The disclosure

Ben is 18, in first year environmental science, and the biggest assessment of his semester is a research report worth 40% of the unit. He uses AI a lot, and none of it is cheating. He asks a chatbot to suggest three possible structures and picks the one that fits his data. He pastes in two of his own drafts and asks for the harshest critique it can give. And when he is unsure whether his data needs one statistical test or another, he talks his reasoning through with the tool until he understands why the answer is what it is. The data, the analysis, the decisions and every final sentence are his.

The unit guide is clear: AI tools may be used, and any use must be declared on the cover sheet. The night before submission, his friend Priya sees him typing the declaration and winces. Do not hand them that, she says. Markers see an AI declaration and mark you down without even noticing they are doing it. Nobody in this unit declares. You will be the only one.

Ben nearly deletes the note. Priya is not being dishonest, she is describing a fear plenty of students share, and he cannot prove the fear is wrong. So he rereads the unit guide instead, and notices the actual shape of the bet. Declared AI use is explicitly permitted, so the worst case of declaring is a sceptical marker. Undeclared use is listed as academic misconduct, so the worst case of not declaring is a misconduct finding, and that bet has to keep winning forever, every time anyone asks him anything about the report. One risk is small and ends at marking. The other is large and never expires.

He keeps the declaration and writes it properly: I used an AI chatbot in three ways. It suggested candidate structures for this report, and I selected and reworked one. It critiqued two drafts, and I made my own revisions in response. I used it to check my reasoning when choosing the statistical test in section 4, and the final choice and analysis are my own. All data collection, analysis and writing are mine, and I verified every factual claim against the cited sources. Four sentences. It takes him six minutes.

The response is an anticlimax, which is exactly the point. His tutor writes one line about the declaration, calling it clear and specific, and marks the report on its merits: strong method, slightly padded introduction, good statistics. In the following week's lab she asks, casually, why he chose that test. Because he had argued about it with the chatbot until he actually understood it, he answers in thirty seconds. The conversation moves on. Nothing happened, and nothing happening was the win.

His classmate Dom used AI in a similar way on a similar report and declared nothing. No detector flagged him. What caught him was ordinary bad luck: the unit randomly selects a handful of students each semester for a short interview about their methods, and Dom's number came up. Asked why he chose his statistical test, Dom could not say. He had pasted the tool's justification into his report without ever understanding it, and the one question he faced was the one part of the report that was never really his. On its own, that might have read as weak understanding. Combined with an undeclared tool and a signed cover sheet saying no AI was used, it became a misconduct case, and the process that followed cost him far more than any marker's scepticism could have.

The gap between Ben and Dom was never virtue. It was strategy. Ben's declaration cost him six minutes and changed how he worked: knowing he would describe his AI use, he only used the tool in ways he could stand behind, which is why the lab question was easy. Dom's silence felt free right up until a routine conversation turned it into evidence. Priya's advice assumed honesty was expensive and concealment was free. The real prices were the other way around.

There is a coda worth adding, because Priya was not wrong about everything. Marker scepticism is real, some students do get unlucky, and her advice came from wanting to protect a friend. What she had wrong was the accounting. She compared the certain small cost of declaring against a concealment cost she priced at zero, because she had never once seen that cost charged. Dom's semester repriced it for everyone who knew him, and declarations in that unit stopped feeling like a risk and started feeling like the cheap insurance they always were.

For Ben the habit outlived the assignment. When he applied for a summer research placement the following year, the supervisor asked in the interview how he used AI in his work, a question that is quickly becoming standard, and he answered the way he had written the declaration: the specific uses, the boundary, the checks. It took him thirty seconds and it read as competence rather than confession, because that is what disclosure becomes once it stops being about a single cover sheet. It is simply being able to give an account of your own work, and people who can do that get trusted with more of it.