A 2023 transparency report shows only 0.9% of reviewed content for hateful behavior resulted in visible enforcement. This article examines the data behind detection technologies, platform policies, and the scale of impact, including the 2.3 million U.S. youth who experienced cyberbullying in a given year.
Key Takeaways
- In X’s 2023 transparency report, the company reported enforcing actions for 0.9% of the global content visibility it reviewed for hateful/harassing behavior
- Meta’s transparency reporting shows that, in a recent quarter (Q4 2023), automated systems detected and initiated action for the majority of policy enforcement cases (72% initiation rate)
- In the EU, the average notice-and-action requirement under the DSA includes removing or disabling access to illegal content with “diligence” timelines specified by procedure
- UK Online Safety Act 2023 requires in-scope services to assess and mitigate risks of harm, including harassment, with duties commencing for services in phased timelines (2024–2025)
- Google’s Perspective API study reported that its toxicity detection model could identify toxic comments with an AUROC of up to 0.93 on benchmark datasets
- A 2020 paper on cyberbullying detection reported F1-scores between 0.70 and 0.86 depending on model and dataset
- A 2021 systematic review reported that machine-learning approaches for cyberbullying detection commonly achieve accuracy in the 70%–90% range on benchmark datasets
- In the U.S., 2.3 million children and teens experienced cyberbullying in a given year (youth self-report estimate)
- In a 2017 meta-analysis, online victimization showed a moderate association with depressive symptoms (standardized mean difference around 0.30)
- A 2019 peer-reviewed study found that cyberbullying victimization increased odds of suicidal ideation by approximately 2x (odds ratio reported as ~2.1)
- 11% of U.S. teens reported that they have been harassed or bullied online in the past year and it caused them distress severe enough to affect schoolwork/grades (survey-based severity indicator)
- 37% of LGBTQ+ youth reported experiencing cyberbullying or online harassment (percentage reporting online harassment within a defined 12-month reference period)
- 64% of social media users in the UK believed platforms should remove harmful content more quickly, indicating strong demand for faster moderation pipelines (survey-based opinion metric)
- EU DSA requires very large online platforms to mitigate identified systemic risks within set timeframes following risk assessment (time-bound mitigation obligation)
- UK Online Safety Act 2023 creates a requirement for in-scope services to carry out risk assessments for harmful content including harassment and cyberbullying, including publication of certain risk-related information (duty with measurable scope)
Cyberbullying affects millions, and faster automated detection plus better moderation responses can reduce harm.
Related reading
01 · Category
Reporting And Response1 stats
01
In X’s 2023 transparency report, the company reported enforcing actions for 0.9% of the global content visibility it reviewed for hateful/harassing behavior
Interpretation
Reporting And Response Interpretation
In X’s 2023 transparency reporting, the platform enforced actions on just 0.9% of the global content visibility it reviewed for hateful or harassment, suggesting that for the reporting and response category, only a small share of flagged cyberbullying related content leads to enforcement.
02 · Category
Market And Policy9 stats
01
Meta’s transparency reporting shows that, in a recent quarter (Q4 2023), automated systems detected and initiated action for the majority of policy enforcement cases (72% initiation rate)
02
In the EU, the average notice-and-action requirement under the DSA includes removing or disabling access to illegal content with “diligence” timelines specified by procedure
03
UK Online Safety Act 2023 requires in-scope services to assess and mitigate risks of harm, including harassment, with duties commencing for services in phased timelines (2024–2025)
04
U.S. states have introduced at least 60 bills related to cyberbullying or social platform responsibility between 2019 and 2023 (tracked legislative initiatives count)
05
European Commission reports that platforms subject to DSA transparency obligations include about 19 very large platforms as of 2023 (designation count)
06
In 2020, the UN Guiding Principles on Business and Human Rights referenced the responsibility to prevent and mitigate adverse human rights impacts (including online abuse) for companies operating digitally
07
The U.S. STOP CSAM Act updates definition and platform obligations; in the cyberbullying context, it is part of broader online safety legislation affecting platform reporting requirements with deadlines set by DOJ guidance
08
Australia’s Online Safety Act 2021 established a framework with enforceable obligations for platforms to respond to cyberbullying and harmful content, with penalties up to AU$555,000per breach
09
Germany’s NetzDG law required platforms to remove illegal content within 24 hours for certain complaints and within 7 days for others (statutory deadlines)
Interpretation
Market And Policy Interpretation
Across Market and Policy, 2023 reporting and regulation efforts are converging fast as the EU applies notice-and-action duties, the UK Online Safety Act 2023 pushes platforms to assess harassment risks, and the US has seen at least 60 cyberbullying or platform responsibility bills from 2019 to 2023.
