30 AI Disclosure and Research Integrity
This chapter is part of a book in active development and has not yet been through the author’s review. Content may change as the review advances.
The research decision. Decide exactly what you will say about every AI tool you used: which tool, on which task, and how you checked what came back. Then record it, so when someone asks “did you use AI, and how did you verify it?” you answer from a checkable record instead of from memory.
30.1 Why this decision matters
The decision on the table: which AI uses you disclose, and what you claim you did to check each one.
“I am not going to penalize you for using an AI tool. I am going to penalize you for not being able to tell me where you used it and how you checked it. A result I cannot trace back to a source and a verification is a result I cannot trust.” — a conference reviewer, standing at your poster and asking the AI-use question
At your research conference, one question a stranger can ask you is the AI-use question. Many journals and conferences now ask authors to state where and how AI tools were used, so this is not a formality you can wave off. If your only answer is “yes, a bit, and it seemed fine,” you have turned good work into an untrustable result on the spot. The disclosure is what keeps an honest use of AI honest under questioning.
30.2 The concept
AI disclosure is a truthful statement of where and how you used an AI tool in your work (International Committee of Medical Journal Editors 2026). Example: “I used Gemini to draft the code that computed my coverage ratio, and I recomputed that number by hand before reporting it.” Research integrity is the practice of keeping your work honest and traceable, so a reader can see where each result came from and trust it (ALLEA – All European Academies 2023). Example: a claim that names its source and its check has integrity; a claim that names neither does not.
The tool that makes disclosure easy is the one you have kept since you opened it at the start of your project: the AI Research Ledger, a running table with one row per AI use, recording the task, the tool, your decision, and the verification method. A complete disclosure is not written from scratch at the end. It is read straight off your ledger (Munafò et al. 2017).
Three parts make a disclosure complete by this book’s standard, and each maps to a ledger column: which tool you used, for which task, and how you verified the output. Drop any one and it fails. “I used AI” with no task is meaningless; a task with no verification is exactly the gap a reviewer is trained to find.
30.3 A worked example
You are presenting a business-analytics poster on whether a public company can comfortably pay the interest on its debt. Your headline number is its interest coverage ratio, the ratio of a company’s operating income to the interest it owes in a year. A coverage ratio of 2.0 means the company earned twice what it owed.
You used an AI tool twice: once to find the company’s reported operating income, once to write the short calculation that turns that income and its interest expense into a coverage ratio. Both are honest uses, and neither is safe until it is disclosed and verified.
Know where each part of that rule comes from. Current editorial guidance requires you to name which tool you used and for what purpose, and it keeps responsibility for the accuracy and attribution of the work with you; a tool can never be an author (International Committee of Medical Journal Editors 2026). The third column, how you verified the output, is this book’s own requirement on top of that guidance, because a disclosure that cannot say how something was checked leaves a reader nothing to trust.
Your ledger already holds two rows. For the reported figure, the verification method is primary-source reading: you opened the company’s published annual report and confirmed the operating income the tool gave you. For the calculation, the method is alternative code, recomputing the same number a second way:
# The number the AI code reported, recomputed by hand.
operating_income = 276 # USD millions, confirmed against the annual report
interest_expense = 131 # USD millions, from the same annual report
coverage_ratio = operating_income / interest_expense
print(round(coverage_ratio, 2)) # 2.11 — matches the AI-generated cellWhen the reviewer asks the AI-use question, you do not improvise. You read your two rows aloud: the tools, the tasks, and the two checks. The claim you defend is bounded. The coverage ratio is about 2.1 for this fiscal year, with the income figure confirmed against the annual report and the arithmetic reproduced by hand. Its boundary is honest too: you read a single year of reported results, not a downturn scenario and not the debt coming due next year.
30.4 An AI failure case
You ask a tool to “write my AI-use disclosure from these notes,” and it hands back a polished paragraph: “All AI-generated outputs were independently verified against primary sources.” It reads beautifully, and it is false. You verified the operating income against the annual report, but you never independently checked the interest figure the same tool helped you pull. The sentence upgraded one real check into a blanket claim of “all.”
You catch it by reading the disclosure line by line against your ledger. Every sentence claiming a verification must point to a row that names that check. The word “all” has no such row, so it is fabricated and comes out. Your true disclosure is narrower and stronger: you name the two checks you ran and the one input you did not independently confirm. A confident paragraph is not a truthful one. You verify the disclosure against the record, not against how responsible it sounds.
30.5 It is your turn
You are working inside Studio 9: Write, bound, and disclose. Keep what you write here; the studio’s milestone chapter is where it joins the other lessons’ pieces into one artifact you can defend.
Your claim-evidence table stands behind every claim; this step writes the one about AI, and backs it with a record.
The hands-on half of this section lives in the chapter’s companion notebook: open it in Colab with the badge at the top, and work the steps there.
Commit your own answer first, then delegate. Each prompt is a checkable job, not a request for a verdict.
Run the loop, and log it. Each round is a prompt, an output, an objection, and a rerun, and the ledger row records what you finally accepted and why. That matters more here than anywhere else in the book. If an agentic tool ran ten invisible rounds to produce your paragraph, your disclosure still has to describe only what you can personally vouch for.
The honesty of the disclosure is yours alone. You decide which uses to disclose (including the one you would rather not mention), whether a verification you claim actually happened, and whether the final statement is true. A tool can format your ledger into a clean paragraph. It cannot know what you really did or certify that you checked it. Signing your name to a disclosure you cannot back is the integrity failure this chapter exists to prevent.
Read your AI Research Ledger end to end and list every use, including the one you would rather not mention.
For each use, write the three parts a disclosure needs: which tool, for which task, and how you verified the output. A use with no check stays blank. Do not fill it in from memory.
Locate the standard.
Act as a research-integrity assistant. Name the elements a complete AI-use disclosure should contain for a research poster, and cite one real, retrievable source such as a university research-integrity office or a major publisher's AI policy. Only name a source you are confident exists.After running, verify: open the cited policy and confirm each named element is actually in it. Counters confident fabrication (an invented policy element arrives as confidently as a real one).
A second angle, optional:
Turn notes into a disclosure table to verify.
Here are my raw notes on every time I used an AI tool on this project: [paste]. Build a table with columns tool, task, and verification method, one row per use. Do not invent a verification I did not mention. If a use has no check, leave that cell blank and flag the row.After running, verify: match each row against your ledger; a verification filled in where your notes had none is fabricated, so cut it. Counters illusion of completeness (a tidy table that quietly reports checks you never ran).
Turn the rows into a short statement in your own words, then check every sentence against a row. Any sentence claiming a check you cannot point to comes out.
Red-team your disclosure.
Here is my drafted AI-use disclosure: [paste]. Act as a hostile reviewer who suspects I am hiding a use or overstating my checks. Name every place I claim a verification I have not shown, or omit a use a reviewer would expect. Do not rewrite it for me.After running, verify: for each flagged line, produce the ledger row that backs it or delete the claim. Counters sycophantic agreement (a reviewer that praises your honesty instead of testing it).
Say what you did not verify. Naming the one AI-assisted input you never independently confirmed is what makes the rest of the statement believable.
Log it in your AI Research Ledger, and verify the finished statement with a named method from the Verification Guide, such as peer reasoning. An AI reviewer may run the check with you; the decision to accept or reject stays yours.