Grant writing is mostly writing under pressure with a deadline, a fixed word count, and a reviewer who has a hundred other proposals to score. It is exactly the kind of work where a capable language model helps — but only if you prompt it like a grant professional, not like someone asking for "a grant proposal about youth programs." Generic prompts produce generic proposals, and generic proposals get declined.

This guide collects AI prompts for every stage of grant writing: prospect research, needs assessment, problem framing, objectives, methods, logic models, budgets, evaluation plans, sustainability, and the final reviewer-style critique you should run on every draft before submission. Each prompt is structured so you can paste in your own organization's context, the funder's requirements, and your existing materials, and get a usable first draft you then refine.

Use these with ChatGPT, Claude, Gemini, or any capable model. They pair well with Promptwright, where you can keep your full prompt library versioned and ready to run across your team.

Where AI actually helps in grant writing

Models are not going to write a winning proposal for you. What they do well, with the right prompts:

Drafting first versions of sections you have written a hundred times — needs statements, organizational background, methods — so you start from a structured draft instead of a blank page.
Reframing weak, vague language into the specific, measurable, evidence-backed prose that reviewers reward.
Aligning your narrative to a specific funder's stated priorities, scoring rubric, and word limits.
Critiquing drafts as a hostile reviewer would, before a real reviewer gets to them.
Translating programmatic jargon into plain language that a generalist reviewer can score without a glossary.
Generating evaluation frameworks and logic models from your stated objectives, in seconds instead of hours.

What AI cannot do: verify your facts, decide if a program is fundable, ensure you actually have the staff capacity you're claiming, or guarantee alignment to a funder's unstated priorities. You still own the truth claims and the strategy. Use AI to draft and pressure-test; verify everything before submission.

An ethics note before you start

Grant proposals make real claims about real programs and real budgets. Three rules that should be obvious but bear repeating:

Never invent statistics, citations, or outcomes. Models fabricate numbers and study references confidently. Every figure, every quote, every research citation must be verified in your source materials before it goes into a proposal. If a model produces a statistic, treat it as a hypothesis to check, not a fact.
Don't fabricate capacity or partnerships. If you don't have a partnership with the local school district, do not let AI draft language implying you do.
Don't submit AI-drafted text unreviewed. A reviewer can often tell when prose has the smooth, generic, slightly over-confident rhythm of an LLM. Edit for voice and specificity.

With those out of the way, here are the prompts that work.

1. Prospect research and funder alignment

Before you write a word, you have to know what the funder actually funds and how they score. AI is great at digesting a funder's guidelines into a checklist.

Prompt: funder alignment checklist

You are a grant strategist. I will give you a funder's guidelines and our organization's program. Produce an alignment analysis.

Funder guidelines (pasted or summarized):
[paste guidelines — priorities, eligible activities, funding range, deadlines, scoring rubric if given]

Our organization and proposed program:
- Organization: [one-line mission]
- Program: [one-line description]
- Budget request: [amount]
- Target population: [population]
- Geography: [location]

Tasks:
1. Alignment score: 1-10 on each of the funder's stated priorities, with one sentence on why.
2. Eligibility flags: any explicit "no" that disqualifies us, and any gray area.
3. Fit risk: anything that could be read as off-priority by a reviewer.
4. Recommended framing: 2-3 angle options for how to pitch this program to maximize alignment, with pros and cons of each.
5. Missing information: what we need to gather before drafting that the guidelines imply but don't state.

This single prompt replaces an hour of squinting at PDFs. Run it on every funder you're seriously considering and you'll write fewer doomed proposals.

Prompt: scoring rubric reverse-engineering

Below is a funder's stated review criteria with point values (or relative weight). Reverse-engineer how a top-scoring proposal allocates space and emphasis.

Criteria:
[paste]

For each criterion:
1. Maximum points or weight.
2. What a top-scoring response demonstrates, in concrete terms (not "clear narrative" — say "specific outcome data with source, named staff responsible, timeline").
3. Common reasons proposals lose points on this criterion.
4. Suggested word allocation for a [total word count] proposal, given the weighting.

Then produce a section-by-section outline of the proposal prioritized by where points are won.

Most losing proposals lose not because they're bad, but because they spend 60% of the words on the section worth 15% of the points. This prompt surfaces that imbalance before you write.

2. Needs assessment and problem framing

The needs statement is where most proposals die. "There is a great need in our community" is not a needs statement. Reviewers want evidence, specificity, and a logical line from the problem to your program.

Prompt: structured needs statement

Draft a needs statement for a grant proposal to [funder name], maximum [word count] words.

