AI-Powered Grant Applications: What Actually Works

Honest guide to using AI for charity grant applications in the UK. Why AI only helps when it drafts from real evidence, and how to build that evidence base before you write.

By Plinth Team

Everyone in the charity sector has heard that AI can help with grant writing. Most have tried it. According to the Charity Digital Skills Report 2025, 76% of UK charities have used AI tools at work, with grant fundraising and communications among the most common use cases (Charity Digital Skills Report 2025). But here is the uncomfortable truth: most of them are doing it badly.

The typical approach goes something like this. You open ChatGPT, paste in the funder's criteria, type "write a grant application for a youth mentoring programme," and get back 800 words of fluent, confident prose that says absolutely nothing specific. It sounds professional. It reads well. And any funder who has reviewed more than a dozen applications in 2026 can spot it immediately. Funders have reported that reviewers are now noticing a rise in generic, AI-generated language across applications — and it is not helping applicants.

The problem is not that AI is useless for grant writing. It is that most charities ask it to do the wrong job. They treat AI as the author, when the hard part of a strong application was never the prose: it is having the evidence. Outcome figures, recent case studies and a clearly documented programme are what make an application fundable, and no writing tool can supply them for you. This guide explains how to use AI well — and why the real work happens before you open any writing tool at all.

What you will learn

  • Why most charities are using AI for grant writing incorrectly — and how to fix it
  • The difference between prompting from nothing and drafting from evidence
  • Which parts of an application AI handles well, and which it does not
  • How to build the evidence base that makes any approach to grant writing faster

Who this is for

  • Charity fundraisers and grant writers looking to use AI without producing generic applications
  • CEOs and managers at small to medium charities who want to apply for more grants without hiring more staff
  • Anyone who has tried ChatGPT for a grant application and been disappointed with the results

Why Does Generic AI Grant Writing Fail?

Generic AI grant writing fails for a simple reason: large language models produce text based on patterns, not evidence. When you ask ChatGPT to write a grant application without giving it your specific data, it generates plausible-sounding claims that are not grounded in anything real. "Our programme has demonstrated significant positive outcomes for young people in the community" is the kind of sentence AI produces easily — and it is exactly the kind of sentence that makes funders move on to the next application.

Funders frequently cite weak evidence of need, poor fit with eligibility criteria, and doubts about an organisation's capacity to deliver as common reasons for rejecting applications. AI without data makes this problem worse, not better, because it creates an illusion of completeness. The application looks finished, reads professionally, and covers all the sections — but none of the claims are backed by anything.

Charities that include structured outcome data in their applications consistently outperform those submitting narrative-only proposals. The issue is not whether you use AI. It is whether you have anything worth writing about.

What Is the Difference Between Prompting From Nothing and Drafting From Evidence?

This is the most important distinction in AI grant writing, and most charities have not grasped it yet. The table below illustrates the practical difference.

FactorPrompting from nothingDrafting from your evidence
Data sourceFunder criteria + your promptYour outcome data, case studies, surveys and programme records
Evidence qualityGeneric claims, no verifiable figuresSpecific numbers taken from your actual records
Case studiesFabricated or vague examplesReal beneficiary stories from your case notes
Time to first draft5-10 minutes5-15 minutes, once the evidence is gathered
Time to final version3-5 hours (heavy editing needed)1-2 hours (refinement, not rewriting)
Funder recognition riskHigh — generic language is increasingly flaggedLow — specificity makes it read as authentic
Consistency across applicationsLow — each prompt produces different claimsHigh — same underlying data, tailored framing
Fit with funder requirementsVariable — depends on prompt qualityStronger — you choose the metrics each funder asks for

The difference is stark when you look at actual output. A prompt with no evidence behind it might produce: "Our mentoring programme has helped dozens of young people improve their confidence and life skills." A draft written from the same charity's actual records might read: "In the 2024-25 programme year, 47 young people aged 14-18 completed our mentoring programme. Pre- and post-programme surveys showed a 34% average improvement in self-reported confidence scores, and 89% of participants reported feeling more prepared for employment or further education."

The second version is not better written. It is better informed. And that is what funders are looking for.

The applications that stand out are the ones where the charity genuinely knows its numbers. Whether the text was drafted with AI or by hand, if the claims can be verified against reality, that builds funder confidence. Specificity signals competence — and competence signals that the funding will be well used.

Where Does Plinth Fit In?

