The development director at Springfield Youth Mentoring, a small nonprofit in Springfield, Missouri, has the research problem every under-resourced fundraiser has. The professional databases that would tell her who funds youth mentoring in the Midwest are priced for organizations with a research budget, and she does not have one. What she does have is a general-purpose AI assistant on a free tier and an afternoon. This lesson covers how far that combination actually takes you, where it stops, and what you have to verify before you act on anything it tells you.

Why AI Changes Grant Research

Professional grant research databases carry a subscription price that small nonprofits simply cannot afford. That has always been the quiet inequity of grantseeking: the organizations with the least capacity to hunt for funders are the ones locked out of the tools that do the hunting. General-purpose AI assistants, most of which are free or cheap, can replicate roughly 70% of what those expensive databases do. Specifically, they can help you find funders, understand what those funders care about, and identify peer organizations that have already received money for work like yours.

It is worth being precise about why that works, because the mechanism tells you where the limits are. Foundations publish. They post priorities, they file returns, they appear in the annual reports of the organizations they fund, and journalists write about them. A model trained on public text has absorbed a great deal of that material and can summarize the patterns in it on demand. What it cannot do is check today's version of any of it. So AI cannot replace hands-on funder research entirely, but it dramatically accelerates the process and costs zero dollars. The remaining portion of the work, the current, local, verified detail, is exactly the part you still do yourself.

Four Prompts That Carry the Research

Most people get disappointing results from AI research because they ask disappointing questions. The four templates below are structured so the assistant has enough context to be specific. Each one produces a different artifact, and together they produce a prospect list. Fill in the bracketed fields with your own details before you send them.

Prompt 1: Identify Funder Prospects

"I run [organization name], a [type] nonprofit in [geography]. We serve [population] by [mission]. We need grants of $[amount] for [program]. Using your knowledge of philanthropic funding, what types of foundations might fund our work? Give me 15-20 foundation types or specific foundations (if you know real ones) that typically fund [issue area] in [geography]."

AI will list foundation types, specific foundations if it knows them, and funder characteristics. Treat all three as leads rather than facts. Cross-reference the list with free tools, including Foundation Directory Online and the IRS 990 database, to verify each name and pick up current information. The value of this prompt is not that it hands you a funder; it is that it converts a blank page into a categorized starting set you can work through methodically.

Prompt 2: Understand Funder Priorities

"I found a foundation called [Name]. Based on your knowledge, what are their typical funding priorities? What issue areas do they fund? What geographic areas? What's their typical grant size? What types of organizations do they support?"

AI will give you a funder profile. Verify it on their website, because AI knowledge might be outdated. Used well, this prompt saves you the slowest part of qualification: reading a foundation's whole site to work out whether they would ever consider you. You get a hypothesis in seconds and then spend your reading time confirming or killing it, which is a much better use of an hour than open-ended browsing.

Prompt 3: Find Peer Organizations That Got Funded

"What nonprofit organizations in [geography] work on [issue area] similar to ours? Can you name 10-15 specific organizations?"

Then search each organization's funder list or 990 filing to see who funded them. If they got funded, you should too. This is the highest-yield prompt in the set, because it works backwards from proof. A foundation's stated priorities tell you what they say they fund; a peer organization's donor page tells you what they actually funded, in your geography, for work that resembles yours. Peer lists also surface small local funders that never advertise.

Prompt 4: Strategic Funding Advice

"Our nonprofit has [annual budget]. We want grants to represent [%] of revenue. Based on this, what grant strategy would you recommend? Should we pursue federal grants, foundation grants, or both? What size grants? How many per year?"

AI will give strategic guidance that helps you prioritize. This is the one prompt where you are not asking for facts to verify but for a framework to argue with. If the answer suggests a mix you had not considered, the useful next step is to ask why, and to test the reasoning against what you know about your own capacity to write and manage applications.