Context:
- Population: [specific population, with geography and any demographic specifics]
- Problem: [the specific problem, not the category — e.g. "85% of 9th graders at [school] are reading below grade level", not "low literacy"]
- Existing data we have (use exactly these, do not invent): [paste any stats, citations, community input, prior program data, all with sources]
- Root causes (our analysis): [list]
- Gap between current state and desired state: [describe]
- Why this problem, this population, this geography, this moment: [urgency / external events / policy change]

Structure the statement as:
1. A one-sentence hook that frames the problem and population.
2. Evidence: 2-4 sentences with sourced statistics (only from the data I provided; mark any number that is not in my data as [UNVERIFIED — REPLACE]).
3. Root cause analysis: 2-3 sentences on why this problem persists.
4. The gap: what current services or programs don't address.
5. The opportunity: why our proposed response is positioned to work.

Rules:
- No "devastating," "staggering," "heartbreaking" — adjectives aren't evidence.
- No invented statistics. If I didn't supply a number, leave a [INSERT SOURCE: X] placeholder.
- Plain language. A non-expert reviewer should follow it.

The placeholders are the key thing. They force you to confront what you don't have — which is usually a lot. Better to know now than when a reviewer docks you for unsupported claims.

Prompt: data-driven problem reframing

Here is a problem statement we drafted. Reframe it to be sharper, more specific, more evidence-driven, and shorter. Do NOT add new data — work only with what's in the original.

Draft:
[paste]

Tasks:
1. Identify every vague adjective ("significant," "widespread," "many") and replace with either the specific number from the original or a [INSERT DATA] placeholder.
2. Cut at least 20% of the words without losing any claim.
3. Reorder so the most compelling single fact appears first.
4. Suggest 2 alternative opening sentences.
5. Output the rewritten version, then a before/after word count.

3. Objectives and methods

Reviewers score objectives on whether they are SMART (specific, measurable, achievable, relevant, time-bound). AI is excellent at turning soft objectives into SMART ones — but only if you force it.

Prompt: SMART objectives

Rewrite these draft program objectives as SMART objectives suitable for a grant proposal. Each objective must be Specific, Measurable, Achievable, Relevant, Time-bound.

Draft objectives:
[paste]

Program context:
- Duration: [months]
- Participants served: [n]
- Setting: [describe]
- Existing capacity: [e.g. "we currently serve 200 youth/year"]

For each objective, output:
- SMART objective (clean, single sentence).
- The measurable indicator (what number goes up, by how much, by when).
- Data source (how we'll measure it).
- A "stretch" version and a "floor" version, so reviewers see a credible target.
- Why this objective matters to the stated problem (one sentence).

Constraints:
- No verbs you can't measure: "improve," "enhance," "build capacity" are banned unless followed by a measurable indicator.
- No "by at least X%" without a baseline.
- Achievable: each objective must be plausible given the stated capacity and budget.

Prompt: methods section

Draft the methods section for a proposal to [funder], [word count] words maximum.

Objectives (already SMART):
[objectives]

Program design (bulleted summary):
- Core activities: [list]
- staffing model: [roles, FTE]
- delivery setting: [e.g. "in-school, after-school"]
- frequency and duration: [e.g. "12 weeks, 3 sessions/week"]
- curriculum or intervention: [name and brief description]
- partnerships: [named partners, their role]

Constraints:
- Map each activity to at least one objective explicitly ("Activity X addresses Objective 2 by…").
- Justify why this approach over alternatives (1-2 sentences).
- Address feasibility — staffing, recruitment, retention, logistics — for each major activity.
- If an activity depends on a partner, name the partner and what they're contributing.

Output as subsections with H3 headers. No fluff. No "we believe" or "we are passionate about."

4. Logic models and theory of change

A logic model is one of the most useful tools in grant writing, and one of the most avoided because formatting it is painful. AI drafts them fast.

Prompt: logic model draft

Draft a logic model for our program.

Inputs:
- [resources — staff, funding, facilities, partnerships]

Activities:
- [list core activities from methods section]

Outputs:
- [direct products of activities — e.g. "120 youth complete 12-week curriculum"]

Outcomes (short-term, 0-12 months):
- [from objectives]

Outcomes (medium-term, 1-3 years):
- [if applicable]

Long-term impact (3+ years):
- [if applicable — be honest, most programs can't credibly claim long-term impact]

External factors and assumptions:
- [list things that must be true for the logic to hold — partner stability, recruitment working, no policy changes]

Produce as a table: Column | Definition | Our content. Then a 3-4 sentence narrative theory of change underneath that connects the chain in plain English (if we do X, then Y happens, because Z).

Prompt: theory of change narrative

Write a 200-word theory of change narrative for this program. Plain language, no jargon, suitable for a funder reviewer who is not in our field.

Program:
- Problem: [needs statement summary]
- Activities: [list]
- Population: [specific]
- Goal: [outcome]

Structure the narrative as a chain:
1. The problem we're responding to (1 sentence).
2. What we will do (1-2 sentences).
3. Why we believe this approach produces change — root cause → mechanism → outcome (2-3 sentences).
4. What success looks like (1 sentence).

