"What do I say when I hand the panel each piece?"
Don't hand over raw files and hope. Each exhibit has one sentence that makes it land — and one moment to hold it back. Here's the sentence, why it works, and the catch, for everything worth raising.
| Your exhibit | Say this — close to verbatim | Why this sentence lands | The catch / when to hold it |
|---|---|---|---|
| Drafts, outlines, notes | "Here's the messy path — false starts, a thesis that changed halfway, paragraphs I cut. This is what writing actually looks like." | A finished paper looks the same whoever wrote it. A real draft can't be reverse-engineered — faking the middle means knowing an ending you hadn't reached yet. The mess is the part a panel can't explain away. | Incompleteness helps — don't tidy it. Present it as a sequence ("outline → draft one where the argument was different → the section I cut"), not a pile. |
| Version history (Google Docs / Word) | "The edits track iterative work — research, revision, growth — in proportion to the assignment." | It shows the document growing over time, on a system you didn't run, in a shape that matches the work that was asked for. | Thin? Don't raise it. A near-empty history reads against you. And tools now exist that fake a natural-looking history — which is exactly why a history can't carry your case alone. Lead with drafts and outside records; raise history only if it's genuinely thick. |
| LMS / email / cloud timestamps | "An outside system recorded when I worked — I didn't make these." | It's the one exhibit you couldn't have authored. So it survives the "self-serving" objection that sinks everything you made yourself. Name the system out loud — "Canvas logged every save" — not just the date. | Your strongest single exhibit. Can't reach the log? Ask the school to pull its own — that puts the checking on the side that made the accusation. |
| Your prior writing (baseline) | "This paper reads like everything else I've written. Here's my baseline — same register, same tics, same level." | A panel's first instinct is to compare the work against your earlier writing. So a consistent voice is a defense that predates the accusation. If it reads like last term's papers, "this isn't your voice" collapses. | Decisive if you're a non-native speaker or neurodivergent and "write formally" — the pattern the tool reacted to is your baseline, not a tell → Accused while writing in a second language → |
| The same checker on your old, pre-AI work | "My human writing from before these tools existed also comes back 'AI.' The tool is flagging my style, not my honesty." | A pre-AI essay can't have been machine-written. A high reading on it shows the instrument misreads you specifically — it attacks the foundation, not just this paper. | Strong support, not a standalone. Pair it with the record above — don't stand on it alone → Why a checker's number isn't proof → |
| The policy timeline | "There was no AI rule when I submitted this." | A clean procedural point that asks the panel to believe nothing about your writing. Either the rule postdates the work, or it doesn't. | Verify the dates against your handbook before you say it. A wrong date hands them a reason to doubt the rest. |
Two rules sit under all six. Name the independent system out loud. Present process as a sequence, not a pile. And lead with the exhibit you didn't make. Which exhibit actually holds up, and the forensic mechanics behind each → What evidence actually stands →
"Which exhibit goes first?"
Open with what an outside system recorded. Build with the record only you made. Never volunteer a thin one.
the exhibit you COULDN'T have authored
- LMS / email / cloud timestamps ("I didn't make these")
- the same checker flagging your OWN pre-AI writing
the record only you made, but can't fake
- genuine drafts — the messy path, shown as a sequence
- your prior-writing baseline, consistent over time
the procedural point that needs no belief
- the policy timeline ("no AI rule when I submitted")
raise ONLY if thick — never volunteer thin
- version history (auto-typers fake it; lead elsewhere)
"What stands in the room, and what gets waved away?"
Outside-system records stand. A denial alone doesn't. Here's the full split, and the one line that decides each.
| Exhibit | In the room | Why it cuts that way |
|---|---|---|
| Outside-system timestamps (LMS, email, cloud) | ✅ Stands | You couldn't have authored your school's own log — it survives the "self-serving" objection. |
| Genuine drafts / outlines / the messy path | ✅ Stands | Real struggle is hard to fake. The false starts are the part a panel can't explain away. |
| Your prior-writing baseline | ✅ Stands | It predates the accusation — so it can't have been built for it. |
| The same checker on your own pre-AI work | △ Stands as support — not alone | It discredits the instrument on your writing. Pair it with the record; don't stand on it. |
| Substantial version history | △ Stands if thick; backfires if thin | A full, proportionate history helps. A near-empty one argues their case for them. |
| A tool's number — for or against you | ✗ Not proof on its own | The same paper reads differently across tools. A number is a prompt to look, not a finding → Why a checker's number isn't proof → |
| A polished, after-the-fact "recreation" | ✗ Reads as staged | Anything built after the accusation looks built for the accusation. |
| A confident "I didn't do it," alone | ✗ Not evidence | An assertion is the thing evidence has to support — not a substitute for it. |
| A manufactured or "cleaned-up" history | ✗ Makes it worse | Concealment after the fact is the corroboration the school was missing — a separate, heavier charge → What evidence actually stands → |
"What do I say when they push back?"
