46% of grant professionals surveyed use AI to draft proposals, and 71% say AI lets them write and submit a proposal in less than a week — versus a two-week average without it (survey of 300+ grant professionals)Source ↗
Nearly 25% of nonprofits use AI for grant writing assistance, yet 76% still lack a formal AI strategy, per TechSoup's 2025 AI Benchmark ReportSource ↗
Only 10% of 527 U.S. foundations surveyed by Candid said they would accept grant applications with AI-generated content; 67% were undecided and 23% said noSource ↗
NIH will not review applications 'substantially developed by AI' effective with the September 25, 2025 receipt date, while allowing AI as a limited assistive toolSource ↗
writingClaudeChatGPT

Drafting a needs statement from your own community data

The statement of need is where most proposals stall — you have the census figures, intake numbers, and community survey results, but turning them into a tight narrative takes a full afternoon per proposal. Drafting is the single most common AI use among grant professionals (46% in Instrumentl's survey), and it works precisely because you supply the evidence and the AI supplies the first arrangement of it.

Prompt
You are a grant writer's drafting assistant for {{organization_name}}, a nonprofit. Draft the statement of need for a grant proposal.

Program context: {{program_summary}}

Community data and evidence (the ONLY facts you may use): {{community_data}}

Funder priorities to align with: {{funder_priorities}}

Output format:
- A statement of need of no more than 400 words, in plain, specific language.
- Open with the problem in concrete local terms, not a global statistic.
- Weave in the data points I provided, keeping every number exactly as written.
- Close with why this organization is positioned to address the need.

Rules:
- Use only the facts, numbers, and sources I provided. Do not add statistics, research findings, or citations from your own knowledge — where stronger evidence would help, insert [DATA NEEDED: describe what to find] instead.
- No filler phrases like "in today's rapidly evolving landscape" and no exaggerated urgency.
- Write in our voice: direct, factual, community-focused.
- After the draft, list any claims that still need a citation before submission.

Fill in your details and the prompt updates live — then copy.

planningClaudeChatGPTGemini

Turning a funding announcement into a compliance checklist

Federal NOFOs and state RFPs routinely run 60-100 pages, and a single missed attachment or formatting rule can disqualify an otherwise fundable application. Grant writers now paste announcement text into an AI tool to extract every requirement into a checklist and map the narrative outline to the scoring criteria — then verify it against the document themselves.

Prompt
You are a grants compliance assistant. I am pasting the text of a funding opportunity announcement below. Build two deliverables for a {{org_type}} planning to apply.

1. Compliance checklist — a table with these columns: Requirement, Where stated (section reference), Status (leave blank for me). Cover eligibility criteria, all deadlines (letter of intent, application, reporting), required forms and attachments, formatting rules (page limits, fonts, file types), cost-share or match requirements, and submission method.

2. Response outline — map each required narrative section to the stated evaluation criteria and point values, with a suggested word budget per section given the stated limits.

Rules:
- Use only what is in the pasted announcement. Quote the announcement's own wording for each requirement and give its section reference.
- If the announcement does not address an item on the checklist, write "Not specified in announcement — confirm with program officer" rather than guessing.
- If a requirement appears in two places with different wording, mark it [CONFLICT] and quote both versions.

Announcement text: {{announcement_text}}

Our project in one sentence: {{project_summary}}

Fill in your details and the prompt updates live — then copy.

communicationClaudeChatGPT

Writing a letter of inquiry that mirrors the funder's priorities

Most foundations gate their process with a one-page letter of inquiry, and program officers can spot a mail-merge LOI instantly. Tailoring each letter to a funder's stated priorities is exactly the kind of restructuring AI does well — as long as the outcomes, history, and fit come from you, because Candid's funder survey found some foundations decline AI-shaped applications precisely because their grantmaking is relationship-based.

Prompt
You are helping a grant writer draft a letter of inquiry to {{funder_name}}.

What the funder says it funds (pasted from their guidelines or website): {{funder_priorities}}

Our organization and program (the only facts you may use): {{org_and_program_facts}}

Requested amount and use: {{request_details}}

Draft a one-page LOI of 350-450 words with: an opening sentence that connects our work to the funder's stated priority (quote their own language once); a short paragraph on the need; what the program does and its track record, using only the outcomes I supplied; the specific request; and a closing that invites a conversation rather than assuming an award.

Rules:
- Do not invent outcomes, partnerships, budget figures, or organizational history. If a standard LOI element is missing from my facts, insert [NEED: element] instead of filling the gap.
- Keep every number exactly as I provided it, with its time period.
- Match the funder's vocabulary where it is honest, but never claim alignment we don't have — if our program doesn't fit one of their priorities, omit it rather than stretch.
- Plain, warm, professional tone. No superlatives.

Fill in your details and the prompt updates live — then copy.

writingClaudeChatGPTGemini

Cutting a narrative to strict word and character limits

Online portals impose hard limits — 250-word boxes, 2,000-character fields — and compressing a narrative without losing a required element is slow, fiddly work. Editing is where experienced teams lean on AI most: Instrumentl found 59% of large organizations managing 50+ grants prioritize AI for proofreading and editing, and Candid specifically recommends it for editing for length.

