Methodology

GuardKids grades platforms on child safety using publicly available data. Here's exactly how we do it — our dimensions, scoring rubric, data sources, and update process.

Grading Dimensions

Each platform is evaluated across four dimensions. The final grade is a weighted composite of all four:

CSAM Reporting Volume

30%

How many CyberTipline reports the platform generates relative to its user base. High absolute numbers alone don't mean a platform is unsafe — a platform that proactively detects and reports CSAM is doing its job. We normalize by monthly active users and consider year-over-year trends.

AStrong detection + declining reports relative to user growth
BConsistent reporting with moderate detection improvements
CReporting volume growing faster than user base
DDisproportionately high reports or major detection gaps
FMinimal or zero reporting despite known exploitation

Safety Features

25%

What tools and protections the platform provides — parental controls, age verification, content moderation AI, communication restrictions for minors, and proactive scanning technologies like PhotoDNA or perceptual hashing.

AComprehensive controls, proactive scanning, strong age verification
BGood controls with some gaps in enforcement or coverage
CBasic features exist but require manual activation or have loopholes
DMinimal protections, safety features are opt-in and hard to find
FNo meaningful safety features for minors

Enforcement Response

25%

How the platform responds to reported exploitation — speed of content removal, account suspension, cooperation with law enforcement, and track record with NCMEC and DOJ investigations.

ARapid response, proactive law enforcement cooperation, dedicated trust & safety team
BGenerally responsive with room for improvement on speed
CInconsistent response times, selective cooperation
DSlow response, adversarial relationship with law enforcement
FNon-responsive or actively obstructive

Transparency

20%

How openly the platform communicates about child safety — publishing transparency reports, disclosing CSAM metrics, participating in industry coalitions (like the Tech Coalition), and supporting independent research.

ADetailed transparency reports, independent audits, coalition participation
BRegular transparency reports with reasonable detail
CSome public data but significant gaps
DMinimal disclosure, resists transparency
FNo transparency reports or public data

Data Sources

NCMEC CyberTipline

Annual and quarterly CSAM report counts by ESP (electronic service provider)

Visit ↗

FBI IC3

Internet crime complaints including sextortion, trafficking, and online exploitation

Visit ↗

Platform Transparency Reports

Self-reported content moderation, account actions, and safety metrics

Internet Watch Foundation (IWF)

UK-based CSAM analysis, AI-generated CSAM statistics

Visit ↗

Thorn

Research on technology's role in child exploitation

Visit ↗

Polaris Project

National Human Trafficking Hotline data

Visit ↗

State Sex Offender Registries

Registered sex offender counts by state via NSOPW

Visit ↗

Congressional Research Service

Legislation analysis and bill tracking

Update Frequency

Platform Grades

Updated quarterly when new transparency reports and NCMEC data become available. Major incidents (lawsuits, breaches, policy changes) may trigger interim updates.

NCMEC & FBI Data

Updated annually when official reports are published (typically Q1 for prior year data). Preliminary data may be included with notation.

Legislation

Tracked continuously during legislative sessions. Status changes reflected within 1 week of official action.

Sex Offender Registry

State-level data updated annually. National totals aggregated from individual state registries and NSOPW.

Limitations

  • NCMEC reports ≠ unique incidents. One image can generate multiple reports. Platforms with better detection technology may report more — which can look worse in raw numbers.
  • Self-reported data. Platform transparency reports are self-published and not independently audited. We note where data may be incomplete or biased.
  • Grades are relative. No platform earns an A by being perfect — they earn it by being measurably better than peers across all four dimensions.
  • Subjectivity in enforcement scoring. Enforcement response is partially assessed through reported experiences, lawsuits, and investigative journalism rather than purely quantitative data.
If you suspect child abuse:📞 1-800-843-5678