It is worth being clear about this, because it is easy to assume otherwise from a guide like this one: Plinth does not write grant applications for you. We used to offer an AI grant writer, but bid writing is no longer part of what Plinth does. You will not draft your next application to a trust or foundation in it.

What Plinth does is the part that comes first. It is where charities record case notes, run outcome surveys, track attendance and produce impact reports. When a deadline arrives, the figures and stories an application needs are already collected, up to date and in one place — so whoever writes the application, with or without AI, is working from fact rather than memory.

The writing itself stays with you: in your own words, in the funder's portal or form, or with whatever general-purpose AI tool your charity has chosen to use. The one place Plinth touches the application itself is when a funder runs its programme on Plinth: their application form may offer to suggest answers from documents you upload, which you then review and edit before submitting. That is a feature of the funder's form, not a tool for writing applications elsewhere.

How Should Charities Actually Use AI for Grant Applications?

The effective approach treats AI as a drafting aid, not an author. You remain the expert on your charity's work. AI's job, if you use it at all, is to help organise evidence you have already collected into the structure the funder wants. Here is a practical workflow.

Step 1: Build your evidence base first. Before you touch any AI tool, make sure your outcome data is current. This means up-to-date attendance records, recent survey results, documented case studies and clear programme descriptions. Charities that maintain ongoing data collection spend significantly less time preparing grant applications than those that gather evidence retrospectively.

Step 2: Match your data to the funder's priorities. Read the funder's criteria carefully and identify which of your existing data points are most relevant. A youth employment funder wants employment outcomes and progression data. A community wellbeing funder wants beneficiary feedback and demographic reach. The same charity might use completely different data subsets for each.

Step 3: Provide verified evidence, not a blank prompt. If you use a general-purpose AI tool, give it only the source material relevant to the question — anonymised and aggregated, never raw beneficiary records — and keep a note of where every claim came from. The charity still writes and owns the response.

Step 4: Edit for voice, accuracy and nuance. AI does not know the political context of your local area, the personal relationships you have with partner organisations, or the subtle reasons your approach differs from others. These are the things you add in editing. A good first draft should need refinement, not rewriting.

Step 5: Verify every claim. Check that every statistic, case study reference and outcome figure in the draft matches your actual records. Drafting from evidence dramatically reduces fabrication risk, but human verification is non-negotiable. The Charity Commission's revised CC20 guidance on fundraising (refreshed in 2026) emphasises trustees' duty to protect the charity's reputation and ensure fundraising is carried out responsibly — which in practice means claims made to funders should be accurate and supported by evidence.

Which Parts of a Grant Application Can AI Handle Best?

Not all sections of a grant application benefit equally from AI. Understanding where AI adds value — and where it does not — helps you allocate your time effectively.

AI handles well:

  • Statement of need — Given local deprivation data, ONS statistics and your own needs assessments, AI can help shape them into a clear case. The Office for National Statistics makes extensive datasets freely accessible, so the source figures are easy to find and cite.
  • Programme description — If your programme model is documented, AI can help describe it clearly and consistently across multiple applications.
  • Evidence of impact — Once you have pulled the outcome figures, survey results and beneficiary numbers from your impact reports, AI can help weave them into the narrative without losing the specifics.
  • Case studies — Given anonymised case notes, AI can help condense them into funder-appropriate narratives while preserving the detail that makes stories compelling.

AI handles poorly:

  • Organisational strategy — Why your charity exists, what makes your approach distinctive, and how this application fits into your broader mission. These require human insight.
  • Partnership and relationship context — AI cannot explain the nuances of your local partnerships or stakeholder relationships.
  • Budget justification — While AI can format budgets, the reasoning behind specific cost allocations needs human input. Unrealistic or poorly justified budgets are a well-documented reason funders lose confidence in an application.
  • Sustainability planning — What happens after the funding ends requires strategic thinking that AI cannot replicate.

What Are the Risks of Using AI for Grant Applications?

Ignoring the risks would be dishonest. AI for grant writing carries real pitfalls, and charities need to navigate them deliberately.

Fabrication and hallucination. AI tools will invent statistics, fabricate quotes and create plausible-sounding evidence that does not exist. Research consistently shows that large language models produce factually inaccurate claims at significant rates when not grounded in source data. Giving the tool your real evidence reduces this significantly but does not eliminate it.

Homogenisation. When many charities use the same AI tools with similar prompts, applications start to sound identical. The National Lottery Community Fund has published guidance warning that AI tools "often produce generic content or include buzzwords that don't capture your unique perspective or your community's voice" (TNLCF, AI Guidance). Funders reviewing hundreds of applications notice when text follows the same structures and phrases.