Avoid: "build capacity," "empower," "stakeholder engagement," "leverage" (unless describing actual financial leverage).

5. Budgets and budget narratives

A budget narrative is where bad proposals get caught. The numbers in the narrative have to match the line items, and the narrative has to justify each cost as reasonable and necessary. AI is helpful here because it forces consistency.

Prompt: budget narrative

Draft a budget narrative for this grant budget. Each line item below should produce a paragraph that:

- Names the line item and the amount.
- Explains what it covers specifically (not "salaries" — "Program Coordinator, 0.5 FTE, $X for 12 months at $Y annual salary").
- Justifies the cost as reasonable (market rate, prior year actual, comparable to peer org).
- Links the cost to a specific program activity or objective.

Budget table:
[paste budget as table — line item, amount]

Assumptions and rates I want used:
- Salary assumptions: [paste]
- Fringe rate: [%]
- Indirect rate: [% per negotiated rate or funder cap]
- Cost per unit (e.g. per participant, per session): [if known]

Rules:
- The narrative total must match the budget total. State both numbers at the end.
- Do not pad. If a line is a flat $500 for snacks, say so plainly.
- Flag any line item where the assumption seems missing or implausible with [VERIFY: reason].

The "narrative total must match budget total" instruction will catch transposition errors that would otherwise cost you points.

Prompt: budget stress-test

You are a skeptical reviewer. I will give you a program budget and narrative. Find every weakness a critical reviewer would flag.

Budget:
[paste]

Budget narrative:
[paste]

Program duration: [months]. Participants: [n].

Tasks:
1. Per-participant cost — calculate and comment on whether it's plausible compared to similar programs.
2. Staff coverage — is the FTE load realistic for the activities described?
3. Cost categories that look over- or under-funded.
4. Any hidden cost that's missing (e.g. evaluation, indirect, equipment, transportation).
5. Indirect rate — does it match the funder's cap?
6. Sustainability red flags — costs that imply the program can't continue without this grant.

Output as a numbered list of issues with severity (Blocker / Major / Minor) and a suggested fix for each.

6. Evaluation plans

Reviewers want to see you can measure what you claimed you'd achieve. The prompt below produces an evaluation plan tied explicitly to your SMART objectives.

Prompt: evaluation plan

Draft an evaluation plan for the program described below. Maximum [word count] words.

SMART objectives:
[objectives]

Program activities:
[activities]

Existing evaluation capacity:
- Data we currently collect: [list]
- Tools we currently use: [e.g. "pre/post surveys, attendance logs, program staff observation]
- Staff time for evaluation: [FTE]
- External evaluator: [yes/no, and if yes name]

Output sections:
1. Evaluation questions — 3-5 questions tied to objectives.
2. Indicators and measures — table: evaluation question | indicator | data source | frequency | responsible staff.
3. Data collection plan — who collects what, when, with what tool.
4. Analysis approach — how data will be analyzed (descriptive stats, comparison to baseline, qualitative coding).
5. Reporting and use — how findings will be used internally, when, and reported to funder.
6. Limitations — what this evaluation can and cannot tell us. Be honest.

Rules:
- Every objective must have at least one indicator.
- No evaluation approach that requires capacity we don't have. If a method needs a skill or system we lack, flag it and suggest a realistic alternative.
- Distinguish outputs (what we count) from outcomes (what changes). Reviewers conflate these — your plan should not.

7. Sustainability

Funders want to know their money won't evaporate after the grant period. Most proposals lie here. AI can help draft a more credible sustainability plan if you give it real options.

Prompt: sustainability section

Draft a sustainability plan, [word count] words max. The plan must be credible, not aspirational.

Program: [description]. Grant period: [months]. Annual operating budget after grant: $[amount].

Realistic revenue options the organization has (only list ones that are actually plausible — do not invent):
- [option 1: e.g. "state contract with [agency] renewable annually at $X"]
- [option 2: e.g. "individual giving program currently raising $Y/year, growth plan to $Z"]
- [option 3: e.g. "fee-for-service sliding scale"]
- [option 4: e.g. "in-kind and volunteer"]
- [option 5: e.g. "additional grants — list realistic prospects"]

For each revenue option:
- Expected annual contribution, with assumption.
- Probability (high/medium/low) and why.
- What has to happen for it to materialize.

Then:
- A "bridging gap" section — if these revenue sources sum to less than the post-grant operating budget, say so plainly and describe the bridging strategy.
- A "wind-down plan" — if revenue doesn't materialize, what happens to participants.

Reject the standard "we will diversify our funding" boilerplate. Specific numbers and named prospects only.

The "wind-down plan" requirement is what makes this prompt powerful. Reviewers rarely see it, and it signals maturity.

