or conduct harms concentrate on apps that have meaningful messaging features
Several studies place Snapchat and Instagram among the apps with higher reported contact or conduct harm rates, with YouTube often ranking lower. The ordering within the top group is not consistent across studies, but the overall groupings are. For example:
- USC / Stanford, unwanted conversations in the last four weeks: text messaging 10.2%, Snapchat 9.5%, Discord 9.4%, Instagram 7.3%. YouTube’s rate is 0.2%.
- Common Sense, strangers making uncomfortable contact: Instagram 58%, Snapchat 57%, YouTube 30%.
- Thorn, online sexual interactions across five waves: Snapchat 18–23%, Instagram 18–22%, YouTube 7–9%.
- Pew, harassment on the platform: Snapchat 27%, Instagram 19%, TikTok 18%.
3. Content harms do not follow the same ranking and often vary depending on the specific harm
The ranking of apps for content harms can differ from their ranking for contact or conduct harms and varies considerably depending on the specific content harm measured. For example:
- Ofcom: X leads on rates of serious violence, serious injuries, suicide and pornography, while TikTok leads dangerous stunts, self-harm and both eating-disorder items. YouTube ranks anywhere from second to eighth of fourteen across the content types shown here — second on dangerous stunts at 15.3%, eighth on pornography at 1.9%.
- USC / Stanford has Instagram and Snapchat near the top for unwanted sexual images (7.7 and 6.1% last four weeks, respectively). Instagram is also near the top for violent / disturbing images (9.1% last four weeks), while Snapchat is near the bottom (0.4% last four weeks). Instagram has the highest reported rate of users feeling worse about themselves or their lives because of others’ posts, messages or communications (16.6%).
- The Youth Endowment Fund has TikTok (44% last 12 months) and X (43%) as the top two apps for violent content.
4. Some less commonly used apps may have among the highest risks
Some of Thorn’s apps with the highest prevalence of sexual interactions are Telegram (24% lifetime prevalence; 20% have ever used the app) and Kik (20% lifetime prevalence; 17% have ever used the app). In Ofcom’s content matrix, X, with a relatively low 14% usage in the last four weeks, tops four of the eight content types by prevalence — pornography exposure is reported by 8.8% of its users against YouTube’s 1.9% (X has a fifth of YouTube’s usage in the last four weeks). The Youth Endowment Fund finds the same pattern for violent content: X has 43% prevalence among its users and 22% usage in the past 12 months, second only to TikTok’s 44% prevalence and 69% usage in the past 12 months.
Limitations and areas for future work
1. Selection effects may be driving some of the per-app differences
Differences in harm rates could be due to differences in the apps themselves as well as differences in the children who choose to use each platform. Both are likely to contribute to effects and may interact, as platform design can shape and respond to norms and user bases, and vice versa. Going forward, it would be valuable to better understand the relative size of these effects.
2. Dosage effects may be driving some of the per-app differences
Similarly, children may use apps at different levels of intensity. Apps used for 1–2 minutes/day may have lower measured harm rates than those used for 1–2 hours/day, all else equal. Although intensity is arguably at least partially an app-controllable characteristic, going forward, it would be valuable to better understand the size of dosage effects.
3. Only two of the seven studies are probability samples, and both relatively narrow
USC / Stanford covers a broad set of harms and apps with a weighted analytic sample of 472, while Pew covers a more limited set of harms and apps on 1,458 respondents. The largest samples here, and all three non-US studies, are non-probability panels. Additionally, very few studies publish uncertainty. Only USC / Stanford provides a standard error for each estimate shown here, using separately supplied unsuppressed crosstabs for cells omitted from the article, and only Pew publishes subgroup margins of error. Going forward, larger-scale broad probability samples with transparent uncertainty calculations, as well as ways to calibrate non-probability samples to probability samples, would help.
4. Very few studies provide consistent tracking over time
In the US, only Thorn does this, and only for a limited set of harms. Going forward, establishing standardized, ongoing surveillance similar to other areas in public health would allow for a better understanding of changes over time.
5. Coverage gaps (age, apps, and harm categories)
Some areas where limited data exist include: (a) Harms to under-13s: Thorn reaches age 9 and eSafety age 10, but the rest start at 11 or 13; (b) Gaming and voice apps: where studies do measure them they matter: USC/Stanford reports similar rates of seeing self-harm encouraged for online gaming (6.3%) and TikTok (6.4%); (c) Eating disorder, NCII, and sextortion harms: these have limited coverage and often on small cells. Extending this methodology to other popular online platforms will be essential for fully understanding online harm.
Studies included
Studies that report per-application harm data typically do it in two ways:
- Prevalence among an app’s users — of everyone who uses this app, what share experienced the harm on it.
- Attribution among victims — of everyone who suffered this harm, what share name this app. Helps size where the most children are being harmed, but a high attribution share could be driven by high reach or by high risk.
Our goal was to include a study if:
- It measures at least one harm, per application. Note that underage use is considered a risk factor, so it is not included.
- It reports prevalence directly, or provides a way to derive prevalence via a per-app usage rate — over the same sample and the same window — from which a per-user rate can be derived.
- It has sufficient sample size per platform to provide meaningful data. In practice, we require at least 100 respondents reporting usage of at least one platform in the study.
We plan to periodically update this page, and feel free to contact us if new studies should be included. The table below shows the studies currently included. For studies that are multiple waves, we only show one row.
Studies were considered for inclusion up to . Surveys added to the index after that date are not reflected here. The USC / Stanford entry was updated to include data from the published article (vs. preprint) and separately supplied unsuppressed crosstabs on September 21, 2026.
| Study | Country | Ages | n | Sampling | Apps | Harm coverage | Recall | Measure |
|---|---|---|---|---|---|---|---|---|
| USC Neely / Stanford CDH 2026 | US | 13–17 | 472weighted | Probability | 8 published16 measured | 10 harms | 4 weeks / general (sleep) | Prevalence |
| Pew Research Center 2026 | US | 13–17 | 1,458 | Probability | 3 | 4 domains + 3 harassment items | Lifetime / general | Prevalence |
| Thorn Youth Perspectives 2020–24 | US | 9–17 | 1,048 | Non-probability | 22 | Sexual interactions | Lifetime | Prevalence |
| Common Sense (girls) 2023 | US | 11–15 | 1,397 | Mixed | 4 | 8 harms + offline | Mixed thresholds | Prevalence |
| Youth Endowment Fund 2024 | UKEngland & Wales | 13–17 | 10,387 | Non-probability | 14 published17 measured | Violent content | 12 months | Prevalence |
| Ofcom Children’s Online Safety Tracker 2025 | UK | 11–17 | 1,652Module 2 | Non-probability | 17 | 8 content types | 4 weeks | Prevalence |
| eSafety Keeping Kids Safe Online 2025 | Australia | 10–15 | 2,629 | Non-probability | 24 | 6 harms | Lifetime | Attribution → derived |
Notable exclusions of studies with per-app data
- Under-13 use. Ofcom’s Children’s User Ages waves, the ABCD study and eSafety’s Behind the Screen all publish per-app figures, but for use and account ownership rather than harm, so none of them answers how risky an app is for a child who uses it. Use below an app’s minimum age is a risk factor rather than a harm, per the criteria above. Those sources are collected on How many under-13s use each app?