Prompt
You are an editor helping a grant writer fit a proposal narrative into a strict limit. The application portal allows {{limit}} for this field.

Original text: {{original_text}}

Elements that must survive the cut, in priority order: {{must_keep}}

Produce:
1. Version A — cut to fit the limit using deletion and tightening only, with minimal rewording.
2. Version B — cut to fit with fuller rewriting for flow.
3. The exact word count and character count (with spaces) of each version.
4. A list of everything removed, so I can confirm nothing essential was lost.

Rules:
- Keep every statistic, dollar amount, date, and proper noun exactly as written. Do not round, update, or "improve" any number.
- Add no new facts, claims, or transitions that introduce meaning absent from the original.
- Preserve first person plural ("we") and the exact names of programs and partners.
- If the text cannot fit the limit without dropping one of my must-keep elements, stop and tell me which elements conflict instead of silently dropping one.

Fill in your details and the prompt updates live — then copy.

analysisClaudeChatGPT

Turning program data into a funder progress report

Post-award reporting piles up in clusters — three funders, three templates, one program spreadsheet. Grant professionals increasingly use AI to assemble outcome data and staff notes into report narratives, an application the Grant Professionals Association has covered for both writing and reporting. The discipline is feeding it only real figures and real explanations, because a grant report is a formal account of how restricted money was used.

Prompt
You are helping a grant writer draft an interim progress report for a funded program.

Grant objectives as written in the funded proposal: {{grant_objectives}}

Actual results this period (the only numbers you may use): {{program_data}}

Staff notes on context and variances: {{variance_notes}}

Reporting period: {{reporting_period}}

Draft the narrative with these sections:
- Progress toward each objective — one short paragraph per objective, pairing the target with the actual figure.
- Variances — explain gaps using ONLY my staff notes. If I provided no explanation for a missed target, write [EXPLANATION NEEDED] rather than inventing one.
- Looking ahead — planned activities drawn from my notes only.

Rules:
- Where we missed a target, state it plainly and factually. Do not spin, minimize, or reframe a shortfall — funders read hundreds of reports and trust plain accounting.
- Keep every figure exactly as provided, showing target versus actual side by side.
- Claim no impact beyond what the data supports.
- 600 words maximum, ready to paste into a funder portal.

Fill in your details and the prompt updates live — then copy.

automationClaudeChatGPT

Building a reusable answer library from past proposals

Every portal asks the same questions in different shapes — organizational capacity, DEI statement, evaluation plan — and most grant writers keep a messy folder of past answers with facts that have drifted out of date. Candid identifies this assembly work — consistency, eliminating repetition, refining word choice — as one of the strongest uses of generative AI, and it compounds: build the library once, adapt per application.

Prompt
You are helping a grant writer build a reusable answer library. Below are {{number_of_versions}} answers we have written to essentially the same application question for different funders.

The question they all answer: {{common_question}}

Past answers: {{past_answers}}

Produce a library entry:
1. Master answer — the strongest single version, merging the best material from all of them.
2. Three cut lengths: 100 words, 250 words, and 500 words.
3. Fact inventory — every statistic, date, name, and dollar figure used, listed with which source answer it came from, so each can be verified against current records before reuse.
4. Conflict flags — wherever the past answers disagree (different founding years, staff counts, outcome figures), mark [CONFLICT] and show both versions side by side. Do not pick a winner; we will resolve it.

Rules:
- Use only material from the pasted answers. Add no new facts, statistics, or claims.
- When merging, prefer specific, concrete sentences over general mission language.
- Flag any figure tied to a year, since it will need updating before reuse.

Fill in your details and the prompt updates live — then copy.

Common questions from grant writers

Will funders reject my proposal if I use AI to help write it?

It depends on the funder, so check every time. NIH will not review applications substantially developed by AI (effective September 25, 2025), NSF encourages disclosure of generative AI use in the project description, and among foundations Candid surveyed, only 10% explicitly accept AI-generated content while 67% are undecided. AI-assisted drafting of your own ideas is broadly tolerated; AI-generated substance is not.

Is it safe to paste client or beneficiary information into ChatGPT or Claude?

Not into consumer accounts. Free tools may retain and train on what you paste, and NSF warns that information disclosed to public AI tools cannot be protected from further sharing. Aggregate data before drafting (counts and percentages, never names or case details), and use an enterprise or zero-retention tool if your organization handles sensitive populations.

Can funders tell when a proposal was written by AI?

Not reliably — in Candid's survey, 57% of foundations said they don't know whether they've received AI-generated proposals. The bigger risk isn't detection but blandness, because reviewers reading hundreds of applications notice generic, voiceless prose. Use AI for structure and speed, then rewrite until it sounds like your organization.

Will AI replace grant writers?

The evidence points to compression, not replacement — 71% of grant professionals using AI can submit in under a week, but the parts that win grants (program design, funder relationships, honest data, and judgment about fit) remain human work. Funders who reject AI-heavy applications say their giving is relationship-based, which is exactly the part AI can't do.

Related professions