Truth through
transparency.
Witness does not ask users to trust one opaque score. Reports move through visible confidence states using corroboration, flags, source history, and review context.
Witness does not ask users to trust one opaque score. Reports move through visible confidence states using corroboration, flags, source history, and review context.
Default state for all new reports. The report has been submitted but not yet corroborated by other users.
At least 10 independent users have confirmed this report from the same location. High confidence.
An admin or institutional partner has independently verified this report. Highest tier of trust.
Multiple users have flagged this report. Under community review. Context is attached visibly.
A user submits a geotagged report with text, photo, and category. It appears on the map immediately as "unverified".
Other users in the area can corroborate the report. Each corroboration increases the report's trust signal.
At 10+ corroborations from distinct users, the report automatically moves to "corroborated" tier. The original reporter earns +5 trust score.
If 5+ distinct users flag a report, it's automatically marked "disputed". Context from both sides is visible. Admins review the queue.
Admins can verify reports independently, dismiss flags, or take action. All moderation is transparent, never silent deletion.
A real-time public record will attract manipulation attempts. Witness treats hostile behavior as a first-order design constraint.
Distinct corroborators, source history, location/time consistency, and flag velocity are exposed before a report gains confidence.
Public coordinates are softened, identity modes separate display identity from internal accountability, and sensitive media requires review paths.
Moderation decisions should be logged with visible reasons, appeals, and preserved context instead of silent deletion.
Feeds are chronological and map-native. Disputed context stays attached to the event instead of disappearing into replies.
Content is never quietly removed. Disputed content is flagged with visible context. Users see why something is disputed.
Even when content is removed, the record exists. No hard deletes. Transparency in every action.
The community flags and corroborates. Admins review. Algorithms never decide what you see. Chronological feeds only.
GPS coordinates are never shown publicly. Only reverse-geocoded location names. Identity modes protect contributors.
Ever. Under any circumstances. This is a non-negotiable principle of the platform.
No truly anonymous posting. Identity modes provide privacy while maintaining internal accountability.
Common questions about how Witness verifies reports, builds trust tiers, and resists misinformation and abuse.
Witness verifies reports through corroboration from other people on the ground, public flagging, and transparent trust tiers. Reports begin unverified and move through visible states as evidence accumulates, so reliability is shown rather than asserted.
Trust tiers are visible levels that reflect how much corroboration and credibility a report or contributor has earned. They let readers and newsrooms judge reliability at a glance while keeping the underlying evidence transparent.
Witness combines community flagging, corroboration, and transparent moderation to surface accurate reports and demote false ones. Moderation standards are published openly, and decisions can be appealed, rather than relying on silent removal.
No. Witness deliberately avoids engagement-optimizing ranking. Content is organized around location, time, and evidence, so what surfaces is driven by corroboration and relevance rather than what generates the most clicks.
The community can publicly flag reports, and those signals feed into transparent trust tiers and moderation review. The goal is resistance to coordinated abuse while keeping the process visible and accountable to contributors.
If you work in journalism, safety, research, or moderation, request access and help stress-test the verification model before broader release.