---
license: other
language:
- en
tags:
- nonprofits
- ngos
- charity
- finance
- vectors
- embeddings
- rag
pretty_name: GiveAI — World's Largest Vetted Nonprofit Database
size_categories:
- 10M<n<100M
task_categories:
- feature-extraction
- text-retrieval
- question-answering
---

# GiveAI — World's Largest Vetted Nonprofit Database

The world's largest vetted nonprofit dataset — **8 million+ organizations across 80+ countries**, combining structured financial data with vector embeddings so AI applications can search, verify, and reason about any nonprofit in the world.

---

## What's in this dataset

This dataset has two components that work together:

### `all_ngos/` — Master Nonprofit Records (~8M records)

One document per nonprofit, containing the full organizational profile:

- **Identity** — legal name, EIN/registration number, country, formation year
- **Location** — full address, formatted address, country code
- **Contact** — phone, email, website
- **Mission** — stated mission, beneficiaries served, program details, geographic reach
- **Financials** — up to 3 years of revenue, expenses, assets, liabilities, donations received/made, salaries, government grants, investment income
- **Aggregated stats** — max/min/avg across years for all financial fields
- **Tax & compliance** — tax exempt status, report type (990/990PF/990EZ), political affiliation, donation permissions (domestic and foreign)
- **References** — source URLs from official registries
- **Vector embedding** — 256-dimensional semantic embedding of the organization

### `vectors_optimized/` — Attribute-Level Embeddings (~14M records)

Where `all_ngos` has one embedding per organization, this dataset has a separate embedding for each **individual attribute** — mission, revenue, volunteer count, program details, and more. This enables fine-grained semantic search:

- Find nonprofits whose *mission* matches a query
- Find nonprofits with a specific *financial profile*
- Find nonprofits by *cause area* or *geography* independently

Each record links back to `all_ngos` via `ngo_id` and includes the organization's full financial summary alongside the targeted embedding.

---

## Coverage

Data sourced from IRS 990 filings, national charity registers, and proprietary data pipelines across 80+ countries. All organizations are verified against official government records.

| Region | Organizations |
|---|---|
| United States | ~1.6 million |
| Canada | ~86K |
| United Kingdom | ~171K |
| Rest of world | ~6.2 million |
| **Total** | **8M+** |

---

## Embeddings

All `vector` fields are generated using **OpenAI `text-embedding-3-large` with `dimensions=256`**.

To perform semantic search against this dataset, embed your query with the same model:

```python
from openai import OpenAI
client = OpenAI()

response = client.embeddings.create(
    model="text-embedding-3-large",
    input="your search query here",
    dimensions=256
)
query_vector = response.data[0].embedding  # 256-dim list of floats
```

Then run nearest-neighbour search against the `vector` column using your vector DB of choice (Qdrant, Weaviate, pgvector, FAISS, etc.).

---

## Why structured + vectors?

Most nonprofit databases are either purely structured (good for filters, bad for semantic search) or purely text (good for search, bad for analytics). GiveAI combines both:

- **Structured fields** → filter by country, revenue range, tax status, cause area
- **Vector embeddings** → semantic search in natural language
- **Together** → hybrid queries like *"climate nonprofits in Southeast Asia with assets over $1M"*

This makes it natively compatible with RAG pipelines and conversational AI — an AI assistant can look up, verify, and reason about any nonprofit in the world using this dataset as its knowledge base.