- The Anti-Defamation League’s 2024 survey publishes per-platform figures for its teen sample across 11 platforms, as a share of the 264 teens who were harassed rather than of users, with no teen usage denominator published — so no per-user rate can be recovered from it.
- Ofcom’s Online Experiences Tracker has the longer run of waves, but its whole teen sample is 442–976 per wave and its per-harm, per-platform teen bases fall to 0–24 respondents — below Ofcom’s own reporting threshold — so no per-app harm rate is reportable in any wave. It also publishes no usage denominator after Wave 2. COST is the better Ofcom representative; OET stays in the index for the long-run trend.
- Thorn’s sextortion survey. It breaks non-consensual intimate imagery down across 33 apps, but as attribution among victims with no published usage denominator — the same case as the ADL bullet above, so it cannot answer how risky an app is for someone who uses it.
- Earlier waves of trackers. Only the most recent representative wave of each tracker is included as a study. Where that wave’s own report includes a time series, as Thorn’s 2024 report does for 2020–24, the trend is shown inside that section. Ofcom’s Wave 1 publishes the same per-service items for 14 named services and is not charted here: Wave 2 widened the harm list and replaced the generic gaming code, so the two waves are not directly comparable.
United States: USC Neely / Stanford Center for Digital Health (2026)
US·ages 13-17·weighted n = 472·Probability·16 apps measured; 8 with published rates·4 weeks; general for sleep·sources
This is the widest harm-by-app matrix in the index: ten distinct harms measured on each of eleven named contexts, on a US probability sample, all on the same four-week window except the sleep item, which has no recall period, with an offline comparator where applicable. Other surveys reach more apps — Thorn covers 22 — or larger samples, but for narrower harm sets, without true per-app prevalence reporting, and/or on non-probability samples.
Per-app prevalence rates of harms among each app’s users
Each cell is the weighted % of that app’s past-4-week users who experienced the harm on that app, with its standard error (SE) in parentheses; harms are grouped by the site’s taxonomy categories (Conduct & Contact, Content, Problematic Use).
Click a app heading to sort harm rows, or a harm heading to sort app columns. Click again to reverse the order.
| Harm | In personweighted n=472 | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1.3 (0.6) | 8.5 (1.7) | 0.4 (0.4) | 7.2 (1.9) | 4.3 (1.3) | 6.4 (1.8) | 0.9 (0.6) | 4.7 (1.7) | 0 (0) | 7.6 (2.6) | 5.3 (2.8) | 18.5 (2) | |
| 0.2 (0.2) | 10.2 (1.9) | 0.7 (0.5) | 7.3 (1.7) | 1.9 (0.8) | 9.5 (2.2) | 3.9 (1.5) | 5.5 (1.8) | 0.4 (0.4) | 9.4 (3) | 2 (1.2) | 12.8 (1.7) | |
| 0.6 (0.4) | 1.3 (0.7) | 0 (0) | 1.3 (0.6) | 1.1 (0.6) | 3.5 (1.4) | 0.1 (0.1) | 0.1 (0.1) | 0 (0) | 1.6 (1.2) | 2.4 (1.3) | 2.9 (1) | |
| 0 (0) | 2.1 (0.9) | 0 (0) | 0.2 (0.1) | 0.7 (0.4) | 2.2 (1.2) | 1 (0.7) | 0 (0) | 0 (0) | 1.8 (1.8) | 1.7 (1.2) | 1.4 (0.7) | |
| 4.1 (1.2) | 5.6 (1.4) | 0.6 (0.4) | 16.6 (2.5) | 13.4 (2.3) | 10.8 (2.4) | 0.1 (0.1) | 3.1 (1.5) | 5.8 (1.9) | 2.3 (1.3) | 2.6 (1.5) | 8.7 (1.4) | |
| 4.3 (1.2) | 0.3 (0.3) | 0.6 (0.4) | 7.7 (1.8) | 5.6 (1.5) | 6.1 (1.8) | 0 (0) | 1.8 (1.1) | 1.2 (0.8) | 3.6 (2.1) | 3.4 (1.8) | 2 (0.7) | |
| 1.9 (0.7) | 0.6 (0.4) | 0.7 (0.4) | 9.1 (1.9) | 5.4 (1.4) | 0.4 (0.4) | 0 (0) | 0.8 (0.4) | 0.8 (0.6) | 3.5 (1.8) | 1.7 (1.2) | 1.6 (0.6) | |
| 1.4 (0.6) | 2.2 (1) | 0 (0) | 3.4 (1.2) | 7 (1.8) | 3.6 (1.3) | 0.3 (0.3) | 2.6 (1.1) | 0 (0) | 7.8 (3) | 0 (0) | 8.6 (1.5) | |
| 2.2 (0.9) | 2.8 (1) | 0.6 (0.4) | 5.8 (1.7) | 6.4 (1.7) | 4.6 (1.7) | 0.2 (0.2) | 6.3 (2.2) | 0.4 (0.4) | 4.8 (2.1) | 1.5 (1.1) | 5.2 (1.2) | |
| 27.7 (2.7) | 10.2 (1.8) | 0.6 (0.4) | 27.8 (3) | 42.2 (3.4) | 20 (2.9) | 9.8 (2.2) | 18.8 (3.3) | 8.2 (2.3) | 8.2 (2.7) | 5.6 (3.1) | N/A |
Click an app column above to see a plain-language summary for that app.
Question wording & response options
- Conduct & ContactBullying / harassment“In the past 4 weeks, whether online or in person, has anyone done any of these things to you? • Insulted or disrespected you • Contacted you in an inappropriate way • Damaged your reputation • Threatened you • Excluded you or left you out”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services did you experience being a target of harassment (e.g. an insult, disrespect, inappropriate contact, reputational harm, threat, or exclusion)?”
- Conduct & ContactUnwanted conversation“In the past 4 weeks, whether online or in person, have you gotten involved in an unwanted conversation?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services have you gotten involved in an unwanted conversation?”
- Conduct & ContactUnwanted sexual advances“In the past 4 weeks, whether online or in person, have you received unwanted sexual advances?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services have you received unwanted sexual advances?”
