Grant writing is half research, half assembly, and a small, decisive part persuasion. AI helps most with the first two. It reads a 60-page notice of funding opportunity and turns it into a requirements checklist, drafts a needs statement from your program's real data, and plays the tired reviewer scoring you against the rubric. It gives you back the hours that used to disappear before the actual writing started.
It will also invent the exact things that get proposals rejected: a statistic with no source, a citation that does not exist, an outcome your program never achieved, a partner commitment nobody made. Reviewers who read hundreds of proposals also recognize the flat sameness of unedited AI prose. Every number gets a source you can hand over, and every paragraph gets rewritten until it sounds like your organization.
The funders are paying attention too. NIH and NSF have published notices on AI use in applications and peer review, and a growing number of foundations ask applicants to disclose it or restrict it. Read the current policy for each funder before you paste anything, and treat client, participant, and donor information as confidential by default.
Quick wins this week
- Paste the full funding announcement and ask for a requirements matrix: every required element, page limit, formatting rule, attachment, and deadline, with where each appears. Then check it against the original.
- Give the AI your program's last three reports and ask for the ten strongest evidence points about outcomes, with the exact figure and where it came from.
- Paste your draft narrative and the funder's scoring rubric and ask for a score on each criterion with the sentence that cost you points.
- Draft a two-paragraph letter of inquiry from your one-page program summary and the funder's stated priorities, then cut it until every sentence earns its place.
What AI can do for grant writers, task by task
Prospect research and fit screening
Paste your program summary, geography, budget range, and populations served, and ask what to confirm about a funder before pursuing it. Use a search-connected tool to gather candidates with links, then verify every funder, deadline, and giving priority on the funder's own site or in your grants database. Models invent foundations and misremember deadlines.
Turning the funding announcement into a compliance matrix
Paste the full notice and ask for a table of every requirement: section, page or word limit, formatting, required attachments, eligibility conditions, scoring weights, and deadline, each tagged with where it appears. Check it line by line against the original; a missed attachment is an automatic rejection, and the model will occasionally merge two requirements into one.
Drafting the needs statement and narrative from your evidence
Give the AI your program description, your real outcome data with sources, the funder's priorities in their words, and the section outline, and ask for a draft that uses only the evidence provided and flags any place it needs a fact you have not supplied. Rewrite in your organization's voice. Anything that reads like a general claim about the world gets a real source or gets cut.
Logic models, outcomes, and evaluation plans
Paste your activities and intended outcomes and ask for a logic model with inputs, activities, outputs, short- and long-term outcomes, and indicators you can measure with data you already collect. Push back on any indicator that needs data you do not have. Check that the outcomes match what the narrative promises; a reviewer will.
Budgets and budget justifications
Give the line items, the funder's allowable-cost rules, and your indirect rate, and ask for a justification that ties each line to an activity in the proposal. Recompute every total yourself and against the funder's budget form; models make arithmetic errors and will apply an indirect rate to costs the funder excludes.
Reviewer simulation and revision
Paste the draft and the funder's scoring criteria and ask for a score per criterion, the weakest sentence in each section, and the questions a skeptical reviewer would write in the margin. Then ask for a rewrite of only the weakest paragraph. Keep the final rewriting to yourself so the proposal sounds like a person who runs the program.
Prompts for grant writers
Replace the bracketed placeholders, paste into any chat assistant, and iterate on the result.
Compliance matrix from a funding notice
Act as a senior grants manager. Here is the full funding announcement: [PASTE THE FULL NOTICE OR ATTACH THE PDF]. Build a compliance matrix as a table with these columns: requirement, proposal section it belongs in, page or word limit, formatting rule, required attachment (yes or no, and what), eligibility condition, scoring weight, deadline, and where in the notice it appears. After the table, list anything ambiguous I should ask the program officer about, and any requirement that appears in two places with different wording. Do not summarize; be exhaustive.
Tip: Check the matrix against the notice line by line. A merged or missing requirement is the kind of error that disqualifies a proposal.
