Collection approach
Data is sourced from public, primary records: official APIs and bulk downloads from federal, state, and judicial agencies; non-profit data archives; and verifiable journalism. Where a record originates from multiple sources, we prefer the source closest to the underlying government filing.
Source hierarchy
- Primary government records (FEC, congress.gov, court PACER/CourtListener, agency dockets, official disclosures).
- Recognized non-profit aggregators that publish their methods (OpenStates, CourtListener, GovTrack, Public Citizen).
- Established investigative outlets with documented sourcing.
- Derived measures computed by VigilData, clearly labeled as such.
Update frequency
Update cadence varies by dataset. Frequently changing datasets (votes, trades, donations) are refreshed on scheduled cron jobs; slower-moving reference data (member biographies, committee assignments) is refreshed less often. Each dataset surface shows its last-verified date when available.
Scoring overview
Where VigilData computes derived signals (e.g. Accountability Risk Scan, Donor Influence Signals), the score reflects documented relationships, timing, spending patterns, and statistical associations across primary records. Scores are intended as research aids, not legal findings.
Known limitations
- Public records can be late, redacted, or partial.
- State-level coverage is being expanded and is labeled Coming soon where data is incomplete.
- Judicial filings and ethics disclosures are surfaced where electronically available.
Human review
Editorial features, label changes, and methodology updates are reviewed by humans. Algorithmic flags from AI-assisted scans are not published as final claims without verification against primary sources.
AI disclosure
Some research-aid features use large-language-model assistance (the “AI Watchdog” tools). These features are clearly labeled, include confidence indicators where possible, and direct readers to the primary sources they reference.