- Conduct & ContactPhotos used inappropriately“In the past 4 weeks, whether online or in person, has anyone used your photos in an inappropriate way?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services has anyone used your photos in an inappropriate way?”These are unfiltered photo-misuse results, not NCII-specific rates.
- ContentFelt worse about self“In the past 4 weeks, whether online or in person, have you felt worse about yourself or your life because of other peoples' posts, messages, or communications?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services did you feel worse about yourself or your life because of other peoples' posts, messages, or communications?”
- ContentUnwanted sexual images“In the past 4 weeks, whether online or in person, have you seen nudity or sexual images that you didn't want to see?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services did you see nudity or sexual images that you didn't want to see?”
- ContentViolent / disturbing images“In the past 4 weeks, whether online or in person, have you seen any violent, bloody, or disturbing images that bothered you?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services have you seen any violent, bloody, or disturbing images that bothered you?”
- ContentSaw self-harm“In the past 4 weeks, whether online or in person, have you seen someone harm themselves, or threaten to do so?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services have you seen someone harm themselves, or threaten to do so?”
- ContentSaw self-harm encouraged“In the past 4 weeks, whether online or in person, have you seen anyone encourage someone to harm themselves?”Response options: Yes / NoPlatform follow-up (select all that apply): “In which contexts/online services have you seen anyone encourage someone to harm themselves?”
- Prob. UseDisrupts sleep“How often do you feel that your use of technology gets in the way of your sleep?”Response options: Never / Rarely / Sometimes / Often (reported as Sometimes + Often)Platform follow-up (select all that apply): “Which of the below online services get in the way of your sleep?”
Values are weighted percentages with standard errors from the unsuppressed crosstabs corresponding to Table 2 of the JAMA Pediatrics article (pp.5–6). Rates reflect experiences that occurred specifically on the named platform, among that platform’s past-4-week users. The In person column uses the full weighted sample of 472, so it is on a different base and is not shaded; N/A for sleep means no offline counterpart. Sleep has no specified recall period. Excluded: non-consensual intimate imagery (NCII), which the survey measured with overall and context follow-ups but for which the supplied table does not provide NCII-specific per-app rates; underage use and eating-disorder content, which the survey did not ask about; and second-order “witnessing” items (harm seen happening to other people).
This table shows all eleven contexts from the article’s Table 2: eight branded apps plus Text Messaging, Email and Online Gaming. The study measured sixteen branded apps; only eight have published harm rates. See the reference page for the full measurement scope. The source sheet places 0 (0) in the offline sleep cell; it is shown as N/A here because the questionnaire has no offline sleep item.
United States: Pew Research Center (2026)
US·ages 13-17·n = 1,458·Probability·3 apps·lifetime / general recall·sources
Pew’s April 2026 report covers Instagram, Snapchat and TikTok, on a US probability sample. It asks about 4 life impacts (sleep, productivity, mental health, and friendships) on a help/hurt scale as well as 3 bullying/harassment experiences. It is the largest US probability sample with per-app harm reporting.
TikTok users are most likely to say the app has hurt how much sleep they get, while Snapchat users are most likely to say they’ve experienced harassment on the app.
Per-app prevalence rates of harms among each app’s users
Click a app heading to sort harm rows, or a harm heading to sort app columns. Click again to reverse the order.
| Harm | |||
|---|---|---|---|
| 37 | 24 | 26 | |
| 18 | 19 | 27 | |
| 15 | 16 | 23 | |
| 10 | 12 | 18 | |
| 5 | 7 | 10 |
Question wording & response options
- Prob. UseSays it hurt how much sleep they get“Do you think using [Instagram / Snapchat / TikTok] has helped or hurt…” — “How much sleep you get”Response options: Helped a lot / Helped a little / Neither helped nor hurt / Hurt a little / Hurt a lot / No answerReported here: NET of “Hurt a little” and “Hurt a lot”.The modal answer is “neither helped nor hurt” (56–68%), and “hurt a lot” alone is 6% on TikTok, 3% on Snapchat and 3% on Instagram.
- Conduct & ContactAny of the three harassment items“Have you ever experienced any of the following on [Instagram / Snapchat / TikTok]?” — “Any of the three”Response options: Yes, this has happened to me / No, this has not happened to me / No answerReported here: Yes to any of the three items. Published only in the report chart, not the topline.
- Conduct & ContactCalled an offensive name“Have you ever experienced any of the following on [Instagram / Snapchat / TikTok]?” — “Being called an offensive name”Response options: Yes, this has happened to me / No, this has not happened to me / No answerReported here: “Yes, this has happened to me”. Lifetime-ever: no frequency, severity or repetition qualifier, and no bullying definition is shown to respondents.
- Conduct & ContactRumors spread about them“Have you ever experienced any of the following on [Instagram / Snapchat / TikTok]?” — “Having someone spread rumors about you”Response options: Yes, this has happened to me / No, this has not happened to me / No answerReported here: “Yes, this has happened to me”. Lifetime-ever: no frequency, severity or repetition qualifier, and no bullying definition is shown to respondents.
- Conduct & ContactPhysically threatened“Have you ever experienced any of the following on [Instagram / Snapchat / TikTok]?” — “Being physically threatened”Response options: Yes, this has happened to me / No, this has not happened to me / No answerReported here: “Yes, this has happened to me”. Lifetime-ever: no frequency, severity or repetition qualifier, and no bullying definition is shown to respondents.
On life impact, sleep is the one that most directly maps to problematic use. The other three — productivity, mental health and friendships — were also measured. Notably, while sleep and productivity were net negative for all three apps, friendships were net positive on all three and mental health on TikTok and Snapchat — on Instagram the reported shares for helped and hurt are similar (13% against 11%). Net helped-minus-hurt across the three (percentage points): sleep -30 to -17; productivity -13 to -7; mental health +2 to +10; friendships +20 to +37.
Sleep from the report p.8 (printed p.7), chart “Teen TikTok users stand out for saying it hurts their sleep and productivity”; harassment from the report p.10 (printed p.9), “About a quarter of teens on Snapchat report being harassed there, higher than Instagram or TikTok”. Item-level figures from the topline pp. 9–16. Heat shading is on one scale across all rows, but the two batteries are different kinds of measure — see the badges, and the question wording above for what each row asks and reports. Pew publishes a margin of error per platform subgroup rather than per cell (percentage points) — TikTok ±4.0, Instagram ±4.2, Snapchat ±4.5 — see the report p.28 (printed p.27).
