AI can write the words. It cannot write you.
There is a shortcut everyone is taking. You have probably taken it too. You open ChatGPT or Claude, paste the call for proposals, and type: write me a grant application for this. What comes back looks like a grant. It has the right structure. The sentences are clean. It mentions your discipline. It is completely hollow.
The reviewers know. They have read thousands of applications. They can feel when no one is home.
But the problem is not the tool. The problem starts earlier, before you open any application, before you touch any software. The problem is that most artists skip the hardest step.
The step no one wants to do.
Before you write a single word of a grant, you need to sit with your work and ask yourself some uncomfortable questions.
Why are you making this? Not the elevator pitch version. The real answer. The one that takes a few minutes of silence to arrive at.
What does this work mean to you? What is it trying to hold, or resist, or repair? Where did it come from — not the technical origin, but the emotional and intellectual one? What poem, what image, what piece of music, what conversation, what loss cracked something open in you and made this work necessary?
These are not rhetorical questions. They are the substance of your application. A grant panel is not just evaluating a project. They are evaluating whether an artist understands what they are doing and why it matters. That understanding cannot be faked. And it absolutely cannot be generated.
A machine has no access to what Édouard Glissant meant to you the first time you read him. It does not know which film stopped you in your tracks at twenty-three. It does not know what you noticed in the archive that nobody else had noticed, or why that moment felt like a door opening. It cannot know your grandmother’s hands, or the particular quality of light in the place you are trying to document, or the conversation with a colleague that reoriented your entire practice.
That subjectivity is your voice. And your voice is what makes a grant application land.
What LLMs are actually doing?
When you hand a language model your grant brief and ask it to write the application, here is what it produces: a statistical average of every grant application it has ever encountered. It gives you the center of the bell curve. The most expected sentences. The most common structure. The safest, most unremarkable version of what a grant could sound like.
Grant panels sit on the other side of that bell curve all day. They are not looking for the average. They are looking for the artist who sees something specific, who has done the thinking, who has something at stake.
Generic prose does not signal competence. It signals absence.
The reflection practice
Before you begin any application, give yourself thirty to sixty minutes alone with these questions. Write the answers by hand if you can. Do not edit. Do not perform.
What is the core question this work is asking? Not what it is about. What it is asking.
Who made work that opened a door for you? A filmmaker, a poet, a weaver, a theorist. Name them. Describe what they did and why it mattered to your practice. The lineage you come from is part of your argument.
What is the specific thing only you can make? Not because you are the most talented, but because of your particular position, history, and set of obsessions. What is yours to make?
What does success look like for this project — not for your career, but for the work itself? Who does it reach? What does it change or hold or refuse?
When you can answer these questions in your own words, without borrowed language, without performing expertise, you have something real. That is what you bring to the application. That is what no model can generate on your behalf.
A balanced approach
This is not an argument against using tools. I use them. They are useful for structure, for editing, for checking that your argument is legible, for translating a thought you have already formed into cleaner prose. That is a legitimate use.
The mistake is using them as a starting point instead of an endpoint. If the thinking has not happened first, the tool has nothing real to work with. It will fill the space with noise that sounds like signal.
Start with yourself. Do the reflection. Write the rough version in your own words, even if it is clumsy. Then, if you want to use a tool to tighten the language or check the structure, you can. But the thinking, the meaning, the specificity — that has to come from you first.
A few prompts to get you started
Once you have done the reflection work, here are prompts you can bring to an LLM; not to write for you, but to help you develop what you have already found.
“Here is a rough paragraph about why I am making this work. Help me make it clearer without changing the voice or removing the specificity.” [Past your text here]
“Here is my artist statement. What is the central argument? Where does the logic feel weak?” How can I improve the document submitted? [Past your text here]
“I have written a draft of my project description. The grant requires me to write about […fill the gap with the grant section]. Help me decipher what this section is about and help me revise my document.” [Past your original and personal text here]
This prompt is the most useful↓
“I am applying for this grant [weblink/URL]. The specific criteria for evaluating the application are [Paste all available criteria or paste the entire call here]. Based on the criteria, write a critical report about the documents I want to submit such as CV, portfolio, project description [upload your documents]. Tell me what I missed. Tell me how to improve. Tell me about my blind spots. Be a critical friend. I am expecting to significantly transform my application from a basic level to an advanced winning application.”
Notice what these prompts have in common. They start from your material. They ask the tool to serve your thinking, not replace it.
There are more advanced techniques to yield superior outputs from LLMs, but this requires an entire separate conversation and specific methods that I taught to my successful students at University within a research context (Course name: “Advanced AI methods in Media production”).
