# Why the student writes

> AI can carry an application; it must not write it. The rule at the center of Admit Protocol, why it is enforced by the server rather than requested on an honor system, and what it means for families and schools.

- Published: 2026-11-17
- Author: Michael Janzen
- Audience: parents, students, schools
- Canonical: https://admitprotocol.com/articles/why-the-student-writes
- Publisher: Admit Protocol, https://admitprotocol.com

Every school that asks an applicant to write something is asking the same question underneath: who are you, when nobody is helping? AI tools have made that question harder to answer and more important to ask. Admit Protocol's answer is a rule, and the rule is enforced by machines so that people do not have to argue about it.

## The rule

The agent facilitates the application. The student writes. The AI does not write for the student.

In practice: an agent may read the student's files, ask the student questions, transcribe something written by hand, and send what the student wrote exactly as it is, typos included. It may not draft, rewrite, polish, suggest wording for, or "improve" the student's words, even when the student or a parent asks it to. If the student wants help, the help is a question ("what do you mean here?"), never a sentence.

## How it is enforced

Every piece of writing sent through the protocol carries an attestation: the student wrote it, no agent edited it, and a named parent or guardian confirmed that before sending. A school's server refuses writing that arrives without the attestation. An agent that wants to send writing has to send it under that statement, and an agent that lied in it would be putting the family, not the software, at risk of a false statement on an application.

This does not make cheating impossible. Nothing does. A student could write with AI help in another window and paste the result into the folder, the same way a student could have a tutor write the essay in 1995. What the rule does is remove the easy, deniable path. The tool that is carrying the application will not do the writing, and the family has signed a statement that nobody else did either. That is the same standard schools have always applied to a parent's help, made explicit.

## Why not just ask nicely?

Because an honor system that depends on the AI tool's own good behavior is not a standard, it is a hope. Different assistants have different defaults, and most of them are built to be helpful in exactly the way that is harmful here: offered a rough paragraph, they tidy it. The protocol has to say, in the instructions every agent reads first and in the server that receives the result, that tidying is forbidden. Agents follow instructions; this one is written where they cannot miss it.

## What schools get from it

Something they have been losing: a sample of how the student actually thinks and writes, at the level they actually write at. Admissions readers at independent schools are not grading essays. They are trying to tell whether a student is curious, whether they notice things, whether they can be honest about a failure, and whether they will be good to live with. Polish hides all of that. A rough, real paragraph shows it.

Schools on the protocol are asked to read accordingly, and to tell families so. The schools' page says it plainly: read for thinking, interest and honesty, not for quality. A school that wants beautiful prose should keep its old form. A school that wants to know who is applying should try this.

## What families get from it

A clear rule, which is a relief. Most families do not want to cheat. They want to know what is allowed, and the answer in most admissions processes is a shrug. Here the answer is written down: the student writes, a parent confirms it, and the tool will not cross the line for you. Parents who were going to "just help with the wording" are spared the temptation, and students who were going to do it right are no longer competing against students who did not.

It also puts the student in charge in a way that matters. A student can run the whole application themselves, from the first conversation with the assistant to the final read-back. The only step that needs a parent is the approval code at the end. The writing is theirs because the process is theirs.

## What the rule does not say

It does not say AI is bad, or that students should not use it. The protocol exists because AI agents are good at the part of applying that was never meant to measure anything: finding the documents, filling in the fields, keeping track of what each school wants. Let the machine do that. Keep the one part that was always a question about the student.

It does not say the writing has to be good. Word counts are guidance. Spelling is not graded. A student who writes two hundred plain words about a podcast they liked has done the assignment.

It does not say schools may not use AI for anything. It says the protocol has no place for AI judgment of applicants, which is a different thing, and a line we have drawn on purpose.

## For other schools and other tools

The rule is part of the open specification, not a policy of one school. Any school that implements Admit Protocol inherits it, and any agent that wants to apply through it has to follow it. If you build admissions software and you are deciding what to do about AI and essays, you are welcome to copy this approach wholesale: write the rule where the agent reads first, require the attestation, refuse what lacks it, and tell families and readers exactly what you are doing. It costs nothing and it answers the question everyone is asking.