United States: Thorn Youth Perspectives (2020–2024)
US·ages 9-17·n = 1,048·Non-probability·22 apps·lifetime recall·sources
Thorn covers two harm categories — online sexual interactions and online bullying/harassment — across 22 apps and 5 consecutive annual waves on a fixed instrument, for a US 9–17 non-probability sample. The annex reports three per-platform blocks across all 5 waves — reach, online sexual interactions (Fig a-08), and a “potentially harmful online experience” net (Fig a-09). The table here charts online sexual interactions. The net is a union that folds in non-sexual harm and is printed on the all-minors base, and Thorn never reports the bullying slice on its own, so it is not comparable with the per-user rates below and is not charted here. The sample size in the strip above, n=1,048, is the 2024 wave; the four earlier columns are four separate samples whose sizes the 2024 report does not print. Thorn additionally asks about interactions with those the minor believes to be an adult; rates are lower but overall ordering is broadly the same.
Snapchat’s reported rate was 21–23% in 2020–2023 and 18% in 2024. YouTube runs at 7–9% of its minor users, consistently below Thorn’s own cross-platform average (13–15%). Instagram (18–22%) and Snapchat (18–23%) are above that average in every year. TikTok sits lower at 12–14%. For comparison, pornography sites as a category were 30–38% in 2023 and 2024, the only two years Thorn included this category.
Per-app prevalence rates of online sexual interactions among each app’s users, by year
Click a year heading to sort app rows. Click again to reverse the order.
| App2024 lifetime usage | |||||
|---|---|---|---|---|---|
| Thorn’s printed average | 13 | 14 | 14 | 15 | 15 |
| YouTube98% ever used | 9 | 8 | 8 | 7 | 8 |
| Minecraft81% ever used | 5 | 5 | 7 | 7 | 8 |
| Roblox80% ever used | 5 | 5 | 8 | 8 | 9 |
| TikTok77% ever used | 14 | 13 | 12 | 14 | 14 |
| Fortnite75% ever used | 8 | 7 | 8 | 7 | 7 |
| Messenger71% ever used | 18 | 20 | 18 | 19 | 18 |
| Facebook69% ever used | 16 | 20 | 19 | 17 | 20 |
| Instagram69% ever used | 22 | 20 | 18 | 20 | 20 |
| Snapchat69% ever used | 23 | 22 | 21 | 23 | 18 |
| Call of Duty62% ever used | 11 | 7 | 9 | 6 | 7 |
| Grand Theft Auto56% ever used | 10 | 6 | 10 | 8 | 8 |
| Discord51% ever used | 14 | 15 | 17 | 14 | 19 |
| Google Hangouts/Meet49% ever used | 15 | 16 | 14 | 13 | 17 |
| Twitch49% ever used | 9 | 9 | 12 | 10 | 15 |
| WhatsApp48% ever used | 21 | 20 | 18 | 18 | 19 |
| X (Twitter)48% ever used | 15 | 19 | 19 | 15 | 16 |
| Reddit46% ever used | 7 | 11 | 14 | 10 | 14 |
| Among Us43% ever used | 6 | 4 | 6 | 7 | 11 |
| Tumblr33% ever used | 20 | 23 | 18 | 16 | 19 |
| Telegram20% ever used | 21 | 19 | 28 | 22 | 24 |
| Kik17% ever used | 23 | 26 | 24 | 23 | 20 |
| Pornography Sites(not an app)13% ever used | — | — | — | 30 | 38 |
| BeReal12% ever used | — | — | 14 | 17 | 11 |
Question wording & response options
Conduct & ContactBoth sexual-interaction measures come from one question (Q20), asked as a platform grid:“For each of the following platforms, please indicate if you have ever experienced any of the following. Please select all that apply then click next. Remember that your answers are anonymous.”Response options: Multiple select across the platform grid; no frequency or severity scale.
- An online sexual interactionDenominator: Minors 9–17 who use that platformReported here: Charted above. Thorn also publishes the same numerator against all minors 9–17.
- An online sexual interaction w/ someone they thought was 18 or olderDenominator: Minors 9–17 who use that platformReported here: Not charted. Levels run roughly half the “any” column, with broadly the same ordering; Thorn’s printed average for it runs 7–10%.
Platform usage (Q14B): “How often do you use/check/play each of the following?” The usage figures show the share who have ever used each platform. Fig a-09, report p.48.
For sexual interactions, recall is lifetime (“ever”) — Thorn publishes no time-bounded per-platform variant. The report prints the grid’s column labels but not the item text shown to respondents, so the two labels above are Thorn’s wording for the derived columns rather than verbatim question items. Thorn describes its platform list as “diverse but non-exhaustive” and changes it between waves, and flags its own undercount risk: “Entrenched stigma and sensitivity surrounding the topic may lead to an undercounting of the scale and frequency.”
Report p.47 (Annex); 2024-only counterpart Fig 8, p.15. “Fig a-08 | Online sexual interactions involving both peers and adults by platform, by year”. The percentage beside each platform is the share of all surveyed US 9–17-year-olds in the selected year who had ever used it (Fig a-09, report p.48). Clicking a year sorts the harm rates and shows that year’s usage figures. The initial view and Reset use 2024 usage, with apps ordered largest first. Heat shading applies to the five harm-rate year columns only, on a shared scale. Thorn’s “average” row reproduces the report’s printed cross-platform benchmark, not overall prevalence among all minors. Em-dash = platform not carried that wave. Thorn publishes no per-platform base n and no margin of error.
United States: Common Sense Media, girls 11–15 (2023)
US·ages 11-15·n = 1,397 (girls only)·Mixed·4 apps·mixed frequency thresholds·sources
Common Sense Media asks about a broad range of harms across 4 applications (TikTok, YouTube, Instagram, and Snapchat, plus a messaging-apps bucket) among a mixed (probability and non-probability) sample of girls aged 11–15. They (along with USC / Neely) also ask for “in real life” prevalence of harms when applicable. While their survey instrument provides a consistent set of threshold options, their published report does not always provide aggregates for each threshold. The table below includes a subset of reported harms from the study, with the threshold noted for each row.
TikTok users (41%; at least weekly) are most likely to report it gets in the way of sleep, while Instagram (58%; ever) and Snapchat (57%; ever) are most likely to report experiencing uncomfortable contact from strangers.
