You found a grant that fits. Now what?

Start here: a search result is a hypothesis, not a fit. When our tool surfaces a grant, what you have is a lead worth an hour of checking, not a project worth six weeks of writing. The most expensive mistake small organizations make is pouring effort into a grant they were barred from on day one. So the first job after "this looks like a fit" is not to start writing. It is to try to rule yourself out quickly, before you spend a single evening you cannot get back.

The one thing to know about AI here

Before you use AI, find out what the funder allows

Everything below assumes something you have to confirm yourself: that this funder permits AI on this application. Agencies set their own rules and they do not agree with each other. Some ask you to disclose AI use. Some treat an application substantially written by AI as not your own work. Some say nothing at all. The only rule that governs you is the one attached to the money you are asking for.

Two real policies show how wide the gap is. NIH states it will not consider an application substantially developed by AI to be the applicant’s original idea, and use discovered after an award can be referred to the Office of Research Integrity (NOT-OD-25-132, issued July 2025). NSF only encourages proposers to note in the project description whether and how they used generative AI (notice dated December 2023). Same federal government, opposite postures. Neither one tells you what your funder requires.

Look in three places before you draft: the funding notice itself, the agency’s application guide or general terms, and the instructions on the submission portal. If all three are silent, email the program officer and ask. Put the question in writing and keep the reply, because a written answer is what protects you if the question ever comes up again.

Do this first: confirm what this funder permits before you put a single sentence of this application through an AI tool. Policies change between cycles, so check the notice you are actually applying to. Do not rely on what the agency said last year, and do not rely on this page.

The workflow, start to submitted

1. 01

   Try to disqualify yourself first

   Use AI

   Open the actual funding notice and hunt for the reasons you cannot apply: who is eligible, entity type, geography, any exclusion for past recipients, required certifications, and the cost-share (match) you would have to fund yourself. If any single gate rules you out, stop now and move to the next lead. An hour spent here can save you six weeks. While you have the notice open, find its AI terms too, as described above.

   AI here: good for pulling the eligibility rules out of a long document into a short list. Checkpoint: it also misreads eligibility, so confirm every gate against the notice's own words before you trust it.

2. 02

   Confirm you can even submit

   Do it yourself

   Federal submission runs through registrations that take real time and expire. You need an active SAM.gov registration (which assigns your organization a Unique Entity ID and must be renewed every year), a Grants.gov account, and a Login.gov account to sign in to both. New registration can take a couple of weeks or longer if anything needs to be verified, and it is free. If you are not registered, you cannot submit, no matter how good the application is. Start this the day you decide to pursue a grant, not the week it is due. Links to the official sites are at the bottom of this page.

   AI here: not the tool. It will invent current processing times and steps. Go straight to the official sites and follow their instructions.

3. 03

   Read the whole notice, then build a compliance matrix

   Use AI

   The short summary is not the rules. Read the full notice and pull out every requirement: scoring criteria and their point values, required activities, page limits, formatting rules, required attachments, the match, allowable and unallowable costs, and any deadline quirks. Turn all of it into one checklist you work against until the day you submit. This single document is what keeps a small team from missing the requirement that gets them screened out.

   AI here: the strongest use in the whole process. Paste the notice and ask it to build a requirements checklist with point values and every attachment. Then read the notice yourself to catch anything it dropped.

4. 04

   Make an honest go / no-go call

   Do it yourself

   Before you draft, decide as a team: can we actually win this, fund the match, staff the work, collect the data it demands, and hit the deadline without wrecking everything else? A grant you cannot deliver on is worse than one you never won. Saying no to the wrong grant is a skill, not a failure.

   AI here: it can lay out the questions to ask, but it cannot make this call. It does not know your capacity, your cash, or your people. The decision is yours.

5. 05

   Gather your evidence and real numbers

   Do it yourself

   Now collect what the application actually rests on: need data for your area, your track record, real client counts, budget figures, and commitment letters from partners. Every statistic needs a source you can name and link. Every partner contribution needs a real person who agreed to it.

   AI here: the single most dangerous step. AI fabricates statistics and citations that look flawless, complete with plausible source names that do not exist. Never let it supply a number or a reference. Pull those from primary sources yourself.

6. 06

   Assign the work and count backward from the deadline

   Use AI

   Take the real due date and work backward to today, leaving room at the end for review, signatures, and the technical submission itself. Give each section an owner and an internal deadline that is earlier than you think you need. Note early who has to sign: the person authorized to submit on behalf of the organization is often not the same person running the project.

   AI here: good at building a reverse timeline from your due date and turning your checklist into an owner-by-owner task list. Adjust it to your team's real availability.

7. 07

   Draft to the scoring rubric, not to the blank page

   Use AI

   Reviewers score against published criteria, so write against those criteria in the order and proportion they are weighted. Put your heaviest effort where the most points live. Answer every scored question plainly and directly. Make it easy for a tired reviewer to find each point and award it.

   AI here: genuinely useful for first drafts, outlines, and tightening prose. Two checkpoints: reviewers can feel generic AI writing, so rewrite it in your own organization's voice; and current agency priorities change, so make sure the language reflects this cycle's stated priorities, not last cycle's (see the note below).

8. 08

   Build and document the budget

   Do it yourself

   Every dollar has to be allowable under the notice, tied to an activity, and explained in the justification. Confirm your match, and name where it comes from. Check the notice for what it will and will not pay for before you build a line around it.

   AI here: fine for the shape of a budget and a justification template. Never for the arithmetic or your actual costs. AI makes quiet math errors and cannot know your real salaries, rates, or match.

9. 09

   Review in three separate layers

   Use AI

   Do not review everything at once. Pass one: check the draft against your compliance matrix, is every required piece present and within limits? Pass two: read it as a skeptical reviewer scoring against the rubric, where would you dock points? Pass three: plain proofreading. Then have a human who has not seen it read the whole thing.

   AI here: a strong compliance checker and a useful skeptical reviewer. Ask it to check your draft against the requirements list and to argue why a reviewer might score each section low.

10. 10

    Submit early, on purpose

    Do it yourself

    Submit days before the deadline, not hours. Portals get slow near the close, uploads fail, validation throws errors you did not expect, and some programs stop accepting applications once they hit a cap. An application finished on time but submitted late scores zero. Give yourself a full business day of cushion.

    AI here: not a factor. This is a discipline step. Put the internal deadline days ahead of the real one and hold it.

A note on language and framing

Federal priorities, and the words reviewers reward, shift with each administration and each funding cycle. The safest move is to find the funding agency's current strategic plan and stated priorities before you write, and to mirror that language honestly wherever your work genuinely fits it.

This is a place to be especially careful with AI. Because it learns from older text, it will often hand you the previous cycle's vocabulary, or its best guess at "what funders want to hear," which can be out of step with what the agency is actually asking for now. Anchor your framing to the current notice and the agency's current priorities, in your own words, not to an AI's memory of what grants used to say.

The three accounts you cannot submit without

All three are free. Any site that charges you to register is not the official one. Timelines and steps on these systems change, so treat the sites themselves as the source of truth.