philanthropy.org is built by gathering public data from many sources, cleansing it into one normalized substrate, and serving it through an API. Every figure is stamped with its source, so any number traces back to the filing or register it came from. Here is the full provenance — by how authoritative each source is.
Most recent bulk ingest: 2026-09-09 16:53:01 (52 archives processed).
Internal Revenue Service
The universe of registered exempt orgs (~1.97M): EIN, name, NTEE, location, subsection, ruling date, foundation code, BMF assets/income/revenue.
Internal Revenue Service
Baseline rows (EIN, name, city, state, recognition date) for orgs newly recognized by the IRS that haven't yet reached the BMF/Pub78/990-N bulk extracts, so their hub exists ahead of their first 990.
IRS (annual TEOS archives) + GivingTuesday data lake (per-filing mirror)
The complete e-filed return — every element of the core form and all appended schedules (A public support, B contributors, D donor-advised funds, G fundraisers, L insider transactions, R related orgs, plus 990-PF Part XV / Schedule I grants): financials, officers, programs, contractors, governance, narratives.
Internal Revenue Service
The EIN → OBJECT_ID filing locator that maps an org to its e-filed returns, so a filing fetch is an O(1) probe instead of a bulk scan.
Internal Revenue Service
Auto-revocation status and date (exemption lost for non-filing) — a hard compliance signal.
U.S. GSA / Federal Audit Clearinghouse
Single-audit findings for orgs above the federal-expenditure threshold: going-concern doubt, material weaknesses, questioned costs.
State AG / Secretary-of-State charity regulators (pulled directly)
State charitable-registration status and lapses, pulled DIRECTLY from each state's own open-data registry (no aggregator) into state_registrations, with state_lapse events derived for orgs not in good standing. Live: Colorado. Portal-only states (CA, NY) are backlog pending a direct scraper.
Charity Navigator, Candid/GuideStar, BBB Wise Giving Alliance
Third-party ratings, transparency seals, and donor advisories — shown ONLY as badges (their assessment = X). They never feed a core field; if a provider drops us we lose the badge, not data.
Each nonprofit's own homepage
Homepage text for the semantic-similarity layer (embeddings) and program-language enrichment.
YouTube Data API, Brave Search, GDELT
Public reach/engagement and news-presence signals used for organizational-vigor scoring — diagnostic, never authoritative.