Skip to content
Phase 7: Review, Compliance and Submission

AI Grant Proposal Red Team Prompt

By · Updated · Free to copy and use
Quick Answer

This AI grant proposal red team prompt reads your draft from several skeptical points of view, such as a subject-matter reviewer, a finance reviewer, a program officer and a member of the community served. For each one it lists the strongest objections, the claims that reader would not believe and why, contradictions between sections, and what evidence would change their mind. It then gives one ranked list of vulnerabilities. It finds weaknesses and does not fix them.

The prompt

Copy and paste
You are red-teaming my grant proposal. Your job is to find its weaknesses before a reviewer does. Do not praise it, do not fix it and do not rewrite it.

Use only the draft and the funder information I paste below. Quote the draft for every point you make. Do not invent facts about my organization, my field or the funder.

Read the draft separately from each of these perspectives:
[LIST PERSPECTIVES, OR USE THESE FOUR:
1. A subject-matter reviewer who knows this field and doubts unsupported claims
2. A finance reviewer who reads the budget and asks whether the money matches the work
3. A program officer who cares most about fit with the funder's stated priorities
4. A member of the community this project is meant to serve]

For each perspective, give:
a. The three strongest objections this reader would raise
b. The specific claims this reader would not believe, with the quoted passage and the reason for doubt
c. What this reader would want to know that the draft does not say
d. What evidence or change would be needed to satisfy them

Then, across the whole draft:
e. Contradictions between sections, quoting both passages
f. Promises that nothing in the draft shows we can keep
g. Assumptions the plan depends on that are never stated or defended
h. The single weakest point in the proposal, and why

Finish with one ranked list of vulnerabilities, most serious first. For each, say whether it is a problem of substance (the plan or the evidence), a problem of presentation (the plan is fine and the draft does not show it), or a problem of fit (the project may not match what this funder wants).

If you cannot tell whether something is a real weakness, say what you would need to know.

Funder's priorities and review criteria: [PASTE FUNDER PRIORITIES AND CRITERIA]
Draft proposal, including the budget summary: [PASTE DRAFT]

Written for ChatGPT, Claude, Gemini and similar AI chat tools. Replace each bracketed placeholder with your own details before you send it. Results vary by tool and version, so always check the output.

What to fill in

[LIST PERSPECTIVES, OR USE THESE FOUR ...]
Keep the four defaults or replace them with the readers who matter for this application: a practitioner, an evaluator, a board member, a peer organization. Choose readers who would see different problems.
[PASTE FUNDER PRIORITIES AND CRITERIA]
The funder's stated priorities and review criteria in their words, so the program officer perspective has something real to work from.
[PASTE DRAFT]
The full draft with headings, plus a summary of the budget. A red team can only find contradictions between sections it has been given. Remove confidential and personal information.

Why this prompt works

Red teaming means deliberately attacking your own work to find its weak points while you can still fix them. The term comes from exercises in which one group plays the opponent. For a grant proposal, the opponent is a tired, skeptical reader with a stack of other applications.

  • Several readers, not one. A subject expert, a finance reviewer and a program officer look for different things. Asking for each in turn finds problems a single read would not.
  • No praise and no fixes. AI tools default to encouragement and to rewriting. Both are switched off so the output is a list of problems and nothing else.
  • Claims that would not be believed. This is different from asking what is missing. It targets statements that are present and unconvincing, such as an ambitious outcome with nothing behind it.
  • What would change their mind. Each objection is paired with the evidence that would answer it, which turns criticism into a to-do list.
  • Whole-document checks. Contradictions, unkept promises and unstated assumptions only show when the proposal is read as one piece.
  • Three kinds of problem. Substance, presentation and fit need different responses. A presentation problem is an edit. A substance problem is a decision. A fit problem may mean this is the wrong funder.
  • Quotes required. Every point is tied to your text, so vague criticism is ruled out.

This complements a mock review. A reviewer simulation scores the draft against the rubric. A red team looks for doubts the rubric may not capture.

Why it is appropriate to use

Use this on a full draft, with enough time left to respond. It is an appropriate use of AI because the AI adds nothing to the proposal. It reads and objects, and a person decides which objections are sound.

Two cautions. First, an AI playing a community member is not a community member. It can raise questions such a reader might ask, and it cannot tell you what people in your community actually think. If the project is meant to serve them, ask them. Their answer outranks anything in the output. The same goes for the subject expert perspective, which is no substitute for someone who knows the field.

Second, the AI will always produce objections, because that is what it was asked for. Some will be weak or wrong. A long list does not mean the proposal is poor. Your job is to sort the real vulnerabilities from the noise.

Do not use the output to rewrite the proposal defensively until it promises nothing. A proposal has to make claims. The goal is to back them up. Check the funder's AI policy and remove confidential information before pasting.

How to check and revise the output

  1. Read every quoted passage and decide whether the objection is fair. Discard the ones that misread the draft, but note that a misreading can point to unclear writing.
  2. Sort the ranked list into substance, presentation and fit, and send each item to whoever can act on it.
  3. Take problems of substance to the program team. If a promise cannot be supported, change the plan or scale the promise back.
  4. Fix presentation problems in the text by adding the evidence or explanation the reader was missing.
  5. Take fit problems seriously. If the program officer perspective found a real mismatch with the funder's priorities, reconsider before investing more time.
  6. Resolve every contradiction in both places.
  7. Then ask real people, including someone from the community served and someone who knows the field, to read it.

Frequently asked questions

What does red teaming a grant proposal mean?
It means reviewing your own proposal as a skeptic would, looking on purpose for objections, weak claims and contradictions, so you can address them before the funder's reviewers find them.
How is a red team different from a mock review?
A mock review scores the proposal against the funder's criteria. A red team does not score. It looks for anything a doubtful reader would question, including problems that no criterion names, such as an unstated assumption or a claim that is hard to believe.
Can AI stand in for community members reviewing my proposal?
No. It can suggest questions a community member might raise, which is a useful prompt for your own thinking. It cannot tell you what the people you serve actually think or need. For that you have to ask them.
Should I fix every weakness the red team finds?
No. The AI will always find objections because it was asked to. Judge each one. Fix the real vulnerabilities, clarify passages that were misread and set aside the ones that do not hold up.

More prompts for review and submission

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