CharitySense Data

Public 990 data should be free to search

The IRS publishes the returns. Making them searchable is an interpretation problem as much as an access problem. The form mix in our current archive shows why a useful public answer needs its limits attached.

By the CharitySense Data team

Updated

1.54M
e-Postcard notices among the published filing records
8.87M
published filing records across five forms
2.22M
public organization profiles
17.4%
of filing records are e-Postcards
97.2%
of named transfer records come from two grant sections

01/files are public

The files are public. The questions take work.

The IRS posts machine-readable Form 990-series returns for download. Organizations generally must also make their annual returns available for public inspection. ProPublica's Nonprofit Explorer provides a free search archive and public API. Free access already has serious stewards.

Yet a question such as “How did this organization's grants change?” asks us to join returns across years, understand which form was filed and show the line that supports the answer. The public record deserves that kind of search without a subscription. It also deserves clear warnings when the filing cannot answer the question.

A free answer is most useful when the reader can see what it counts.

02/One archive five different

One archive, five different kinds of return

Our current public statistics count 8,868,546 filing records across 2,222,250 organization profiles. Those are different units. A profile can have returns from several years; an unrelated-business tax return may accompany an annual information return. Counting documents as though they were distinct nonprofits would inflate the sector.

The form mix carries an even bigger caution. 1,540,498 records are Form 990-N e-Postcards, 17.4% of all published filings. The IRS asks for eight basic information items on that notice. It does not provide a full expense statement or a list of programs. A blank program-spending comparison for one of these organizations means the form did not ask the question.

The filing archive is five different forms

Filing records, millions. Select a form to see the question it can answer.

Source: CharitySense published-site statistics, computed September 26, 2026. Units are filing records, not organizations. See Behind the Data.

Form 990-PF belongs to private foundations and uses a different financial structure. Form 990-T concerns unrelated business income tax and can coexist with a primary return. Even among full Form 990 filers, attached schedules vary. Our essay on Schedules B and R shows how much interpretation can turn on what a schedule discloses.

03/relationship map can be

A relationship map can be large and still be partial

Joining filings makes new questions possible. The current named-money-flow map contains 1,588,539 transfer records with organizational endpoints. Of these, 97.2% come from private-foundation grant lists or the domestic-grant schedule on full returns. That percentage describes records in the map, not the share of dollars, organizations or all nonprofit spending.

The concentration tells us where public relationships are easiest to resolve. Some flows have no named counterparty or no reliable identifier. Foreign-grant detail, for example, generally reports a region and purpose rather than a public recipient name. An open graph should let a reader trace the edges it has while making the missing edges legible. Our filing puzzle explores the scale of the financial record behind those links.

Missing evidence is not a negative finding
A missing connection in a relationship map does not prove that no transfer happened. It may mean the form did not ask for a recipient name, the organization filed a different form or the available evidence could not resolve an identity.

04/New sources help only

New sources help only when they keep their labels

The filings now sit beside other kinds of public evidence. We currently have 26,902 organizations with federal-award records and about 33,500 selected website captures. A federal award record is issued by a government source; a website statement is the organization's own description at a point in time. Neither is an IRS filing, and the two should not quietly overwrite one another.

The website set is selected and much smaller than the profile archive. An organization without a capture should never be described as having no website or no program. An organization with an award record should be read in the award's time frame and context. This is the practical case for source labels, dates and direct citations in a public search result. Our grantee diligence essay shows why that distinction matters when a decision has consequences.

05/What a useful public

What a useful public answer can promise

We think free search should identify the organization, distinguish its filing years and form types, and point back to the exact public evidence. Comparisons should name their denominator. Where a value comes from a website or another government source, the reader should see that source and its date.

It should also leave room for the absent parts of the sector. Churches and some church-affiliated organizations generally do not have to file annual returns or notices. Recent tax years arrive over time. Neither a blank page nor a clean chart can fill those gaps by inference. Our faith-and-geography analysis treats filing visibility as part of the finding.

06/Behind the Data

Behind the Data

We read the current published-site statistics on September 28, 2026. They were computed September 26 and report 2,222,250 public profiles and 8,868,546 published filing records. We checked that the five form counts sum to the filing total; each share in the chart divides its form count by 8,868,546. The e-Postcard share is 17.4% after rounding. These are filing records, not unique organizations or a count of every nonprofit in the country.

We separately counted currently available named-transfer records, federal-award records and website captures. The website count was 33,506 when read on September 28 at 23:20 UTC and is rounded to about 33,500 in the article; this independently refreshed set can move between reads. The 97.2% transfer share uses all 1,588,539 named-transfer records as its denominator; 845,475 came from private-foundation grants and 698,924 from full-return domestic-grant lists. Those record counts cannot be converted into a spending share. External-evidence counts describe current coverage, not the number of organizations with or without a website or federal funding. The newest filing years can also change.

Sources

Ask the research team

What would you want us to pull from the filings? Tell us, and a future piece may answer it.

Final Thoughts

The return files are already public. Search can make them easier to question, but the form mix keeps us honest. Nearly one in six records is a brief e-Postcard. New relationship and outside-source evidence reveals useful connections while leaving visible gaps. We would rather show those gaps than fill them with an attractive guess.

A good public answer lets someone ask a better second question and inspect the evidence for themselves.

About the author

The CharitySense Data team

Research team

We read public nonprofit filings and other sources to understand how the sector works. This blog shares what we find, along with the limits of the evidence. Each article explains how its figures were measured and cites the records or primary sources behind them.