AI Grant Writing Tools Compared for Small Non-Profits

For most small non-profits, a general assistant on a non-profit plan does the writing well: ChatGPT Business or Claude Team at $8 a seat a month, or Gemini at no extra cost with Google Workspace for Nonprofits. Pay for a specialist tool such as Grantable, Grantboost or Instrumentl only when you also need funder discovery and tracking, or write many bids a year.

Three things decide it: how many applications you write a year, whether you need help finding funders or already know who you'll apply to, and whether a tool's funder database actually covers your funders. Several specialist databases are built largely from one country's public tax filings, so a charity applying to funders elsewhere should test the search on names it knows before paying for it.

Follow me on Instagram@sagnikteaches

Grant-writing tools and list prices side by side

List prices in USD as checked in September 2026. Promotions and non-profit discounts change, so confirm on each vendor's pricing page.

Connect on LinkedInSagnik Bhattacharya

What each option does well, and where it falls short

A general assistant on a non-profit plan

ChatGPT, Claude and Gemini all handle the core of grant writing: extracting criteria from guidance, drafting to a limit from your evidence, checking a budget narrative against the table, and critiquing a draft as an assessor. Stored in a Project or Gem, your evidence pack is available to every chat. The non-profit prices are low: two Claude Team seats come to $192 a year, and Workspace for Nonprofits adds Gemini at no extra cost. Claude's non-profit plan also includes connectors to Candid, Blackbaud and Benevity; check what Candid data your account can reach before relying on it for prospecting.

Subscribe on YouTube@codingliquids

The job to test first is criteria extraction, because every later draft depends on it. A prompt that works in any of the three:

Here is a funder's guidance for its community grants round.
List, as a table: every question applicants must answer, with its
word or character limit; every attachment required; every
eligibility rule; the deadline and how to submit. Quote the
guidance's own wording for each question. If a limit isn't stated
or something seems to be missing, say so. Don't guess.

An illustrative extract of what comes back:

Q3  "Describe the need your project addresses"      300 words
Q4  "How will you know the project has worked?"      250 words
Q5  "Budget justification"                           No limit stated
Attachments: latest annual accounts; safeguarding policy
Eligibility: registered charities, annual income under 250,000
Deadline: 14 November, 12 noon, via the online portal

Useful, and incomplete in two ways that are typical. The funder's separate FAQ document caps Q5 at 2,000 characters in the portal, which the model couldn't know because only the main guidance was pasted. And a footnote asking for a letter of support from a partner school was missed. The fix is procedural rather than a better prompt: paste every document the funder publishes, including the FAQ and any portal screenshots, then tick the table off against the portal itself before drafting a word.

One pricing quirk catches one-person fundraising teams. ChatGPT Business and Claude Team both need at least two seats, so a lone fundraiser pays for two. On non-profit pricing that's $16 a month, which is still less than a single individual Plus or Pro plan at $20. If your organisation qualifies, the second seat can go to the director who signs off bids, and the evidence pack sits in a shared Project both can use.

Whatever sits in the Project or Gem does most of the work, so it's worth being specific about what goes in. A filled-in evidence pack for an illustrative animal rescue:

• Outcomes table, last three years: animals taken in, rehomed and returned, average days to rehoming, with the source spreadsheet named.
• Latest annual report and accounts, as PDFs.
• Approved paragraphs: mission, history, governance and safeguarding, each signed off by the chair and dated.
• Three case studies with adopters' written consent, first names removed.
• Budget template with current costs: vet fees per animal, kennel costs, volunteer expenses.
• Past applications, successful and unsuccessful, with any funder feedback.

Where they fall short: no grant calendar, no pipeline, no funder database of their own, and they only know your organisation through what you upload. You bring the structure, which is what writing grant applications with AI, step by step sets out. Eligibility for non-profit pricing depends on each programme's verification rules; AI tool discounts for charities and non-profits covers how to claim them.