Per-app prevalence rates of harms among each app’s users
Click a app heading to sort harm rows, or a harm heading to sort app columns. Click again to reverse the order.
| Harm | Threshold | Real lifen=1,397 | |||||
|---|---|---|---|---|---|---|---|
| at least weekly | 28 | 41 | 29 | 29 | 31 | — | |
| at least weekly | 34 | 45 | 30 | 33 | 37 | — | |
| ever | 30 | 46 | 48 | 58 | 57 | — | |
| at least weekly | 20 | 30 | 26 | 29 | 33 | 36 | |
| at least monthly | 29 | 39 | 29 | 42 | 35 | 39 | |
| at least monthly | 28 | 39 | 26 | 37 | 32 | 36 | |
| daily | 11 | 21 | 28 | 23 | 30 | — | |
| at least weekly | 23 | 31 | n/a | 32 | 28 | 27 |
Question wording & response options
Every row comes from one battery, asked separately about each platform:“E1. How often have you had the following experiences on …”Response options: Never / Only once or twice in my life / Once or twice per month / Once or twice per week / About once per day / Multiple times per day / Almost constantlyOffline comparator: “E2. How often have you had the following experiences in “real life” (i.e., in-person, NOT on social media)?”
- Prob. UsePlatform gets in the way of sleep(E1c)“Using it gets in the way of my sleep.”Reported here: at least weeklyFigure 4, p.32 (printed p.26) — "Interference of social media with sleep"
- Prob. UseFeel “addicted” / use longer than intended(E1b)“I feel like I’m “addicted” or end up using it for a longer period of time than I originally wanted to.”Reported here: at least weeklyFigure D, p.12 (printed p.6) — "Overuse of social media, by frequency"
- Conduct & ContactStrangers make uncomfortable contact(E1d)“Strangers try to contact me in ways that make me uncomfortable.”Reported here: everFigure P, p.21 (printed p.15) — "Unwanted contact by strangers on social media"; frequency bands Figure 15
- Conduct & ContactDrama, bullying, mean comments, exclusion(E1h)“Something happens with my friends or peers that makes me feel bad (like drama, bullying, mean comments, fighting, or feeling left out).”Reported here: at least weeklyFigure 10, p.39 (printed p.33) — "Negative social experiences on social media and in “real life,” by frequency"
- ContentUpsetting suicide / self-harm content(E1j)“I see or hear things about suicide or self-harm that are upsetting to me.”Reported here: at least monthlyTable 11, p.46 (printed p.40); bands Figure 17, p.45 (printed p.39)
- ContentUpsetting eating-disorder content(E1k)“I see or hear things about eating disorders that are upsetting to me.”Reported here: at least monthlyFigure 18, p.45 (printed p.39); NETs corroborated in Table 11, p.46
- Prob. UsePressure to be available / respond right away(E1a)“I feel pressure to be available, “on,” or responsive right away.”Reported here: dailyFigure 5, p.32 (printed p.26) — "Availability stress"
- ContentFeel bad about my body or appearance(E1q)“I feel bad about my body or appearance.”Reported here: at least weeklyFigure O, p.20 (printed p.14) — "Negative body image experiences … by frequency"
Figures as cited per row above. Heat shading covers the five platform columns only, on a shared scale; the real-life column is on a different base (all 1,397 girls, not platform users) and is not shaded. Thresholds differ by row — Common Sense reports some items at “at least daily,” some at “at least weekly,” some at “at least monthly,” some as “ever” — so compare apps within the same row. “n/a” = not asked for messaging apps.
England & Wales: Youth Endowment Fund (2024)
UK — England & Wales·ages 13-17·n = 10,387·Non-probability·17 apps measured; 14 with published rates·12-month recall·sources
The Youth Endowment Fund surveys 10,387 teenagers in England and Wales on a non-probability panel. It covers one harm: seeing real-world violent content, over a past-12-month window.
TikTok users are most likely to report seeing violent content on the app (44%), followed by X (43%) and Facebook (33%). YouTube sits in the middle (26%), and the messaging services run lowest — for example, WhatsApp is 18% and iMessage is 5%.
Per-app prevalence rates of violent content exposure among each app’s users, past 12 months
| AppPast-12-month reach | ContentViolent content |
|---|---|
| YouTube79% reach | 26 |
| WhatsApp74% reach | 18 |
| TikTok69% reach | 44 |
| Instagram60% reach | 31 |
| Snapchat58% reach | 32 |
| Facebook47% reach | 33 |
| Facebook Messenger27% reach | 10 |
| iMessage22% reach | 5 |
| X (Twitter)22% reach | 43 |
| Discord15% reach | 12 |
| Twitch11% reach | 6 |
| Reddit11% reach | 26 |
| Android Message8% reach | 6 |
| Telegram7% reach | 21 |
Question wording & response options
ContentSaw violent content on the platformDenominator: users of that platformReported here: any exposure in the past 12 months
Platform usage: “Q28a. Have you used any of the following platforms in the past 12 months?” Please select all that apply (17 platforms listed). Technical report p.44.
Violent content, per platform: “Q28b. Have you seen any violent content on the following platforms in the past 12 months? Again, we’re interested in real-world violence (not including content in the news, TV, video games or films), including people in your local area or that you know.” Yes / No / Not sure, per platform. Technical report p.45.Routed [Only show platforms selected @Q28_A] — the grid shows only the platforms the child selected at the previous question, which is what makes this a per-user rate rather than an attribution share.
Report p.17 (printed p.17); same values listed on p.7. “Figure 1.3: Proportion of children seeing violent content on social media platforms”, right-hand panel, with rates, usage and bases read from Data tables 3.21 and 3.22 (sheets 73 and 74). Darker = higher. Apps are ordered by weighted user n, largest first, with alphabetical ties; app reach (shown after each platform) is the share of surveyed 13–17-year-olds in England and Wales who used it in the past 12 months. Hover a cell for its base. The instrument asks about 17 platforms; Vimeo, Tumblr, Viber are omitted here because fewer than 50 of their users answered yes, below the floor YEF sets for reporting a figure at all. No margin of error is published with these data.
United Kingdom: Ofcom Children’s Online Safety Tracker (2025)
UK·ages 11-17·n = 1,652 (Module 2)·Non-probability·17 apps·4-week recall·sources
Ofcom’s Children’s Online Safety Tracker asks UK children aged 11–17 whether they saw harmful content on each of 17 named services in the last four weeks, on a weighted non-probability online panel. The Wave 2 harms module includes 1,652 children, so per-service usage bases run from 126 to 1,158. Wave 2 covers 38 content types, including violent content, self-harm or suicide content, eating disorder content, and pornography. While they also ask about bullying, the primary framing is around witnessing bullying content (not bullying victimization), so we do not include that in our summary here.