Over-reliance. AI should support your grant writing, not replace your understanding of the funder relationship. Charities that delegate the entire process to AI — including reading the funder's guidance — tend to produce applications that technically answer every question but miss the funder's underlying priorities.

Data protection. Pasting beneficiary information into public AI tools like ChatGPT raises serious GDPR concerns. The Information Commissioner's Office (ICO) publishes detailed guidance on AI and data protection, making clear that data shared with AI processors must meet the same UK GDPR standards as any other data processing arrangement. Keep beneficiary records in a system with a proper data processing agreement — such as Plinth — and share only anonymised, aggregated figures with any writing tool.

Funders are not anti-AI. But a clear split is emerging between applications where AI has been used thoughtfully — with real data behind it — and those where someone has clearly just prompted ChatGPT and submitted the output. The latter group is not doing themselves any favours, and experienced assessors can tell the difference almost immediately.

How Much Time Does AI Actually Save on Grant Applications?

The time savings are real, but they depend almost entirely on your starting point. If you have no outcome data, no case studies and no documented programme model, AI cannot help you much — because it has nothing to work from. If you have a solid evidence base, the savings are substantial, and most of them come from not having to hunt for evidence at all.

Here are realistic time estimates for different approaches:

Application stageNo AI, evidence scatteredGeneric AI, evidence scatteredEvidence ready, AI as a drafting aid
Gathering evidence4-8 hours4-8 hours (no change)30-60 minutes (data already collected)
Writing first draft6-12 hours2-3 hours1-2 hours
Editing and refinement2-3 hours4-6 hours (more corrections needed)2-3 hours
Compliance check1-2 hours1-2 hours30-60 minutes
Total per application13-25 hours11-19 hours4-7 hours

The critical insight from this table is that generic AI barely saves time overall. You gain speed on the first draft but lose it in editing, because text written without evidence sounds right but needs extensive correction to be accurate. The biggest single saving is in the first row — and that has nothing to do with AI. It comes from having your evidence collected and organised before the deadline arrives.

For a charity submitting 15-20 grant applications per year — a common workload for a medium-sized organisation with an active trusts and foundations pipeline — the difference between 20 hours per application and 6 hours per application is the equivalent of reclaiming roughly 280 hours annually. That is more than seven full working weeks.

How Do You Build the Evidence Base for Grant Writing?

The charities that get the most from AI grant writing are not the ones with the best prompts. They are the ones with the best data. Building that foundation is not as daunting as it sounds, but it does require intentional effort — and it is the part of the job where software genuinely helps.

Start with outcome tracking. Define 3-5 core outcomes for each programme and measure them consistently. Use pre- and post-programme surveys where appropriate. Many charities still lack a documented outcomes framework, even though funders increasingly expect to see outcome data in applications.

Capture case studies as you go. The biggest barrier to strong applications is not having recent, relevant case studies when you need them. Rather than scrambling before each deadline, build case study collection into your regular workflow. Plinth's AI case notes let frontline staff record a short conversation with a beneficiary and turn it into a structured, anonymised case study that is stored and ready when an application needs one.

Keep programme information current. Your programme descriptions, staff qualifications, partnership agreements and safeguarding policies should be up to date and accessible. Whoever writes the application — and whatever tool they use — can only work from what they can find.

Centralise everything. The reason most charities struggle with grant applications is not lack of evidence — it is that evidence is scattered across spreadsheets, email inboxes, shared drives and individual staff members' memories. Keep outcomes, case studies and programme records together, then decide what is relevant before writing.

Can Small Charities Use AI for Grant Writing?

Absolutely — and in many ways, small charities have the most to gain. Organisations with annual income under £500,000 typically have one or two staff members responsible for all fundraising, alongside everything else they do. Small charities typically spend a disproportionate share of available staff time on grant applications compared to larger organisations with dedicated fundraising teams.

Used well, AI narrows the gap. A small charity with good data and a general-purpose AI tool can produce applications that hold their own against those from a larger organisation with a full-time grant writer. The key requirements are:

  • Some form of outcome tracking — even basic pre/post surveys or attendance records
  • At least 3-5 documented case studies — refreshed within the last 12 months
  • A clear programme description — what you do, who you serve, and what changes as a result

Plinth helps small charities build that foundation without complex setup: outcomes, case studies and impact reporting sit in one platform from the outset. The application itself is still written by the charity.