8. The reviewer-critique pass

The most valuable single prompt in this guide. Run it on every draft before submission.

Prompt: hostile reviewer critique

You are a competitive grant reviewer scoring this proposal against the rubric below. Be specific and critical — do not be polite.

Rubric:
[paste funder scoring criteria with point values]

Proposal text:
[paste full proposal draft]

Tasks:
1. Score the proposal on each criterion 1-5 (or as the rubric specifies), with a one-sentence justification per criterion.
2. For each criterion, list specific weaknesses with the section/paragraph reference.
3. Identify the single biggest reason this proposal would be declined.
4. Identify the three highest-leverage edits that would lift the score most.
5. Note any claims that are unsupported, vague, or contradict another part of the proposal.
6. Word count check — is the proposal within the stated limit per section?
7. Final verdict: Fund / Fund with revisions / Decline / Borderline.

Be the reviewer you don't want to meet. We need the truth before submission, not a cheerleader.

The "be the reviewer you don't want to meet" line consistently produces more useful criticism than "be critical." It works because models respond to vivid instruction.

Prompt: gap audit

List every requirement from the funder's application guidelines below. For each, show whether the proposal addresses it (Yes / Partial / Missing) with the section reference. If Partial or Missing, suggest the smallest edit that covers it.

Guidelines checklist (extract every "must include," "should describe," "applicants will" requirement):
[paste guidelines]

Proposal:
[paste]

This catches the "we forgot the IRS determination letter attachment" failures that happen to everyone, including people who've written 50 proposals.

Patterns that make these prompts work

Across all of them, a few common moves:

Give the model the funder's exact language. Proposals get declined for missing one buried sentence in the guidelines. Pasting the guidelines verbatim is the highest-yield single thing you can do.
Forbid invention explicitly. "Do not invent statistics," "only use the data I provided," and "use [placeholder] for unknowns" all dramatically reduce hallucinated facts in a domain where fabricated data is fatal.
Lock the output structure. Tables, sections, and explicit headers prevent the model from drifting into prose when reviewers want a grid.
Map every claim to a source. The prompts force every number back to a named source. If you don't have one, the placeholder stares at you, which is exactly what should happen.
Decompose the proposal. Run separate prompts for needs statement, objectives, methods, budget, evaluation, sustainability — not one giant "write the proposal" prompt. Staged prompts produce vastly better sections, and you can iterate on each independently.
Use the critique prompt last, then revise. The hostile reviewer pass finds things you've gone blind to. Always budget time for at least one revision cycle after it.

Mistakes to avoid

Pasting last year's proposal and asking AI to "rewrite it for this funder." This produces a dressed-up version of last year's proposal, not a proposal aligned to this year's funder. Start from the funder's guidelines, not your last submission.
Letting AI write the needs statement from thin air. If you don't feed it your data, it will make data up. Always paste your verified stats or use placeholder tags.
Skipping the SMART pass on objectives. Soft objectives are the second most common decline reason after weak needs statements. Always run the SMART prompt.
Filing the budget narrative as an afterthought. Reviewers read it carefully. Use the budget stress-test prompt before you finalize.
Forgetting the sustainability wind-down plan. It's rare, which means including it signals seriousness and reduces reviewer anxiety about post-grant collapse.
Submitting without running the hostile reviewer pass. Even one round of critique-driven revision meaningfully improves scores.

A realistic workflow

For a typical $50,000-$500,000 proposal, the workflow looks like this:

Prospect research — run the funder alignment checklist before committing to apply.
Run the scoring rubric reverse-engineering prompt to plan section weighting.
Draft the needs statement from your own data, using the structured needs statement prompt.
Stress-test the needs statement with the reframing prompt.
Generate SMART objectives, then methods, then logic model in sequence — each one builds on the previous.
Draft budget narrative only after methods are settled, so the budget reflects the actual methods.
Run the budget stress-test prompt.
Draft the evaluation plan tied to SMART objectives.
Draft the sustainability section with real revenue options, including a wind-down plan.
Assemble the full proposal, then run the hostile reviewer critique prompt on the whole thing.
Run the gap audit against the funder's guidelines checklist.
Revise once based on the critique, verify every statistic and citation against source materials, then submit.

Used this way, AI cuts proposal drafting time meaningfully — usually in half for experienced writers and more for newer ones — while making the proposal better-aligned and better-defended. The judgment work stays with you: funder selection, program design, partnership reality, the courage to write a credible wind-down plan. AI handles the parts that are mostly formatting and structure, freeing you to spend your time on the parts that actually win grants.

For a versioned library of these grant-writing prompts you can edit, share across your team, and trigger from a single workspace, sign up at Promptwright and get started today.

AI Prompts for Grant Writing: Templates for Proposals, Budgets, and Narratives