RivetLore / Evidence standards
Documentary research should make its sources visible, its uncertainty legible, and its review process clear. This page describes what RivetLore supports today—and where the current workflow stops.
How to read a dossier
Each dossier claim is paired with a source title and direct URL. Readers can open the cited source and assess it for themselves. A link records where a dossier points; on its own, it does not certify the source's quality, completeness, or independent confirmation.
A claim displays a confidence percentage. It communicates how confidently the dossier presents that claim; it is not a statistically calibrated probability or an automated fact-check. RivetLore does not claim that percentages are calibrated across topics.
Dossiers record the myth and its origin in separate fields. The explanation should distinguish what story or claim is repeated from what is known about how it began. When an origin is uncertain or disputed, the dossier should make that uncertainty clear rather than force a single answer.
A reviewer must explicitly confirm human QA before a dossier can be published. That is a required review step, not a guarantee that every underlying source has been independently verified or that an error is impossible.
Available today
RivetLore currently supports topic requests and a reviewer-led path from draft to publication. A saved request is not an instant research result.
A creator can save a research topic and optional context. The request and its status stay together in the account.
An authorized reviewer can create a dossier with a timeline, key figures, source-linked claims, and notes about myths and origins.
After a reviewer confirms human QA, the dossier can appear in the public research library. Drafts are not public dossiers.
Planned, not available
AI search and cross-checking are currently unavailable. No automated process searches for sources, gathers them, or verifies claims in the background. Source links and confidence fields belong to the human-reviewed dossier workflow; they are not outputs from an active AI verification service.
Any future research automation should be described as planned until it is shipped and operating. A published dossier signals that a reviewer completed the required human QA step—not that software independently confirmed every claim.
Keep following the question