Review state stays attached
Reviewed, practical, pending-review and restricted are publication states. Research candidates are discovery metadata, not evidence.
Trust center · updated 2026-09-05
What Brali knows, how it knows it, and where the boundaries are.
One canonical surface for evidence handling, citation, provenance, corrections and measurable quality. Brali does not convert these signals into a decorative “AI readiness”, authority or scientific-certainty score.
Review decisions are inspectable
Evidence Decisions record supported claims, unsupported or overstated claims, limitations and editorial outcomes.
Generated artifacts are checkable
The API manifest publishes release metadata and checksums so downstream consumers can pin and verify machine-readable distributions.
Changes do not need a perfect-history story
Material corrections can be recorded in a public ledger. An empty ledger means no entries are currently recorded; it is not a claim that nothing was ever fixed.
Measured signals
Brali exposes the underlying metrics instead of rolling them into one opaque score. Evidence metrics track corpus and review-state coverage. Retrieval benchmark publishes deterministic retrieval cases and recall/precision results. Page & Zone Quality tracks presentation and crawlability separately from evidence trust.
Source and review policy
Source Policy defines source handling. Evidence methodology explains the review layer. These are publisher-declared policies, not independent endorsements.
Citation and reuse
Citation guidance defines canonical attribution. License terms define reuse boundaries; the current public license is non-commercial, so Brali does not describe its content as permission-free open reuse.
History and freshness
Project Updates and the JSON feed expose product and content evolution. Freshness is observable through dated releases and feeds rather than inferred from a generic status badge.
For machines and agents
Site profile describes current interfaces. AI Search & Citation Profile describes discoverability modules, measurement signals and guardrails. Knowledge graph links the publisher, datasets, ontology, evidence, trust and update surfaces.