Funders Say They Can Tell When You Use AI
“When it comes to AI, we can tell who’s using it and who’s not.”
That comment came from a funder during a roundtable discussion I recently attended. Other speakers raised concerns about AI removing important information from narratives or producing responses that didn’t adequately tell an organization’s story. Nobody called for a blanket prohibition on using it, but the overall sentiment about applicants’ use of these tools was fairly skeptical.
If you’re a nonprofit leader exploring how AI can support your grant writing, that’s worth knowing. Funders are discussing tools such as Claude and ChatGPT, which use large language models (LLMs), and some have strong opinions about their use in applications.
There’s more to this conversation than whether or not funders approve of their applicants using AI to help with their applications. These tools can be incredibly useful in grant development, funders’ expectations and opinions vary, and research on people’s ability to recognize AI-written text offers a more complicated picture than that confident statement suggests.
AI Can Be Useful in Grant Writing
I understand why so many nonprofit teams are interested in AI. When you’re responsible for programs, staff, fundraising, and whatever has just landed in your inbox, help getting an application moving can be welcome.
Used thoughtfully, AI tools can help organize approved program information, suggest an outline, identify repetition, or offer a clearer way to explain a complicated process. It can help a writer explore different approaches to a response or work toward a character limit. Those are useful contributions.
The Chronicle of Philanthropy has reported on grant professionals using AI for prospect research, letters of inquiry, and proposal development. One practitioner, Josie White of Shelter the Homeless, described the preparation behind her use of the tools for the publication: “I provide the data I’ve already verified, researched, and know I want to use.” She also described having program and executive leadership approve that information.
That preparation matters. Someone has established which information is accurate, appropriate, and relevant before asking the tool to work with it. Someone also needs to assess what comes back.
AI tools can make parts of that process faster. The grant professional’s expertise helps determine whether the result is useful and effective.
Funders Have Different Expectations About AI
The views expressed in one discussion don’t represent every funder. In Candid’s 2024 research on AI-generated grant proposals, foundation respondents expressed a range of opinions. Some welcomed the possibility of reducing applicants’ workload or improving access. Others opposed accepting AI-generated applications, and many had not decided. One respondent put it plainly: “We are agnostic to the tools that applicants use to create grant applications.” These findings came from a convenience sample of U.S. foundations, so they provide a dated snapshot of respondents’ views.
More broadly, the Center for Effective Philanthropy’s 2025 report, AI With Purpose, found that AI use and concern about AI coexist across foundations and nonprofits. Respondents reported using the technology while expressing concerns about accuracy, security, staff expertise, and bias.
That combination makes sense to me. A tool can be useful and require careful judgment. The difficulty for applicants comes when expectations are unclear, or when permission to use AI is accompanied by comments suggesting that doing so reflects poorly on their work.
Funders can help by explaining their policies and the standards they expect applications to meet. Nonprofits deserve enough clarity to make informed decisions about their process.
What Research Says About Detecting AI Writing
The assertion that someone can alwas recognize AI-written content also deserves a closer look. In a 2025 study published in Advances in Simulation, five healthcare simulation educators and researchers assessed article introductions. They were asked to distinguish among five categories, ranging from entirely human-written text to entirely AI-generated text, with different degrees of AI involvement in between. Their overall accuracy was 19 percent, roughly what chance would produce on that task. A different 2025 study found stronger discernment among people who frequently used AI for writing. The researchers tested judgments about 300 nonfiction articles. The majority vote among five frequent AI-writing users misclassified only one article.
Those studies examined different texts, participants, and tasks. Neither directly establishes how accurately a program officer can identify AI use in a nonprofit grant application. We should be careful about treating either as a universal answer. That distinction matters because a reviewer can correctly identify a weak response while being mistaken about how it was written. An experienced reader may also recognize patterns associated with AI-generated text. Confidence alone doesn’t tell us which situation we’re dealing with.
For applicants, the practical concern is that a reviewer’s impression may influence how they receive the application. For funders, it’s a reason to keep the evaluation grounded in the application’s substance and stated requirements.
What Funders Are Concerned About
Some of the concerns raised in the discussion I attended were quite specific.
One speaker described an applicant using AI to shorten a narrative, only to find that important information had disappeared. The response needed material added back in because the shortened version had lost part of the explanation.
Anyone who has worked against a stubborn character limit can appreciate how that happens. A response can get shorter and read smoothly while becoming less useful to the reviewer.
Imagine, for example, that a program description explains how evening hours allow parents working later shifts to participate. A revision that replaces those details with “accessible, family-centered services” may sound polished, but it has removed the explanation of what makes the service accessible.
That’s a substantive editing problem. The writer needs to recognize which details establish the case and protect them during revision.
The same applies when a response describes an important community issue without explaining the organization’s particular role, or makes a claim about results without identifying the evidence behind it. Those gaps deserve attention however the text was produced.
As grant writers, we need to recognize these gaps before we submit an application. Whether we use AI to help us draft our responses or write every word ourselves, we should be asking what’s missing, which claims need support, and where someone unfamiliar with the organization might need more explanation. That review gives us a chance to strengthen the application before a funder ever sees it.
Why Qualified Grant Expertise Matters
For a nonprofit leader, this conversation raises a larger question about responsibility for the application.
Someone needs to understand the funding opportunity well enough to interpret its requirements. Someone needs to know which staff members can confirm the program information, whether the figures describe the same reporting period, and how the proposed activities connect with the request.
Those responsibilities remain when an LLM helps produce the draft.
Consider a response claiming that an organization served 200 young people. A qualified grant professional may ask whether that figure represents unique participants, program enrollments, or service encounters. They may discover that the budget covers one program while the narrative describes several. Resolving those questions requires access to the right information and judgment about what the application needs to communicate.
An AI tool can assist with identifying inconsistencies or organizing the answers. A professional still has to investigate, verify, and decide how the findings affect the proposal.
That expertise may sit with an employee or an outside consultant. What matters is that the person responsible can assess the work throughout development. Asking someone to proofread a completed draft is a much narrower assignment.
This is where I see the value of qualified grant professionals using AI well. They can draw on its useful capabilities while remaining accountable for the reasoning and evidence in the application. The organization benefits from the tool and from someone who knows how to evaluate its contribution.
Put Grant Expertise Behind Your Application
If AI is becoming part of your grant process, it’s worth looking at who owns the decisions surrounding its use and who checks the substance before submission. AI can help you write a grant - an experienced grant professional can help you win one.
Through Carpenter Nonprofit Consulting’s grant writing and strategy services, I help nonprofits assess funding opportunities, develop their case for support, and coordinate the information needed for credible applications. That work gives your team experienced guidance from the early decision to pursue an opportunity through proposal development. If your team could use that kind of support, get in touch and tell me what you’re working toward or set up a brief, no-pressure consultation with me. I’d love to talk about your funding goals and whether working together would be a good fit.
Explore This Topic Further