The hardest moment isn't presenting. It's the follow-up question built to rattle you. Four you should expect, and the line that meets each. The pattern never changes: don't defend your character — point back to the record.
| If they say | You say | Why it holds |
|---|---|---|
| "Couldn't you have written all this after the fact?" | "The timestamps are from an outside system, recorded as I worked — I couldn't have made them later. They're checkable against the school's own logs." | Moves it off your word and onto a record you didn't author — the one thing an after-the-fact story can't fake. |
| "We've seen students manufacture this kind of trail." | "Then check mine against the independent record — the LMS log, the email timestamps. A manufactured trail wouldn't match an outside system it can't control." | Concedes the general worry, then invites the exact test that separates a real record from a fake one. Offer the verification; don't get defensive. |
| "This just doesn't sound like your other work." | "Here's my prior writing — same register, same patterns, across courses and from before this assignment. If it reads like everything else I've written, it's my voice." | Answers a subjective impression with a documented baseline that predates the accusation — so it can't have been built for it. |
| "The tool's number is very high, though." | "Its makers say a number shouldn't be the sole basis for a finding, and the same paper reads differently across tools. I'm not asking you to take my word over it — I'm showing you the record it can't account for." | Doesn't argue accuracy in the abstract. Reframes from the number to the record — where you're strongest → Why a checker's number isn't proof → |
The rule under all four: a pushback question is an invitation to point at the record, not to plead. The moment you defend who you are, instead of what the record shows, you've stepped onto the ground where you're weakest.
"What exactly do I bring?"
Six things — and each earns its place for a reason. Here's the why on every one.
- Your writing record — drafts, version history, edit timestamps, notes, saved research. Export it to your own device before the day.Why: it's the only evidence that shows process. And the systems holding it can change your access without warning — once it's on your device, no one can quietly close that door.
- The accusation in writing — assignment, the tool, the number, the exact policy section cited.Why: you can't rebut what you can't see precisely. Pinning the exact charge stops it shifting mid-hearing, and tells you which exhibit answers it.
- Your handbook — integrity policy + hearing procedure, every promised step marked.Why: a skipped step is one of the few things an appeal will actually hear. Mark the procedure to hold them to it → What happens at the hearing →
- A timeline — when you wrote, when you submitted, and when the rule took effect.Why: it turns scattered exhibits into one story a panel can follow. It also surfaces the policy-timeline point, if the rule postdates your work.
- A one-page statement — calm, factual, about the record. Drafted, not sent without advice.Why: anything you put in writing becomes part of the record and can be turned around. One page keeps you to the facts, and off the defensive over-explaining.
- An advisor — ask before the day who is permitted to attend.Why: who may speak, and when an attorney is worth it, varies by school and by what's at stake. You want that answer before the room, not in it → What happens at the hearing →
"I already said something I regret — is it too late?"
No. A panicked email is not a finding — so stop adding to it. One panicked message isn't the case against you. Shift to the record you can document, put nothing else in writing without advice, and move the conversation back to the process evidence above. How to recover from an early statement → If you already confessed →
"Will preparing guarantee a win?"
No — and anyone promising that is selling something. Your leverage is the record and the procedure, not a certainty. When a tool's flag stood alone, students have won. When the school had something independent, it usually held. The split, case by case → The cases — coming soon
"What's the one thing I can't put together the night before?"
The writing record at the top of every line above. It only counts if it already existed before the question. That's the one thing you can't reconstruct under pressure.
Read back up the table. The exhibits that stand are the ones an outside system recorded, or that already existed before the accusation. The night before a hearing, you can gather them, order them, and rehearse the sentence for each. But you cannot create them. The strongest position isn't assembled once you're accused. It's a continuous, contemporaneous record of how the work came together — already there before anyone asks.
Already accused?
Build your evidence by hand, today:
Your best evidence already exists →Another assignment due soon?
Set up a record before anyone asks:
Prove it before anyone asks →Where to next? → What evidence actually stands → · The cases — coming soon · Why a checker's number isn't proof → · What happens at the hearing →
Information, not legal advice — read your handbook and talk to a qualified advisor. We don't judge; we help you track the records.