Grantable

A workspace built for grant writers: it keeps your organisation's past content, drafts in your voice, helps find funders and tracks applications. The pricing is unusual in having no per-seat fees, so a volunteer committee can share one account. The free plan's five chat messages a day is enough to evaluate it on a real application over a week, but not to write bids in earnest; Starter raises that to 50 a day, Pro to 200. Grantable advertises a first-year discount for small non-profits that meet its eligibility rules, so check whether you qualify.

Five messages a day goes a long way if you plan them before you log in. A realistic first day on a live bid:

1. Paste the funder's guidance and ask for the questions, limits and attachments as a list.
2. Ask for a draft of the hardest question from the content you've saved, at its word limit.
3. Ask it to cut that draft by 20% without losing any figure.
4. Ask it to score the draft as the funder's assessor would, against the published criteria.
5. Ask it to list every number in the draft and where in your saved content each came from.

A week of this shows it extracting, drafting, editing, critiquing and source-checking your own material, which is a fairer test than a demo. Check too whether funder searches count against the daily message limit, so you can fit a discovery test in.

Good fit: one or two people writing a steady stream of mid-sized bids. Test first: whether its funder search covers the funders you actually apply to.

Grantboost

Discovery and drafting at a lower price point, with limits set by seats and "active projects" (grants you're working on at once): Starter has one seat and five active projects, Growth five seats and 20, Team unlimited seats and 50 or more. Brand voice training starts on Growth. For a charity running a handful of applications at any time, the project caps rarely bite.

They can bite sooner than you'd expect, though, depending on what counts as active. Funders often take two or three months to decide, so bids pile up. An illustrative autumn for a youth charity on Starter: two bids in drafting, four submitted and awaiting decisions, and one funder saved for a spring round. If submitted-but-undecided bids still count, that's six or seven active projects against a cap of five, and the charity either archives bids it may need to report on or moves up to Growth. Ask the vendor exactly when a project stops counting before you choose a tier.

Good fit: small teams who want discovery and drafting in one place without Instrumentl's price. Test first: the quality of matches against your mission, and whether drafts stay within your evidence.

Instrumentl

Primarily a prospecting and grant-management platform with a large funder database (it cites 450,000+ funder profiles), deadline tracking and, on Full Lifecycle, post-award budget tracking. Its AI drafting tool, Apply, starts on the Pre-Award plan at $499 a month billed annually for up to five users, so AI writing alone is an expensive reason to choose it. The Discover plan ($299 a month billed annually) includes prospecting but no AI drafting.

That gap is worth a quick sum. Pre-Award costs $5,988 a year on annual billing. Discover costs $3,588, and two Claude Team seats on non-profit pricing add $192, so prospecting in Instrumentl plus drafting in a general assistant comes to $3,780: about $2,200 a year less. Spread across bids, Pre-Award works out at roughly $200 an application for an organisation writing 30 a year and nearly $500 for one writing 12. Those figures only make sense if the pipeline and deadline tracking are saving real staff time, or if keeping drafting inside the same platform as the funder records matters to the team.

Good fit: an organisation with a development team, many applications and restricted grants to manage. Test first: coverage of your funders during the 14-day trial.

Grant Assistant

A specialist drafting tool that compares drafts against funder requirements to find gaps. The company says it was built on thousands of successful proposals and that customer data isn't used to train its models. Pricing is by quote, based on organisation size, which makes it harder to compare; ask for a quote alongside a trial on a real application.

The gap-checking is the part to judge. On a trial, an illustrative result against a funder's criteria might read: "Criterion 4, 'show how the people you support helped shape this project': not addressed. The draft describes activities but no consultation." That's a genuinely useful catch, since general assistants will only find it if you ask. What to check is whether the gaps it flags are really in the funder's guidance, or are generic best practice it applies to every bid. Compare three of its flags against the guidance line by line.

A quote request that gets you a comparable price in one reply:

"We're a charity with annual income of about [amount], two staff who write bids, and around 15 applications a year, mostly to trusts and foundations, with two or three larger public-sector tenders. Please quote annual and monthly prices for that use, say what's included (users, applications, compliance checks), and confirm whether our uploaded documents are used to train any model."

