Skip to content
Phase 2: Funder Research and Prospecting

AI Prompt for Form 990-PF Funder Research

By Angel Reyes, DrPH(c), PMP, PMI-ACP, MPH, MCHES · Updated · Free to copy and use
Quick Answer

This AI prompt for Form 990 funder research analyzes the grants list from a private foundation's Form 990-PF, which you paste in. It summarizes grant size range, recipient types, geographic pattern and repeat grantees, shows its arithmetic and reports what the filing says about unsolicited applications. It reads past giving. It does not tell you what the foundation will fund next.

The prompt

Copy and paste
You are helping me research a private foundation using only the text I provide. Do not use memory, web search, assumptions or outside knowledge about this foundation.

Foundation (label only, do not look it up): [FUNDER NAME]
Source document and tax year: [SOURCE AND TAX YEAR]
My program in one sentence: [ONE-SENTENCE PROGRAM DESCRIPTION]

Grants list from the Form 990-PF:
[PASTE 990-PF GRANTS LIST]

Other relevant text from the filing, including anything about how to apply:
[PASTE OTHER 990-PF TEXT]

Return:
1. The number of grants listed and their total dollar amount. Show your arithmetic.
2. The grant size range: the smallest and largest grants and the median. Show each step. List any entries you excluded or could not read, and say so if the list appears incomplete.
3. The types of recipients and the issue areas visible in the list.
4. The geographic pattern of recipients. Report recipient locations as listed. Do not treat them as a stated geographic policy.
5. Repeat grantees within this list, with the number of appearances and amounts.
6. Whether the pasted text says the foundation accepts unsolicited applications, says it gives only to preselected organizations, or does not say.
7. A fit note for my program labeled Strong, Possible, Limited or Insufficient Evidence, with reasons tied to specific entries in the list.
8. A checklist of facts I should confirm on the foundation's own website or with the foundation before making contact.

For each finding, point to the supporting entry or quote a short phrase. Label uncertainty clearly. Do not invent priorities, application procedures, deadlines or contact details.

Works in ChatGPT, Claude, Gemini and other AI chat tools. Replace each bracketed placeholder with your own details before you send it.

What to fill in

[FUNDER NAME]
Used only to label the output. The prompt tells the AI not to look the foundation up or add what it remembers.
[SOURCE AND TAX YEAR]
Where you got the filing and which tax year it covers. Filings become public well after the year they describe, so note this on everything you save.
[PASTE 990-PF GRANTS LIST]
The list of grants paid during the year, copied from the filing with recipient names, locations, purposes and amounts intact. Paste the whole list if you can. If you paste part of it, say so.
[PASTE OTHER 990-PF TEXT]
The section of the filing where the foundation describes how to apply, or states that it gives only to preselected organizations, plus any stated restrictions.
[ONE-SENTENCE PROGRAM DESCRIPTION]
Your population, activity and intended result in one sentence, with nothing confidential.

Why this prompt works

Form 990-PF is the annual return that private foundations file with the IRS, and it is a public document. It lists the grants the foundation paid during the year, which makes it one of the few places you can see what a foundation actually funded as opposed to what it says it funds. How much detail each entry carries varies from one filing to the next, and some describe a grant's purpose in only a word or two.

Reading a long grants list by hand is slow. The prompt has the AI do the tallying while keeping the result checkable.

  • A closed evidence boundary. The AI analyzes the pasted filing and nothing else, so its summary is about this foundation's listed grants and not what it recalls about foundations in general.
  • Visible arithmetic. AI tools make calculation errors, especially on long lists. Asking for counts, totals and each step exposes skipped rows and duplicates so you can catch them.
  • Median, not just average. One very large grant can make the average misleading. The range and median describe what a typical grant looks like.
  • Observation kept apart from interpretation. A recipient's address is a fact. A geographic funding policy is an inference. Item 4 tells the AI to report the first without claiming the second.
  • Three possible answers on applications. "Does not say" is allowed, so silence in the filing is not turned into a yes or a no.

Why it is appropriate to use

This is public-record research. The filing is published, the output is an internal note, and you are using AI for the counting and sorting it can do quickly. That makes it a sound use of AI during prospecting.

Keep the limits in view:

  • A 990-PF describes a past year. It does not show current priorities, current capacity or openness to new applicants.
  • Recipient locations show where grantees are based, which may differ from where the funded work happened.
  • Only private foundations file Form 990-PF. Community foundations and other public charities file Form 990, which is a different return.
  • If the filing says the foundation gives only to preselected organizations, a history of grants like yours is not an invitation to apply.

Always check the foundation's current website and guidelines after running this prompt. The guide to family foundation grants covers what to do with what you find.

How to check and revise the output

  1. Recount. Compare the number of grants and the total the AI reports with the filing. If the filing prints its own total for grants paid, the two should agree.
  2. Spot-check entries. Pick several grants, including the largest and smallest, and confirm the names and amounts against the original.
  3. Recalculate the median yourself in a spreadsheet if the number will influence how much you ask for.
  4. Check name matching. Make sure "repeat grantees" are really the same organization and not two with similar names.
  5. Read the application language yourself. Do not rely on the AI's reading of whether unsolicited requests are accepted.
  6. Look at more than one year before drawing conclusions about patterns, and record the tax year beside every figure you keep.

Frequently asked questions

What information should I give an AI prompt for Form 990-PF funder research?
Paste the grants list from the foundation's publicly filed Form 990-PF, any text in the filing about application procedures, the tax year and a one-sentence description of your program. Tell the AI to use only that text and to show its arithmetic.
What can an AI prompt reveal from a foundation's Form 990-PF?
It can summarize patterns visible in the filing: grant size range, recipient types, where recipients are located, repeat grantees and what the filing says about unsolicited applications. Verify the results against the original filing and the foundation's current guidelines before you act on them.
Is a Form 990-PF the same as a Form 990?
No. Form 990-PF is the annual return filed by private foundations. Most other tax-exempt organizations, including public charities, file Form 990 or one of its shorter versions. This prompt is written for the grants list in a 990-PF.
Can I ask AI to look up a foundation's 990 for me?
Do not rely on an AI tool's memory for filing data. Get the filing from an official or established public source, such as the IRS Tax Exempt Organization Search, and paste in the text so every figure can be traced to the document.

More prompts for funder research

Related prompts and guides

← Back to all AI prompts for grant writing

Ready to Master Grant Writing?

The Complete Grant Architect is a free 16-week course that transforms you from grant writer to strategic grant professional. Learn proposal engineering, federal compliance, budgeting, evaluation design, and AI-powered workflows.

Enroll Free in The Complete Grant Architect