The Charity Commission's register lists around 170,000 registered charities in England and Wales, and the vast majority have annual incomes below £100,000. These are the organisations that can least afford to spend 20 hours on each grant application — and the ones that benefit most from getting that number down to five or six.

What Should Charities Look For If They Use an AI Writing Tool?

If you decide to use AI for drafting, you will be choosing between general-purpose assistants (such as ChatGPT, Claude or Copilot) and specialist grant writing products, some of which are essentially a general-purpose model with a grant-specific prompt template. Plinth is neither, so this is independent guidance. Here is what to evaluate:

Control over source material. Can you give the tool exactly the evidence you want it to use, and stop it from filling gaps with invented detail? A tool that encourages you to start from a blank prompt makes fabrication more likely, not less.

Funder-specific tailoring. Can the tool work to a specific funder's word limits and assessment criteria? Different funders want different things, and the output needs to answer the questions actually asked.

Traceability. Can you trace each statistic in the draft back to its source? If not, you will have to check every figure by hand anyway — which you should do regardless.

Data protection. Where is your data processed? Is there a GDPR-compliant data processing agreement? The ICO is clear that data shared with AI tools must be treated with the same rigour as any other data processing arrangement under UK GDPR. The safest default is to share only anonymised, aggregated information.

Cost. For small charities, cost is a deciding factor. Some AI grant tools charge per application — which can add up quickly if you are submitting 15-20 per year. Compare that cost with what a general-purpose assistant offers, and with the staff time still required for review, verification and rewriting.


Frequently Asked Questions

Can Plinth write my grant applications?

No. Plinth used to offer an AI grant writer, but no longer does, and it is not a grant writing tool today. It is where charities collect and organise the evidence applications rely on — case notes, outcome surveys, attendance and impact reports — so that figures and case studies are ready when a deadline arrives. The writing stays with you, in your own words or with a general-purpose AI tool of your choosing. If a funder runs its programme on Plinth, its application form may suggest answers from documents you upload, which you review and edit before submitting.

Will funders reject applications they think were written by AI?

Most major funders have not introduced explicit AI policies for applicants. The National Lottery Community Fund, for example, has stated that it will not reject applications simply because AI tools were used, but it warns that AI-generated text is "often not as strong as it might appear". Funders are increasingly able to recognise generic, AI-produced text — and those applications tend to score poorly because they lack specificity, not because they were flagged as AI-written.

Is it ethical to use AI for charity grant applications?

Yes, provided you use it responsibly. The key ethical requirements are accuracy (every claim must be verifiable), transparency (some funders now ask whether AI was used), and data protection (beneficiary data must be processed in compliance with GDPR). Using AI to draft from verified data is no different from using a template or asking a colleague to help write a section.

Can ChatGPT write a good grant application?

ChatGPT can produce a structurally competent draft, but without your specific data it will fill sections with generic claims that do not stand up to scrutiny. If you use ChatGPT, provide your actual outcome data, recent case studies and programme details as context — anonymised, with no personal data. Even then, expect significant editing and verify every claim before submission.

How do I avoid AI hallucinations in grant applications?

The single most effective measure is to give the AI your actual evidence rather than asking it to generate from nothing, and to tell it not to add facts that are not in the material you provided. Beyond that, always verify every statistic, case study and claim in the final draft against your source records. Never submit an AI draft without human review.

What data do I need before using AI for grant writing?

At a minimum: attendance or participation records for your programmes, 3-5 recent case studies or beneficiary stories, basic outcome data (even simple pre/post survey results), and a clear description of your programme model. The more structured and up-to-date your data, the easier it is for your team to write accurately.

Does AI grant writing work for statutory funding applications?

Yes, though statutory applications often have more rigid formats, specific compliance requirements and detailed financial reporting expectations. AI is particularly useful for the narrative sections of statutory applications, where you need to present evidence of need and impact within tight word limits. Budget sections and compliance declarations still require careful human input.

How much does AI grant writing cost?

Costs vary widely. Using ChatGPT directly costs around $20 per month for the Plus subscription. Specialist AI grant writing tools range from around £30 to £200 per month, and some charge per application. The return on investment depends on your application volume and how much staff review each draft requires.

Can AI help with grant reporting as well as applications?

Yes. In many ways, AI is even more valuable for grant reporting than for applications, because reports draw heavily on data that should already be collected — outputs, outcomes, financial summaries and case studies. This is where Plinth does help directly: its impact reporting feature generates funder-specific reports from your underlying data, so you maintain one dataset and produce multiple tailored reports rather than writing each one from scratch.



Last updated: October 2026