Good fit: organisations writing complex, compliance-heavy proposals. Less obvious fit for a small charity writing short foundation bids.

Five questions that decide it for your organisation

1. How many bids a year? Under about six: a general assistant is almost always enough. Six to twenty: a specialist tool's library and tracking start to save real time. Twenty or more with several writers: a platform like Instrumentl becomes easier to justify.
2. Do you need to find funders? If you already know your funders, discovery features are paying for nothing. If prospecting is your bottleneck, discovery is the main reason to buy a specialist tool.
3. Does the database cover your funders? Search for five funders you've applied to before. If fewer than four appear, the discovery feature isn't for you, whatever the headline numbers.
4. How many people write? Per-seat pricing favours small teams on general assistants; Grantable's no-seat model favours volunteer committees.
5. What's in your existing subscriptions? If you're on Google Workspace for Nonprofits, Gemini is already there. Try it on one application before buying anything.

Three organisations, three different choices

A volunteer-run animal rescue writing four bids a year, all to funders it already knows. It already uses Google Workspace for Nonprofits, so the choice was Gemini with the evidence pack in Gemini Notebook, at no added cost. The volunteers built a criteria-extraction and drafting prompt into a Gem. Annual tool cost: $0. One set-up detail mattered: when a Gem is shared, the people it's shared with can see its instructions and the files uploaded to it. So the Gem holds the evidence pack and approved paragraphs, which any volunteer may see, while the rescue's adoption records and volunteer contact list stay out of it entirely.

A community music charity with one part-time fundraiser writing about 12 bids a year and struggling to find new trusts. The fundraiser trialled Grantable free for a week on a live bid and Grantboost's trial on the same one, searching both for funders she already knew. Grantable found most of them; she chose Starter at $50 a month, roughly $600 a year, and kept Claude on a personal plan for critique. The deciding factor wasn't the writing, which was similar in both. It was discovery coverage.

A tutoring charity with a three-person development team writing 30-plus applications a year, many of them restricted grants with reporting requirements. For them, Instrumentl's Pre-Award plan ($5,988 a year billed annually) or Full Lifecycle made sense because pipeline, deadlines and post-award tracking were the real pain; AI drafting was a bonus. They kept ChatGPT Business on non-profit seats for other writing.

Before committing, the tutoring charity ran the coverage test during Instrumentl's trial. All five of its existing funders appeared. The search then suggested 14 new ones; after reading each funder's own guidance, seven were genuinely eligible, four had closed their programmes or funded only in other areas, and three were a poor match for tutoring. Seven real prospects from one afternoon justified the price for a team writing 30 bids. For the music charity, the same seven would have been welcome but wouldn't have covered $6,000 a year.

How to trial two tools on one real application

Don't judge on demos. Run the same live application through two options in the same week, and score each on this script:

1. Criteria extraction. Paste the funder guidance. Did it capture every question, limit and attachment?
2. Drafting to a limit. Ask for one answer at your character limit from your evidence. Is it within the limit, and does every number come from your documents?
3. Invented facts. Search the draft for statistics you didn't supply. Any at all is a warning sign about how the tool is prompted.
4. Voice. Would your director sign it without rewriting the first paragraph?
5. Discovery (specialist tools only). Of five funders you know, how many did it find? Were its new suggestions actually eligible?
6. Export. Can you get your content out in Word or plain text?

An hour per tool is enough. Keep the scores; they'll help you at renewal time, when a tool that scored well on discovery but badly on drafting might be worth keeping only for its funder search.

The music charity's fundraiser scored her two trials like this (1 to 5, illustrative):

Writing quality was close enough to call a draw. Discovery wasn't, and discovery was the problem she was paying to solve.