Per-app prevalence rates of harmful content exposure among each app’s users, last 4 weeks (Wave 2)
Click a app heading to sort harm rows, or a harm heading to sort app columns. Click again to reverse the order.
| Content type | n=1158 | n=1114 | n=1073 | n=929 | n=842 | n=732 | n=697 | n=544 | n=536 | n=341 | n=306 | n=258 | n=227 | n=223 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Content284 exposed | 5.2 | 2.9 | 8.2 | 4.4 | 5.3 | 10.5 | 1.9 | 2.9 | 0.7 | 5.0 | 2.6 | 4.7 | 11.0 | 2.7 |
| Content196 exposed | 3.7 | 2.2 | 5.8 | 3.1 | 5.2 | 6.7 | 1.1 | 3.1 | 1.5 | 2.6 | 3.6 | 3.5 | 8.4 | 1.3 |
| Content482 exposed | 15.3 | 2.3 | 19.3 | 6.5 | 9.9 | 10.8 | 2.7 | 2.8 | 2.1 | 4.1 | 6.5 | 5.8 | 11.0 | 4.5 |
| Content180 exposed(Only older children asked) | 3.4 | 1.7 | 7.6 | 3.7 | 4.9 | 4.8 | 1.5 | 2.9 | 0.9 | 1.7 | 1.9 | 3.1 | 6.8 | 2.1 |
| Content92 exposed(Only older children asked) | 2.0 | 0.6 | 3.9 | 2.1 | 3.2 | 3.2 | 1.4 | 1.2 | 1.4 | 1.7 | 0.5 | 2.2 | 4.9 | 1.6 |
| Content185 exposed | 3.7 | 0.6 | 7.0 | 2.5 | 4.2 | 4.8 | 0.7 | 2.4 | 1.1 | 2.3 | 2.9 | 4.3 | 4.0 | 2.7 |
| Content246 exposed | 5.6 | 1.8 | 8.7 | 3.2 | 6.5 | 8.1 | 2.9 | 2.4 | 1.5 | 1.8 | 6.5 | 2.7 | 5.7 | 4.0 |
| Content182 exposed | 1.9 | 1.5 | 3.5 | 3.2 | 4.4 | 4.0 | 0.7 | 1.8 | 0.7 | 1.5 | 2.3 | 2.7 | 8.8 | 1.8 |
Question wording & response options
Usage (A1): “A1. Which of these have you used in the last 4 weeks? (Even if you have stopped using them, please tick if you have used them during this time). You can choose more than one.” questionnaire p.13Counted from data tables, Table 29 (sheet ONLINEUSE), column “Module 2”.
Exposure (M2.2): “M2.2. Have you seen or heard any of the below online in the last 4 weeks? This could have been a video, picture, written post, a comment or something you have heard You can choose more than one, if more than one is shown.” questionnaire p.28Asked as ten randomised blocks covering 38 content items. The eight this table uses are listed below, each with the wording the child saw.
Platform (M2.5): “M2.5. Where did you see [content]? You can choose more than one.” questionnaire p.36The M2.5 answer list is piped from the child’s own A1 usage list, with an additional “Somewhere else” write-in option. We divide the published content-specific app counts by each app’s user base to derive rates.
The eight rows, as the child saw them
| Row | M2.2 item, verbatim | Figures from |
|---|---|---|
| Serious violence | “Someone being seriously violent towards another person or encouraging others to do so” | questionnaire p.29·data tables, Table 155.5 |
| Serious injuries caused by others | “A person’s serious injuries that were caused by others” | questionnaire p.29·data tables, Table 155.6 |
| Dangerous stunts or challenges | “Dangerous stunts or challenges online” | questionnaire p.29·data tables, Table 155.9 |
| Self-harm | “Self harm / hurting yourself on purpose: people showing this, giving instructions for this, or encouraging others to do this”[only older children asked — see footnote] | questionnaire p.31·data tables, Table 155.34 |
| Suicide | “Suicide (taking one’s own life): people showing this, giving instructions for this, or encouraging others to do this”[only older children asked — see footnote] | questionnaire p.31·data tables, Table 155.33 |
| Eating disorder | “Showing someone with an eating disorder, or encouraging or providing instructions for one” | questionnaire p.30·data tables, Table 155.31 |
| Extreme eating or exercise | “Someone showing or encouraging extreme eating or extreme exercise behaviours, or providing instructions for them” | questionnaire p.30·data tables, Table 155.32 |
| Pornography | “Pornography” | questionnaire p.30·data tables, Table 155.30 |
On bullying: M2.2’s first block asks about seeing bullying, trolling, griefing and exclusion content. A follow-up (M2.7) asks whether the content was “about you or someone else”, and for bullying content 85% answer “someone else” against 3% “me” (Table 170.1, base 156), so the instrument as designed does not provide sufficient responses on a per-app basis for bullying victimization.
Derived from the Ofcom COST Wave 2 data tables, Table 154 (sheet Module2), with the exception of the suicide and self-harm items. Numerator = weighted children naming the service for that content item at M2.5; denominator = that service’s weighted users at A1, or — for the two rows only older children were asked — its users among those who were asked. Restricted to the 14 services with a Wave 2 user base of 200 or more, which leaves out Messenger Kids (150), Reddit (138) and Telegram (126) — Telegram would sit at or near the top of several rows on this same method. Cells rest on weighted numerators as small as 1, and 39 of the 112 are below 10, so comparisons involving small cells should be treated as indicative. Ofcom publishes weighted counts rounded to whole numbers; these do not always match the raw app-selection records. We use the published counts, so these derived rates are approximate. Darker = higher, on one shared scale; “exposed” is the weighted number of children who saw that content type anywhere, not the cell base. Per-row wording and sources are in the panel above. Ofcom’s own service-level analysis of any harmful content is in Children’s Online Experiences, Figure 11 p.24.The suicide and self-harm items were shown only to children in Year 8 (England and Wales), S2 (Scotland), Year 9 (Northern Ireland) and above — 1,415 of the 1,652 children in the harms module, corresponding to a weighted base of 1,346 (data tables, Table 116). The prevalence rates for these items are derived from the Wave 2 SPSS microdata to reflect the smaller denominator, as opposed to published Ofcom report tables.
Australia: eSafety Keeping Kids Safe Online (2025)
AU·ages 10-15·n = 2,629·Non-probability·24 apps·lifetime recall·sources
eSafety’s Digital use and risk report breaks eight harms out by app on a lifetime window; we chart six of them — five conduct-and-contact harms and one aggregate content harm — among Australian children aged 10–15 on a non-probability panel. Unlike the studies above, it reports attribution — the share of people who experience a harm who attribute it to a given app (e.g. of all people who experienced bullying, what % experienced it on a specific app), rather than prevalence, the share of people who use an app who report experiencing the harm on the app (e.g. of all people who use a specific app, what % experienced bullying on that app).
We can (approximately) convert attribution to prevalence because eSafety also reports app usage on the same lifetime window (see “how the conversion works” for more details).
These figures adjust for how many children have ever used each app, but not for dosage — how frequently or intensely they use it. Children were asked to attribute only their most recent or most impactful experience, so harms on other apps may be omitted; thus, prevalence may be underestimated and app rankings may be distorted. These figures are not directly comparable to other prevalence measures on this page.