The Five-Step Funder Research Workflow

The prompts only pay off inside a sequence that alternates between generation and verification. Step 1 is generating the funder list with AI: use it to identify 20-30 potential funders by type and geography. Step 2 is verifying and detailing with free tools: search Foundation Directory Online, the IRS 990-PF database, and funder websites to verify what you have, and gather actual grant sizes, priorities, and deadlines. The order matters. Generating first is cheap and fast; verifying first is slow and you would be verifying names you have not yet decided are worth the time.

Step 3 is finding peer recipients, using AI together with a search engine. Ask AI to name peer organizations, then search their names alongside "funder list" or "annual report" and see who funded them. Step 4 is developing funder profiles, and it is manual work: create a one-page profile for each of your 8-12 top prospects covering name, grant size, priorities, geography, deadline, contact, and likelihood of fit. Step 5 is prioritizing and applying, also manual: score prospects on fit and apply to the top 8-12 only. The discipline of that last step is what keeps a research exercise from turning into a scattershot application season.

Drafting With AI: Where Help Ends

Research is not the only place AI earns its keep. It is also a competent first-draft engine, provided you stay clear about which parts of a proposal can be drafted by something that has never met your participants.

AI Can Help With

• Drafting narratives: "Write a compelling 2-paragraph problem statement for [issue]" gets you something you then edit.
• Brainstorming outcomes: "What are 5 measurable outcomes for a [program type]?" gets you suggestions you adapt.
• Budget explanations: "Write a 2-sentence justification for budgeting $X for [staff position]" gets you a narrative draft.
• Finding quotes and data: "What statistics support the claim that [problem statement]?" gets you research directions, and you must verify the sources.
• Editing: "Edit this proposal narrative for clarity and impact" gets you suggested edits.

AI Cannot Replicate

• Funder-specific customization, which requires understanding that funder's specific priorities.
• Outcome targets grounded in your organization's own history.
• Compelling impact stories, which require your knowledge of actual beneficiaries.
• Final decision-making, because you are the one who must judge quality.

The pattern in that second list is that every item depends on information the model does not have and cannot acquire. A funder's unwritten preferences live in conversations with program officers. Your realistic outcome targets live in your own data. Your best stories live in your relationships. Anything an assistant writes in those categories will read as plausible and generic at the same time, which is precisely the combination reviewers notice.

Critical AI Limitations

Knowledge cutoff. An assistant's knowledge ends at a specific date, and grant information changes constantly. It might suggest a foundation that has closed or merged, or one whose program has been retired. Always verify current information on funder websites. This limitation is structural rather than a bug to be worked around: the model is describing a world that stopped at some point in the past.

Hallucinations. AI sometimes makes up information. It might name a foundation that does not exist, or describe funding priorities that are incorrect. AI confidence does not equal accuracy, and fabricated answers arrive in exactly the same fluent, assured tone as correct ones. There is no tell in the writing style, which is why the verification step is not optional. Verify everything against primary sources.

Outdated grant data. Even if AI knows a foundation existed in 2024, its current grant size, priorities, or deadlines might have changed. Check the website or call them. A deadline you took from a summarized answer and never confirmed is the single most expensive mistake in this workflow, because you discover it at the moment you are ready to submit.

Misses hyper-local funding. AI works best with national and well-known foundations. It might miss small family foundations, local community foundations, or new funders it was not trained on. That is a serious gap for a small nonprofit, because local money is often the most winnable money. Use free tools for local research, and lean on the peer-organization prompt, which surfaces local funders indirectly through the organizations they have already supported.

Hybrid Research Strategy: AI Plus Free Tools

Put the pieces on a calendar and the whole thing becomes a project rather than a habit you never get to. The schedule below alternates fast AI-assisted generation with slower human verification, so each expensive hour is spent on prospects that have already survived a cheap filter.

The total time investment is 15-20 hours to develop a solid prospect list for 15-20 grants. Compare that with professional research at 30 hours plus a fee. The AI-plus-free-tools route lands at 15 hours with no subscription to pay. The saving that matters most, though, is not the hours or the fee; it is that Week 4 exists at all. Research that stops at a spreadsheet produces applications to strangers. Research that ends in fit-check conversations produces applications to funders who are expecting them.