Narrative draft from supplied evidence only
You are a grant writer for [ORGANIZATION TYPE] applying to [FUNDER] for [PROGRAM NAME]. Use only the evidence I provide; if you need a fact I have not supplied, write [NEED SOURCE] instead of inventing one. Funder priorities in their words: [PASTE PRIORITIES]. Program description: [PASTE DESCRIPTION]. Outcome data with sources: [PASTE DATA AND SOURCES]. Section to draft: [SECTION NAME], limit [WORD OR PAGE LIMIT]. Plain, specific language with no generic claims about the field. Open with a concrete image from the program, tie every paragraph to a funder priority, and end with what the funding makes possible. Then list every factual claim in the draft with its source.
Tip: Search the draft for [NEED SOURCE] and for any statistic you did not supply. Both must be resolved before anyone else reads it.
Reviewer scoring simulation
Act as an experienced reviewer for [FUNDER OR PROGRAM] who has read hundreds of proposals today. Score this draft against these criteria and weights: [PASTE SCORING CRITERIA]. Draft: [PASTE DRAFT]. For each criterion: the score, the two sentences that cost the most points and why, and the question you would write in the margin. Then give an overall verdict in one paragraph, name the single change that would raise the score most, and flag every claim you would want a source for. Do not rewrite the proposal.
Tip: Run it twice with different personas (a program officer and a community member on the panel) and compare where they disagree.
Budget justification with a math check
Act as a grants finance manager. Here are our budget line items with amounts and calculations: [PASTE LINE ITEMS]. Funder rules on allowable and indirect costs: [PASTE RULES OR SUMMARIZE]. Our negotiated or de minimis indirect rate: [RATE]. Project period: [DATES]. Write a budget justification that ties each line to a specific activity in the proposal, shows the calculation for each amount, and follows the funder's ordering. Then recompute every subtotal and the total independently, list any line that looks unallowable under the rules I gave you, and check whether the indirect rate was applied to any excluded cost.
Tip: Recompute the totals in your own spreadsheet anyway. Arithmetic is where models fail quietly.
Honest progress report
You are a program director writing a progress report to [FUNDER] for [GRANT NAME]. What we promised in the proposal: [PASTE PROMISED OUTCOMES AND TARGETS]. What actually happened, with data: [PASTE RESULTS]. What changed and why: [PASTE CONTEXT]. Address every promised outcome in order, state plainly which were met, partially met, or missed, explain the misses without excuses, and describe what we are changing. Under [WORD LIMIT] words. Do not reframe a missed target as a success. End with two specific things the funder's support made possible, drawn only from the data I gave you.
Tip: Funders renew organizations that tell the truth about misses. Keep the 'do not reframe' line in.
Want a prompt for something else? Use the Prompt Builder.
Skills to build
Sourcing every claim before it goes in
Why: Reviewers check numbers, and a fabricated statistic or citation in a federal proposal can end your organization's relationship with a funder. The model will produce plausible figures on request.
How: Keep a fact sheet per program with every figure and its source. Draft only from that sheet, and search each draft for numbers you did not supply.
Reading funder AI policies and disclosing correctly
Why: Federal agencies and a growing number of foundations have rules on AI use in applications, and some require disclosure. Ignorance does not help on a compliance question.
How: Add 'AI-use policy' as a row in your compliance matrix for every funder. Record the date you checked; policies change between cycles.
Building a compliance matrix by reflex
Why: Most rejections are administrative: a missing attachment, an exceeded page limit, an eligibility miss. A matrix built from the notice on day one prevents most of them.
How: Run the compliance-matrix prompt on every notice, verify it against the original, and use it as the checklist for final submission.
Editing generated prose into your organization's voice
Why: Program officers read stacks of proposals and notice when three from different organizations share the same generic rhythm. Voice is evidence of a real organization with real people.
How: Rewrite the opening and closing of every section yourself. Replace every general claim with a specific detail from the program. Read it aloud.