The table below applies this conversion to six of the eight harms eSafety breaks out by platform; image-based abuse (Table 19, 79 victims) and seeing — as opposed to experiencing — online hate (Table 24) are left out. Snapchat is highest on all five conduct-and-contact harms — an implied 22.6% of its users for cyberbullying and 18.3% for sexual harassment. YouTube, the most-used named platform in the sample at 76% lifetime use, sits below TikTok, Instagram and Snapchat on every one of those five, and is first on harmful content at 34.6%, where Snapchat is the lowest of the five major social platforms at 7.7%.
Implied per-app prevalence rates of harms among each app’s users
Click a harm heading to sort app rows, or a app heading to sort harm columns. Click again to reverse the order. Harm columns sort within their category groups.
| Conduct & Contact | Content | |||||
|---|---|---|---|---|---|---|
| App (lifetime use) | 52% overall | 25% overall | 23% overall | 14% overall | 24% overall | 71% overall |
| 86% | 13.3 | 4.4 | 0.5 | 2.4 | 2.2 | 3.3 |
| 76% | 2.1 | 2.3 | 2.1 | 0.6 | 0.9 | 34.6 |
| 56% | 3.7 | 7.6 | 4.9 | 2.8 | 3.9 | 29.2 |
| 47% | 6.6 | 6.9 | 7.8 | 5.1 | 7.1 | 12.1 |
| 46% | 22.6 | 11.4 | 11.0 | 7.0 | 18.3 | 7.7 |
| 44% | 1.2 | 0.6 | 1.6 | 0.3 | 1.1 | 0.0 |
| 42% | 7.4 | 9.5 | 8.2 | 3.7 | 4.6 | 18.6 |
| 40% | 20.8 | 3.1 | 9.8 | 1.1 | 1.8 | 1.8 |
| 32% | 8.1 | 2.3 | 8.6 | 1.3 | 1.5 | 2.2 |
| 29% | 5.4 | 0.9 | 5.6 | 1.0 | 3.3 | 4.9 |
| 26% | 2.0 | 1.0 | 2.7 | 1.1 | 1.8 | 2.7 |
| 22% | 11.8 | 8.0 | 4.2 | 5.7 | 10.9 | 3.2 |
| 21% | 7.4 | 1.2 | 5.5 | 2.0 | 4.6 | 0.0 |
| 20% | 0.0 | 1.3 | 1.1 | 0.7 | 0.0 | 0.0 |
How the conversion works
Let N be the number of people in the sample; H be the number of people who experienced the harm overall, across all apps; A be the number of people who attribute a harm to an app; and U be the number of people who use an app.
We want to derive app harm prevalence (AU). eSafety publishes overall harm prevalence (HN), app harm attribution (AH), and app usage (UN).
Two cancellations have to hold. Neither holds exactly.
N cancels only if usage and overall prevalence come from the same sample. The window matches — both are lifetime on the same panel — but the samples do not: usage is measured on all 2,629 children, while three harms sit on smaller bases because of skipped questions and, for online hate, a smaller recontact survey (online hate 1,594, grooming 1,922, sexual harassment 1,816). eSafety publishes counts but no demographics for those who declined, so the direction of that bias is unknown rather than conservative.
H cancels only if attribution is measured on all of a harm’s victims, and it is not. eSafety states that “to minimise the burden on children who had experienced multiple harm-types, they were only asked follow-up questions about one type” (report p.4). So AH rests on a routed subsample: the platform question reached 42% of the children the survey implies were cyberbullied and 30% of those exposed to harmful content, rising to 83% for online hate. This routing may change the mix of children represented in each column; the coverage figures alone do not establish the direction of that bias.
Question wording & response options & derived app prevalence
Each harm is asked as a battery of options, but the platform question is asked once, about the harm overall. eSafety publishes prevalence for the individual options — harmful content’s eleven are in Table 16, report p.16 — but not attribution for them, so a per-option rate per app cannot be derived. Every column in this section is therefore the net for that harm, as the report presents it.
Harm: “Has anyone online ever done any of the following things to you on purpose because they really wanted to humiliate you or make you feel bad, or make others not like you?” (12 options)
- Had someone say hurtful things to them online
- Were purposely left out of an online activity
- Had humiliating or hurtful things said about them online
- Had humiliating or hurtful lies spread about them online
- Had their private messages, information or secrets shared with other people
- Were sent or tagged in offensive or upsetting photos or videos
- Were told online to hurt or kill themselves, or that they should die
- Had someone say that they were going to hurt or kill them or told other people to hurt or kill them
- Had humiliating or hurtful photos or videos of them shared online
- Had their phone number, email or home address shared online
- Had humiliating or hurtful fake photos or videos made of them and shared online
- Had a fake online identity or profile made of them
Platform: “Which platform or platforms did this happen to you on? You can choose more than one answer.”
| App | |||
|---|---|---|---|
| Online video game | 22% | 86% | 13.3% |
| YouTube | 3% | 76% | 2.1% |
| TikTok | 4% | 56% | 3.7% |
| 6% | 47% | 6.6% | |
| Snapchat | 20% | 46% | 22.6% |
| FaceTime | 1% | 44% | 1.2% |
| 6% | 42% | 7.4% | |
| Messenger Kids | 16% | 40% | 20.8% |
| Text messages | 5% | 32% | 8.1% |
| 3% | 29% | 5.4% | |
| 1% | 26% | 2.0% | |
| Discord | 5% | 22% | 11.8% |
| Messenger | 3% | 21% | 7.4% |
| 0% | 20% | 0.0% |
Harm: “they had someone say or post offensive or threatening things to or about them online because they were hateful of one or more of their identity characteristics”
Platform: “Which platform or platforms did this happen to you on? You can choose more than one answer.”
The report prints no question wording for this item. This is its own description of the measure (p.19 / printed p.36). Identity characteristics itemised on the same page: skin colour, race or culture; sexuality; country of origin; gender identity; religion; disability; being a girl; being a boy; being Aboriginal or Torres Strait Islander.
| App | |||
|---|---|---|---|
| Online video game | 15% | 86% | 4.4% |
| YouTube | 7% | 76% | 2.3% |
| TikTok | 17% | 56% | 7.6% |
| 13% | 47% | 6.9% | |
| Snapchat | 21% | 46% | 11.4% |
| FaceTime | 1% | 44% | 0.6% |
| 16% | 42% | 9.5% | |
| Messenger Kids | 5% | 40% | 3.1% |
| Text messages | 3% | 32% | 2.3% |
| 1% | 29% | 0.9% | |
| 1% | 26% | 1.0% | |
| Discord | 7% | 22% | 8.0% |
| Messenger | 1% | 21% | 1.2% |
| 1% | 20% | 1.3% |
Harm: “Non-consensual tracking, monitoring or harassment” (3 options)
- Had someone use an app or a device to track their exact real-time location without their permission (7% ever / 4% past 12 months)
- Had someone look through their phone (or another device) or their social media or read their messages without their permission (17% / 12%)
- Felt worried or scared because someone kept sending them messages, calling them or posting on their social media page, and they wouldn’t stop even when asked to (10% / 5%)
Platform: “Which platform or platforms did this happen to you on? You can choose more than one answer.”