Prompts That Actually Work

The most effective prompt structure gives the assistant your role, your organization, your population, your geography, your program, its cost, its scale, and then one specific ask: "I'm [role] at [organization], a [type] nonprofit. We serve [population] in [geography]. We want to find grants for [program]. The program costs $[amount] and will serve [number] participants. Using your knowledge of philanthropy, help me identify [specific ask: funder types, peer organizations, funding strategy, etc.]." The more specific you are, the better the response, because every detail you supply narrows the space of plausible answers.

A specific, good prompt reads like this: "I'm the Development Director at Springfield Youth Mentoring, a small nonprofit in Springfield, Missouri. We run a mentoring program that costs annually and serves 75 at-risk high school students. Help me identify 10-15 foundations or funder types that typically fund youth mentoring programs in the Midwest. Include foundation names if you know them, and note which ones have geographic focus on Missouri or Midwest." A vague, bad prompt reads like this: "What grants can we apply for?" The second question has no wrong answer, which is another way of saying it has no useful one.

Ethical Considerations

Using AI in fundraising is not an ethical problem in itself, but it creates several. Most of them come down to whether you are presenting work as yours that you have not actually done, and whether you are passing unverified claims to a funder who will rely on them.

You must:

• Verify all AI suggestions against primary sources before acting on them.
• Never submit AI-generated narratives without significant editing and personalization.
• Disclose if you are using AI, because some funders ask about this.
• Fact-check all statistics, quotes, and claims AI suggests.
• Avoid over-relying on AI; use it as a research assistant, not a substitute for human judgment.

You shouldn't:

• Copy AI-written grant narratives directly into proposals, which lack specificity and authenticity.
• Trust AI's fund information without verification.
• Use AI-generated impact stories; they are generic, and your real stories are better.
• Skip the relationship-building part of fundraising in favor of pure research efficiency.

Anti-Patterns

• Treating the generated list as a prospect list. Names produced in Week 1 have not been verified against anything. Applying from that list directly skips the entire verification half of the workflow and guarantees some effort goes to funders that have closed, merged, or never funded your issue area.
• Submitting a lightly edited draft. A narrative that has been through a spelling pass but still contains no outcome targets from your own history reads exactly like what it is. Reviewers who read dozens of proposals recognize generic writing quickly.
• Trusting a deadline you never confirmed. Of everything an assistant can get wrong, dates are the most costly, because you discover the error at submission time when there is no recovery.
• Asking one broad question and stopping. "What grants can we apply for?" produces a survey of philanthropy in general. The four templates work because each one asks for a single, bounded artifact.
• Letting research efficiency replace relationships. Cutting your research time is only a win if the hours saved go into program officer conversations, not into applying to more strangers.
• Using AI to invent supporting statistics. Asking what data supports a claim is a research direction. Pasting the answer into a proposal without finding and reading the source is fabrication with extra steps.

Practice Prompts

• Run Prompt 1 for your own organization with every bracketed field filled in. Then run it again with half the detail removed and compare the two answers. Note in writing which specific details changed the quality of the output.
• Take the funder list you generate and verify the leading names on it against funder websites and the IRS 990-PF database. Record how many survived verification unchanged, how many needed correction, and how many did not check out at all.
• Use Prompt 3 to name peer organizations in your sector, then look up the funder list or 990 filing for several of them. Write down any funder that appears on more than one list.
• Draft a problem statement with AI, then rewrite it using one real beneficiary story you already know. Put the two versions side by side and mark every sentence that only the second version could contain.
• Build a one-page profile for one top prospect covering name, grant size, priorities, geography, deadline, contact, and likelihood of fit. Time yourself, then estimate what the full set of 8-12 profiles will cost you.
• Write the fit-check email you would send a program officer at your top prospect. Do not use AI for it. Then ask AI to critique it, and decide which of its suggestions you actually accept.