03 · Category
Mitigation Technologies5 stats
01
Google’s Perspective API study reported that its toxicity detection model could identify toxic comments with an AUROC of up to 0.93 on benchmark datasets
02
A 2020 paper on cyberbullying detection reported F1-scores between 0.70 and 0.86 depending on model and dataset
03
A 2021 systematic review reported that machine-learning approaches for cyberbullying detection commonly achieve accuracy in the 70%–90% range on benchmark datasets
04
In a 2019 paper evaluating content moderation, applying blocking filters reduced repeat exposure to harassing content by 33% in controlled tests
05
A 2018 peer-reviewed study reported that human-in-the-loop moderation improved cyberbullying detection precision by 14 points compared with fully automated labeling
Interpretation
Mitigation Technologies Interpretation
Mitigation technologies are getting measurably better, with toxicity and cyberbullying detection models reaching AUROC up to 0.93 and F1 scores up to 0.86, while targeted moderation approaches like blocking filters cut repeat exposure to harassing content by 33% and human in the loop review boosts precision by 14 points.
04 · Category
Impacts And Costs9 stats
01
In the U.S., 2.3 million children and teens experienced cyberbullying in a given year (youth self-report estimate)
02
In a 2017 meta-analysis, online victimization showed a moderate association with depressive symptoms (standardized mean difference around 0.30)
03
A 2019 peer-reviewed study found that cyberbullying victimization increased odds of suicidal ideation by approximately 2x (odds ratio reported as ~2.1)
04
A 2020 review reported that cyberbullying is associated with increased anxiety, with effect sizes in the small-to-moderate range (Hedges g roughly 0.20–0.45 across included studies)
05
In a 2018 report, 44% of students who experienced cyberbullying reported that it negatively affected their school work/ability to focus
06
A 2021 U.S. survey found that 15% of teens who were cyberbullied said it affected their attendance (skipped school or cut back)
07
In a 2022 study, cyberbullying victims reported higher rates of self-harm thoughts, with a reported prevalence difference of 12 percentage points between victims and non-victims
08
In a 2023 Gartner forecast, the global market spend on content moderation solutions is projected to reach $6.8 billion by 2026
09
According to OECD, victimization impacts educational outcomes, with an estimated 10% performance decline associated with bullying exposure (cross-country evidence)
Interpretation
Impacts And Costs Interpretation
For the impacts and costs of social media cyberbullying, evidence shows it is not just harmful emotionally but disrupts daily life, since 44% of affected students reported problems with school work or focus and 15% of cyberbullied teens said it led them to skip school or cut back attendance.
05 · Category
Prevalence Estimates2 stats
01
11% of U.S. teens reported that they have been harassed or bullied online in the past year and it caused them distress severe enough to affect schoolwork/grades (survey-based severity indicator)
02
37% of LGBTQ+ youth reported experiencing cyberbullying or online harassment (percentage reporting online harassment within a defined 12-month reference period)
Interpretation
Prevalence Estimates Interpretation
For the Prevalence Estimates category, the data suggests that cyberbullying is widespread, with 11% of U.S. teens reporting online harassment that caused severe distress and 37% of LGBTQ+ youth reporting cyberbullying or online harassment within a 12-month period.
