Our Commitment to Transparency
Trust is earned with evidence, not claims. This page shows live quality metrics from our platform — computed from real data, updated hourly, with no manual curation.
Quality Scorecard
Aggregate quality metrics across all disease reports. These numbers are computed directly from our database — not curated or cherry-picked.
93%
Avg across 7,872 scored reports
100%
Reports passing all validation tiers
64%
Reports verified against source data
50%
7,920 of 15,964 diseases
Last computed: Feb 8, 2026, 7:38 PM. Metrics refresh hourly.
Evidence Base Distribution
Every report displays its evidence tier so you always know how well-supported the information is.
Backed by extensive structured data from multiple authoritative sources
Moderate data available, actively enriching from new sources
Limited structured data, honestly labeled — not padded with speculation
Evidence tiers are computed automatically from structured data coverage — not from subjective judgment. Learn how tiers are determined
Content Coverage
How much of the rare disease catalog has generated, QC-validated content.
50% of rare diseases have AI-generated, QC-validated reports
6,265
Comprehensive 7-section content from rich data
762
Actionable pathways for moderate-data diseases
893
Honest starting points for data-sparse diseases
81 diseases with PubMed citations
Peer-reviewed literature linked via hybrid AI + PubMed API pipeline — zero hallucinated PMIDs
Data Source Health
Live sync status across all authoritative data sources.
Disease taxonomy and hierarchy
Orphan drug designations and approvals
Approved drugs and gene therapies
Active and completed clinical trials
Validated gene nomenclature (44,748 symbols)
Clinical phenotypes with frequency data
Prevalence and incidence data
Peer-reviewed citation enrichment
25+ sources across 3 tiers
State and federal legislation tracking
Freshness timestamps reflect the most recent data sync for each source. For programmatic access, see the API health endpoint.
Data Provenance & Audit Trail
Every report carries cryptographic metadata for reproducibility and regulatory traceability.
Reproducibility
Every report stores SHA-256 hashes of the prompt, constraint map, and knowledge packet used to generate it. Given the same inputs, the same validation applies.
Traceability
Every generation records what triggered it — auto-fill for missing content, news-triggered update, scheduled refresh, or admin-initiated regeneration.
Version History
Full edit history with diff comparison, rollback capability, and attribution. Every change — whether by AI, community, or expert — is preserved.
Immutable Audit Log
Who changed what, when, and why — supporting regulatory traceability. Validation results, fact-check outcomes, and expert reviews are all timestamped.
Honest Limitations
Trust requires honesty about what we can and cannot guarantee.
Kisho is not a substitute for professional medical advice. Always consult a healthcare provider for clinical decisions.
AI-generated content can contain errors despite multi-tier QC. Our validation catches the vast majority, but no system is perfect.
Data-sparse diseases are honestly labeled as "Early" evidence base — not padded with speculation to appear comprehensive.
Coverage varies by disease. Some have extensive structured data from 10+ sources; others have limited information available globally.
We publish our evidence tier on every report so you always know how well-supported the information is.
Regulatory Awareness
How Kisho aligns with industry standards for data integrity and AI transparency.
ALCOA+ Principles
Our audit trail aligns with ALCOA+ — the pharma industry standard for data integrity: Attributable, Legible, Contemporaneous, Original, Accurate, plus Complete, Consistent, Enduring, and Available.
EU AI Act Awareness
We are preparing for transparency obligations under the EU AI Act (effective August 2026), including technical documentation, risk management, and human oversight requirements.
Data Privacy
Kisho processes no patient PII. All disease intelligence is derived from public authoritative sources (MONDO, FDA, PubMed, HPO). Privacy-by-design principles are followed throughout.
How We Calculated These Numbers
Quality Score is a composite index (0-100%) computed per report from validation cleanliness, fact-check outcome, quality review results, and regeneration history. The page displays the average across all scored reports.
QC Pass Rate measures the percentage of reports in “Ready” status — meaning they passed all automated validation tiers and are actively serving content.
Fact-Check Pass Rate is the percentage of reports where automated fact-checking against source data found no unsupported claims.
Evidence Base tiers (Strong / Developing / Early) are computed from structured data coverage and determinism scores — not from subjective judgment.
Freshness timestamps reflect the most recent successful sync for each data source. All metrics are cached and refresh hourly.
For programmatic access to health and freshness data, see the /api/v1/health endpoint (no authentication required).
Trust Is Built in the Open
Transparency is not a page — it's a practice. Every report, every validation, every data sync is part of the record.