Trust center
Our methodology is designed to help parents understand visible signals across digital surfaces. We use public disclosures, documented product changes, reviewed evidence, and admin approval before publishing updates.
Rooted Reality publishes public-source posture records. We use cited public sources, admin review, and a consistent methodology. We do not make legal conclusions or claims based on private household data.
Agents may draft posture updates. Admin approval is required before publication.
input
Public sources
agent
Agent drafts
human
Admin reviews
output
Published
Rooted Reality reviews digital surfaces — apps, communities, AI products, and model versions — that families with kids and teens may need to understand. The unit of review is a surface or a release, not a company. A company can ship multiple surfaces, and a surface can change posture between versions; the library tracks each one separately.
Rooted Reality uses a signals-based methodology. We review what a digital surface publicly says, what it documents, and what changes over time. Scores summarize visible posture across four axes: moderation, privacy, security, and youth protection. Each score includes evidence and confidence so parents can understand both the signal and its limits.
What we measure
How the surface handles content rules, abuse reports, and harmful behavior at scale.
How we review
We read public policies, transparency reports, abuse-handling disclosures, and documented behavior changes that affect what kids see.
What we measure
What personal information the surface collects, how it is used, and who it is shared with.
How we review
We read privacy policies, data-handling disclosures, app-store privacy labels, and documented changes to data sharing or retention.
What we measure
How the surface protects accounts, devices, and stored data from compromise.
How we review
We read public security disclosures, breach history, account-protection documentation, and changes that affect stored content or session integrity.
What we measure
Whether the surface ships specific protections for kids and teens out of the box.
How we review
We read youth-specific policies, age-band defaults, parental-tooling documentation, and changes that affect age-appropriate behavior.
Every published posture summary cites public sources. The kinds of sources we accept are listed below. Each kind carries a different weight depending on the axis it supports.
A posture summary requires at least one official vendor source per scored axis. Score changes additionally require either a second independent source or a documented release event the vendor published itself.
Confidence
High
Multiple official sources cover every scored axis, and the most recent review window includes documented changes from the vendor.
Confidence
Medium
At least one official source per scored axis, but coverage is incomplete or one or more axes rest on a single source.
Confidence
Limited
Public evidence is sparse on one or more axes, or the most recent vendor disclosure predates the current review window. Treat the page as a starting summary rather than a settled posture.
The confidence label on a state's law card reflects how settled the law itself is — not how clean the most recent agent verification run was. We split those signals intentionally: the agent's run-time verification confidence (e.g. “robots.txt blocked the source on this pass”) is a useful internal diagnostic, but it’s misleading on the public surface. A state can have a fully-enacted statute and still produce a low run-confidence if a single source URL was temporarily unreachable; that is not a meaningful signal for a parent reading the card.
Three buckets, derived from three signals
We derive the public confidence from the parent-facing law status, the enforcement state, and whether the entry has a cited source URL.
Confidence
High confidence.
Specific rule in effect + active (or not-yet-effective) enforcement + at least one cited public source. The statute is on the books and we can point at it.
Confidence
Medium confidence.
Limited or adjacent coverage, in motion (a bill is moving), no state-level rule found, OR a specific rule whose enforcement is currently paused by court order. The signal is real but the picture is unsettled.
Confidence
Low confidence.
No data — Rooted Reality has not yet reviewed this state on this topic. The card should be read as “we don’t know,” not as a finding.
We do not show a numeric percentage. We don’t have a continuous confidence model under the hood, and a number like “87%” would imply precision the rubric doesn’t actually carry. The bucket plus the source link plus the analyst notes are the honest surface. If a state’s confidence reading looks wrong against the evidence, the fix is on us — flag it through the contact link in the page footer.
Rooted Reality runs four levels of review depth depending on how much a surface has changed since the last published version. Every level ends in admin approval before anything publishes.
Agents may draft posture updates. Admin approval is required before publication.
Every posture update goes through the admin queue. An admin reads the agent's proposed scores, the behavior-diff or re-score notes, and the cited evidence before the draft is published or returned for rework. A draft that an admin does not publish never appears on a public page.
Major version releases default to a full re-score. Minor releases default to a behavior-diff review and escalate to a full re-score when the diff surfaces age-, privacy-, security-, or moderation-relevant changes. Version pages document which review depth was used and why.
Rooted Reality reviews what a surface publicly says and what we can document from outside its internal systems. We do not assess internal compliance programs, undisclosed safeguards, or behavior beyond publicly documented evidence. Posture can change between reviews, and individual households may experience the same surface differently depending on settings, account type, or version.
Rooted Reality does not sell placement in the Apps Reference Library. Reviewed entities do not approve or edit our posture summaries before publication. If Rooted Reality has a material relationship with an entity covered in the library, that relationship should be disclosed on the relevant page.
Public library pages are designed to be privacy-respecting. Rooted Reality does not use third-party analytics on these public reference surfaces.
Notified when our scoring rubric, evidence rules, or review depth changes. One email per change.