Protecting client, participant, and donor data
Why: Case notes, participant stories, and donor lists are confidential, and some are covered by HIPAA or grant-specific data agreements. A consumer tool may retain and train on what you paste.
How: Use composite or fully anonymized examples in prompts, and an enterprise tool with no-training terms if your organization has one. Get written consent before any identifiable story is used.
Tools worth knowing
Claude
A careful writing and analysis assistant that shines on long documents.
ChatGPT
The general-purpose AI assistant most of your coworkers already use.
Perplexity
An answer engine that cites its sources, built for research rather than chat.
NotebookLM
A research notebook that only answers from the sources you give it, with citations.
Microsoft Copilot
AI inside Word, Excel, Outlook, and Teams, with your company's data protections.
Gemini
Google's assistant, strongest when your work already lives in Google Workspace.
Cautions for grant writers
NIH and NSF have both issued notices about generative AI in applications and peer review, and many foundations now ask applicants to disclose AI use or restrict it. Some notices address whether AI-generated content counts as the applicant's own work. Read the current policy for every funder before you paste anything from the proposal, and disclose exactly as the funder specifies.
Models fabricate statistics, invent references, and describe partnerships and outcomes that do not exist. A false statement in a federal application can carry legal consequences for your organization, and a fabricated figure in a foundation proposal costs you the relationship. Draft only from your own verified fact sheet and check every claim before submission.
Case files, participant stories, health or legal information, and donor records are confidential, and some are covered by HIPAA, grant data agreements, or consent forms. Never paste identifiable client, participant, or donor data into a consumer AI tool unless your organization has approved that tool for that use. Use composites and get consent.
Unedited AI drafts read alike across applications, and reviewers notice. Some funders also treat substantially AI-developed proposals as not the applicant's own work. The draft is scaffolding; the specific details, voice, and argument have to come from you and your program staff.
Models make arithmetic errors, apply indirect rates to excluded costs, and misread allowable-cost rules. Recompute every total in your own spreadsheet and check allowability against the funder's actual guidance, never the model's summary of it.
Your 30-day plan
- Week 1: Build a fact sheet for one program with every outcome figure and its source. Read the AI-use policy of your top three funders and add a row for it to your compliance template. Use AI for one low-stakes task a day: a thank-you note, a meeting summary, a sentence-level edit.
- Week 2: Run the compliance-matrix prompt on a live funding notice and verify it against the original. Draft one section of a proposal from the fact sheet only and count how many [NEED SOURCE] flags appear.
- Week 3: Put a full draft through the reviewer simulation twice with different personas and rewrite the weakest section yourself. Run the budget prompt and recompute every total in your spreadsheet.
- Week 4: Set up a project or notebook with your master proposal and produce one tailored version for a second funder with a list of every change. Draft a progress report for a current grant and compare it to your last one for honesty.
- End of month: Decide which four tasks you will keep using AI for, write down your verification checklist, and share your prompts and the errors you caught with your development team.
Frequently asked questions
Can you use AI to write grant proposals?
Do funders allow AI-written grant applications?
Will AI replace grant writers?
Is it safe to put client information into ChatGPT for a grant?
How do I keep AI-drafted proposals from sounding generic?
Terms used on this page
Related roles
- AI for Nonprofit ProfessionalsSmall teams, big missions, endless writing. AI takes the first draft of grants, appeals, reports, and memos off your plate so the humans can spend their time with donors and the people you serve.
- AI for Content WritersAI can research, outline, draft, and edit alongside you. The writers who thrive with it use it to reach the interesting part faster, keep their own voice, and are straight with clients about how they work.
- AI for ProfessorsUse AI to take the drafting, summarizing, and course-building hours off your week without putting student records, unpublished research, or a fabricated citation into anything with your name on it.
- AI for ConsultantsAI synthesizes twelve interview transcripts into themes with quotes, builds the issue tree, argues against your recommendation before the client does, and drafts the SOW. The judgment, the relationship, and the confidentiality obligations are yours.