More related reading
06 · Category
Policy And Compliance4 stats
01
64% of social media users in the UK believed platforms should remove harmful content more quickly, indicating strong demand for faster moderation pipelines (survey-based opinion metric)
02
EU DSA requires very large online platforms to mitigate identified systemic risks within set timeframes following risk assessment (time-bound mitigation obligation)
03
UK Online Safety Act 2023 creates a requirement for in-scope services to carry out risk assessments for harmful content including harassment and cyberbullying, including publication of certain risk-related information (duty with measurable scope)
04
US FTC reported that it brought 7 enforcement actions related to deceptive or unfair practices involving children/teens online privacy and safety over a defined multi-year window (enforcement quantity relevant to platform duties affecting youth safety)
Interpretation
Policy And Compliance Interpretation
Policy and compliance is becoming more urgent across regions, with 64% of UK users urging faster removal of harmful content and EU and UK laws requiring structured risk assessments and mitigation steps, while US enforcement by the FTC underscores that privacy and safety practices involving minors are actively policed through actions.
07 · Category
Market Size3 stats
01
$4.7 billion projected trust and safety market size in 2025 (forecasted market value for trust/safety technologies used by platforms to reduce abuse)
02
12.6% compound annual growth rate (CAGR) forecast for the content moderation market through 2029 (growth rate metric for a core industry segment addressing harassment/cyberbullying)
03
$2.7 billion global market value for safety technologies used in online moderation in 2023 (spending metric for safety tooling relevant to cyberbullying mitigation)
Interpretation
Market Size Interpretation
The market size for social media cyberbullying defenses is expanding fast, with a projected $4.7 billion trust and safety market in 2025 and a 12.6% CAGR content moderation growth through 2029, building on $2.7 billion in online safety technology spending in 2023.
08 · Category
Operational Metrics3 stats
01
58% of trust & safety teams reported that they can’t keep up with volume without automation (operational capacity metric tied to moderation workflow constraints)
02
93% of companies in trust & safety reported using some form of automated detection to handle harmful or policy-violating content (automation adoption metric)
03
Automation reduced review queues by 45% in a controlled pilot described in an industry case study (queue reduction metric)
Interpretation
Operational Metrics Interpretation
Operational metrics show a clear push toward automation as 58% of trust and safety teams say they cannot keep up with content volume without it, while 93% already use automated detection and a pilot cut review queues by 45%, reducing operational strain on moderation.
09 · Category
Research Findings3 stats
01
Machine learning classifiers for harassment detection achieve F1-scores of 0.75 to 0.90 on social-media benchmark datasets in a 2023 comparative evaluation (model performance range metric)
02
A 2022 meta-analysis reported that interventions combining education with moderation reporting mechanisms reduced cyberbullying victimization by an average standardized effect size of approximately g = -0.30 (aggregate effect metric)
03
In a 2020 observational study, victims who received timely responses from platform reporting flows showed a 19% lower likelihood of repeated targeted harassment (recidivism reduction metric)
Interpretation
Research Findings Interpretation
Research on social media cyberbullying shows that detection models are reaching F1-scores between 0.75 and 0.90 on benchmark datasets in 2023, while studies further suggest that well-designed interventions and faster platform reporting responses are linked to lower victimization rates and up to a 19% reduction in repeat harm.
report visual · Key figures
Cyberbullying & harassment signals: platform detection, enforcement, and impact
High rates of automated enforcement initiation are paired with measurable downstream effects on teens’ distress and school performance.
72%
Meta’s transparency reporting shows that, in a recent quarter (Q4 2023), automated systems detected and initiated action
0.9%
In X’s 2023 transparency report, the company reported enforcing actions for 0.9% of the global content visibility it rev
11%
11% of U.S. teens reported that they have been harassed or bullied online in the past year and it caused them distress s
44%
In a 2018 report, 44% of students who experienced cyberbullying reported that it negatively affected their school work/a
source-verifiedtransparency.meta.com · transparency.x.com · samhsa.gov · oecd.org2023
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Catherine Wu. (2026, February 13). Social Media Cyberbullying Statistics. Gitnux. https://gitnux.org/social-media-cyberbullying-statistics
MLA
Catherine Wu. "Social Media Cyberbullying Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/social-media-cyberbullying-statistics.
Chicago
Catherine Wu. 2026. "Social Media Cyberbullying Statistics." Gitnux. https://gitnux.org/social-media-cyberbullying-statistics.
Sources & references
39 datasets cited across this report · attribution is report-level
+7 additional datasets cited (not shown individually)