The report prints no question wording for this harm; its three components appear only as body-text bullets on p.9 (printed p.17). Footnote 11 records that children were told to answer "no" where the tracking or monitoring was by "someone like a parent or caregiver doing this to keep you safe, and you were comfortable with this" — though eSafety notes some children may still have reported such experiences.
| App | |||
|---|---|---|---|
| Online video game | 2% | 86% | 0.5% |
| YouTube | 7% | 76% | 2.1% |
| TikTok | 12% | 56% | 4.9% |
| 16% | 47% | 7.8% | |
| Snapchat | 22% | 46% | 11.0% |
| FaceTime | 3% | 44% | 1.6% |
| 15% | 42% | 8.2% | |
| Messenger Kids | 17% | 40% | 9.8% |
| Text messages | 12% | 32% | 8.6% |
| 7% | 29% | 5.6% | |
| 3% | 26% | 2.7% | |
| Discord | 4% | 22% | 4.2% |
| Messenger | 5% | 21% | 5.5% |
| 1% | 20% | 1.1% |
Harm: “Has an adult or someone at least 4 or 5 years older than you ever …?” (6 options)
- Had an adult or someone at least 4 or 5 years older do or say something creepy or inappropriate to them
- Had an adult or someone at least 4 or 5 years older ask them about private parts of their body
- Had an adult or someone at least 4 or 5 years older ask them to send them a nude image of themselves
- Had an adult or someone at least 4 or 5 years older send them a nude image of themselves
- Had an adult or someone at least 4 or 5 years older send them a nude image of other people
- Had an adult or someone at least 4 or 5 years older offer them money or gifts in exchange for a nude image of themselves
Platform: “Which platform or platforms did this happen to you on? You can choose more than one answer.”
| App | |||
|---|---|---|---|
| Online video game | 15% | 86% | 2.4% |
| YouTube | 3% | 76% | 0.6% |
| TikTok | 11% | 56% | 2.8% |
| 17% | 47% | 5.1% | |
| Snapchat | 23% | 46% | 7.0% |
| FaceTime | 1% | 44% | 0.3% |
| 11% | 42% | 3.7% | |
| Messenger Kids | 3% | 40% | 1.1% |
| Text messages | 3% | 32% | 1.3% |
| 2% | 29% | 1.0% | |
| 2% | 26% | 1.1% | |
| Discord | 9% | 22% | 5.7% |
| Messenger | 3% | 21% | 2.0% |
| 1% | 20% | 0.7% |
Harm: “Has anyone ever …?” (7 options)
- Were called an offensive sexual name online (asked only of children 13+, n=974)
- Were sent sexual messages or had sexual comments made about them online that they didn’t want
- Were asked sexual questions or had someone try to get them to talk about sex online when they didn’t want to
- Had sexual rumours or lies spread about them online (13+ only)
- Were sent a nude image of someone that they didn’t want
- Had someone try to pressure them online into sending a nude image of themselves when they didn’t want to
- Had their private sexual information shared online (13+ only)
Platform: “Which platform did this happen to you on?”
| App | |||
|---|---|---|---|
| Online video game | 8% | 86% | 2.2% |
| YouTube | 3% | 76% | 0.9% |
| TikTok | 9% | 56% | 3.9% |
| 14% | 47% | 7.1% | |
| Snapchat | 35% | 46% | 18.3% |
| FaceTime | 2% | 44% | 1.1% |
| 8% | 42% | 4.6% | |
| Messenger Kids | 3% | 40% | 1.8% |
| Text messages | 2% | 32% | 1.5% |
| 4% | 29% | 3.3% | |
| 2% | 26% | 1.8% | |
| Discord | 10% | 22% | 10.9% |
| Messenger | 4% | 21% | 4.6% |
| 0% | 20% | 0.0% |
Harm: “Have you ever seen or heard any of the following things online?” (11 options)
- Offensive, sexist or hurtful things about girls or women
- Fight videos
- Dangerous online challenges
- Things that encourage unhealthy eating or exercise habits
- Offensive or threatening things about other people online because others are hateful of their identity
- Sexual images or videos
- Things that show or encourage illegal drug taking
- Extreme real-life violence
- Things that suggest how a person can hurt or kill themselves on purpose
- Violent sexual images or videos
- Something else upsetting
Platform: “Which platform did you see it on?”
| App | |||
|---|---|---|---|
| Online video game | 4% | 86% | 3.3% |
| YouTube | 37% | 76% | 34.6% |
| TikTok | 23% | 56% | 29.2% |
| 8% | 47% | 12.1% | |
| Snapchat | 5% | 46% | 7.7% |
| FaceTime | 0% | 44% | 0.0% |
| 11% | 42% | 18.6% | |
| Messenger Kids | 1% | 40% | 1.8% |
| Text messages | 1% | 32% | 2.2% |
| 2% | 29% | 4.9% | |
| 1% | 26% | 2.7% | |
| Discord | 1% | 22% | 3.2% |
| Messenger | 0% | 21% | 0.0% |
| 0% | 20% | 0.0% |
Derived from eSafety DUR 2025, Tables 7, 21, 9, 12, 15 and 18 (attribution among victims), Tables 1 and 3 plus report p.13 (lifetime platform usage), and the lifetime prevalence reported for each harm. Restricted to the 14 platforms used by at least 1 in 5 children aged 10–15 and offered in the attribution tables (YouTube Kids reaches 44% but eSafety excludes it from all platform analysis); below that the denominator is too small for the ratio to be stable. Darker = higher, on one shared scale across all six columns. For more details on the derivation, see “how the conversion works” above. Note that the derivation is approximate since eSafety asks where the most recent or most impactful experience happened, so a child harmed multiple different times only contributes one experience. Multiple apps can also be named for one experience. Additionally, prevalence bases differ by harm — children could skip the sensitive blocks, and one harm (online hate) was measured in a smaller recontact survey, so the bases range from 1,594 to 2,629. Attribution is rounded to whole percentage points and rests on victim bases of 215–569, so small values are coarse: YouTube’s 3% on grooming implies anywhere from about 0.5% to 0.6%. The decimals are approximate.