The invented-facts test deserves more than a glance, because it's where a bid can do real damage. In a realistic example, a draft need statement included "one in four children in the area has no access to music tuition". It read well and was nowhere in the evidence pack; the model had produced a plausible statistic to strengthen the paragraph. It only came to light because the fundraiser searched the draft for every figure and matched each to a source. Add a standing instruction to whichever tool you choose: "After every statistic, put its source from my documents in brackets. If there isn't one, write [NEED SOURCE] instead of a number." Unsupported claims then show up as gaps rather than confident sentences.

Checking the choice at the first renewal

Three months in, compare the tool against what you bought it for, using numbers you can pull from your own records. The music charity's first-quarter review with Grantable Starter (illustrative):

The decision: keep it. The new funders were the reason for buying and they materialised. Had the search found nothing new, the right move would have been to drop to Grantable's free plan or a general assistant, because the drafting on its own was no better than what she already had. Don't count the award as proof either way; one grant in a quarter is too small a sample to credit or blame the tool.

What to avoid when buying

• Paying for discovery you don't use. The most common overspend. If your funder list is stable, you're buying a database you'll never search.
• Tools that promise to "write the whole application". The draft is the cheap part. Evidence, fit and tailoring win grants, and funders' attitudes to AI-written bids are covered in using AI without funders rejecting your bid. Some funders now ask applicants to declare AI use; a plain declaration that fits most forms: "We used AI tools to help structure and edit this application. All facts, figures and commitments were written or checked by our staff, who are responsible for its content." Read each funder's guidance first, because a few restrict AI use outright, and no tool choice gets around that.
• Win-rate claims. Vendor success stories are marketing, not evidence about your organisation.
• Uploading beneficiary data. Aggregated numbers and consented, anonymised quotes only, whichever tool you choose.
• Lock-in. Keep your evidence pack and approved paragraphs in your own drive. Keeping your data and prompts portable explains how.

If budget is the constraint, using AI with almost no budget shows how far the free and discounted options go, and for help choosing between the two main assistants for long-form writing, see Claude versus ChatGPT for business writing.

Choosing a grant tool: common questions

Is a free option good enough for grant writing?

Often, yes. Google Workspace for Nonprofits includes the Gemini app and Gemini Notebook at no cost for eligible organisations, and Grantable has a free plan with five chat messages a day. For an organisation writing a handful of bids a year from a solid evidence pack, that can be enough. Free options fall short when you need funder discovery, several people working on bids, or heavy daily use.

Do grant writing tools train their AI on our applications?

Policies differ by vendor and change, so read the data terms of each one before uploading anything. Business and non-profit plans of ChatGPT and Claude don't train on your content by default. Specialist vendors often state that customer data isn't used for training, but confirm it in the terms rather than on a marketing page, and never upload identifiable beneficiary information.

Can we switch tools later without losing our work?

Only if you keep your own copies. Store your evidence pack, approved paragraphs and past applications in your own drive, and treat any tool's library as a working copy. Before committing to a specialist tool, check that it exports applications and saved content in a standard format such as Word or plain text.

Further reads

• Questions to Ask Before Buying AI That Touches Client Data — Questions to put to any vendor before uploading your documents.
• How to Review an AI Tool After 90 Days: Keep, Fix or Cancel — Decide whether to keep, fix or cancel after the first quarter.
• What to Check in an AI Tool's Privacy Policy and Terms — What to read in a grant tool's privacy policy and terms.
• Charity AI Risks: Data Protection, Deepfakes, and Donor Trust — The data and trust risks that matter most for charities.
• Do Small Charities Need Outside Help to Adopt AI? — When it's worth getting help to set tools up.
• Donor Thank-You Letters With AI That Still Feel Personal — Another fundraising job the same general assistant handles well.
• Turning Charity Data Into Impact Reports With AI — From attendance logs and survey scores to a report a funder trusts: calculate with AI, check every figure and claim only what you can show.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Grantable, Grantboost and Instrumentl pricing pages; Grant Assistant FAQ; OpenAI help page on OpenAI for Nonprofits; Claude for Nonprofits guide; Google for Nonprofits help page on AI features (all checked September 2026). Organisation examples are illustrative.