Reflection

Think about the last funding cycle your organization went through. How many of the funders you applied to had you actually spoken with before submitting, and how many were names on a list someone found? If AI research had handed you a larger list that season, would it have improved your hit rate, or would it simply have spread the same writing capacity across more applications? Consider also where your verification discipline is weakest. Most teams are careful about grant sizes and careless about deadlines, or careful about both and careless about whether the foundation still funds the program area at all. Naming your own weak point now is cheaper than discovering it in a rejection letter.

Glossary

• Knowledge cutoff: The date after which an AI assistant has no information. Anything that changed later, including a foundation closing or merging, is invisible to it.
• Hallucination: A confidently stated but fabricated answer, such as a foundation that does not exist or funding priorities that are incorrect. Fluency is not evidence of accuracy.
• 990-PF: The annual return filed by private foundations, searchable through the IRS database and useful for verifying what a funder actually gave and to whom.
• Foundation Directory Online: A funder research database with a free version, used here in the verification step to confirm contact information, grant sizes, and deadlines.
• Peer recipient: An organization working on your issue area in your geography whose published funder list or 990 filing shows you which funders have already backed work like yours.
• Fit check: A conversation with a program officer before you apply, to test whether your program falls inside what that funder is currently looking to support.
• Funder profile: A one-page summary of a prospect covering name, grant size, priorities, geography, deadline, contact, and likelihood of fit.

Related Lessons

• Funder Research for Small Nonprofits: Free and Low-Cost Methods
• AI-Powered Grant Writing: How to Use It Without Sounding Generic
• Grant Strategy for Small Nonprofits: Building a Sustainable Portfolio
• Government Grants: Federal, State, and Local Opportunities
• Building Funder Relationships Beyond the Ask

Closing

The promise of AI-assisted grant research is not that it finds you money. It is that it removes the cost barrier from the part of fundraising that used to require a subscription, and gives that time back to the parts that were always human. A small nonprofit can now produce, in an afternoon, the kind of categorized starting list that previously required a paid database. What has not changed is that every name on it is a hypothesis until you check it, and that funders still give money to organizations they have talked to. Use the assistant for the generation, keep the verification, and spend the hours you save on the conversations.

Key Takeaways

• AI can replicate roughly 70% of what expensive research databases do: finding funders, understanding their priorities, and identifying peers who received funding. The remaining portion is current, local, verified detail that you produce yourself.
• Alternate generation and verification. Generate with AI because it is fast and cheap; verify with funder websites, Foundation Directory Online and the IRS 990-PF database because only primary sources are current.
• Working backwards from peer organizations that already got funded is the highest-yield research move, and it surfaces local funders that national-scale knowledge misses.
• Specificity in the prompt drives quality in the answer. Role, organization type, population, geography, program, cost and scale, then one bounded ask.
• Use AI to draft structure, outcomes language, and edits. Do not use it for funder-specific customization, outcome targets, or impact stories, all of which depend on information it does not have.
• Budget on the order of 15-20 hours to build a solid prospect list for 15-20 grants, and protect the hours set aside for program officer conversations.

Frequently Asked Questions

Is it unethical to use AI for grant writing?

No, as long as you significantly edit and personalize the output. Using AI as a starting point or a brainstorming tool is fine. Copying AI narratives directly into proposals is weak and funders can tell. Use AI for efficiency, not as a shortcut that compromises quality.

Which AI tool is best for grant research?

The mainstream general-purpose assistants all work well, and all have free tiers. Use whichever you are comfortable with. The cost difference is minimal, and the quality of your prompts matters more than the choice of tool.

How do we fact-check AI suggestions about funders?

Always verify on the funder's website or in the IRS 990-PF database. Call the program officer if information seems outdated. Cross-reference with Foundation Directory and peer organization funding lists. AI is a starting point; human verification is essential.

Can AI replace a grants consultant?

No. AI is great for research and drafting. Consultants add expertise in funder relationships, strategic positioning, and complex proposals. Use AI to reduce your research burden and your consultant cost, not to replace consultants entirely for large grants.