Using AI Tools Ethically
AI assistants are now a common research tool. Used well, they save time. Used carelessly, they invent citations, put unchecked statistics into your work, and take over thinking that you are responsible for. For any task you have three choices: go ahead, use the tool but verify what it gives you, or leave it out. This lesson explains how to choose, then gives you eight everyday situations to practice on.
Where AI helps
Plenty of legitimate uses have nothing to do with faking results. AI can explain a cryptic error message or suggest why your code won't run. It can argue against your interpretation, help you brainstorm explanations for a surprising finding, or tidy the grammar and flow of prose you wrote yourself. In each case you stay in charge: you decide whether an idea is any good, you run and check the code, and you keep your own voice. The tool is fast and often useful, but it is sometimes wrong, and your name is on the work.
Where it fails: fluency is not truth
A language model generates plausible text. On its own it does not check facts, look anything up, or calculate. That one property explains its three most dangerous failures.
- Fabricated citations. Ask for references and it will produce authoritative-looking ones (real journal, plausible authors, convincing title) for papers that do not exist. This happens often, because of the way the tool generates text.
- Unverified statistics. Ask it to "compute" a p-value or a mean and it may output a confident number that is wrong. Unless it actually ran the calculation, the number is a prediction of what the answer would look like, and it is formatted the same way in both cases.
- Hallucinated facts. It states made-up historical "facts," study findings, and definitions as fluently as real ones, with nothing to mark the difference.
In short, fluency is not evidence. The model is built to produce confident, well-written text whether or not the content is true. Many assistants can now search the web or run code when they decide to. That makes the problem smaller but does not remove it. You usually cannot tell from the answer whether the tool searched or only sounded confident. A source it found can still be summarized wrongly, and code it ran is only as good as the inputs you gave it. Assume nothing has been checked until you have checked it.
You own every word and number
Whatever help you use, the checking stays with you. If a sentence, statistic, or citation appears in your work, you are vouching for it as if you had produced it yourself. "The AI told me" is not a defense to a supervisor, an examiner, or a journal. Every number must be reproducible from your data in statistics software. Every citation must point to a source you have read, and every factual claim must be checkable. AI can get you to a draft faster, but you are still responsible for what the draft says.
The data-protection trap. Never paste participant data (transcripts, survey responses, anything identifiable) into an external AI tool. Doing so discloses your participants' data to a third party they never consented to, which breaches your data-protection duties and probably your ethics approval. Removing the names is not enough, because personal stories and quasi-identifiers can still identify people (see Privacy & Confidentiality and Data Privacy Basics). Keep participant data inside approved, secure systems only.
Disclosure, and why AI can't be an author
Whether you must declare AI use depends on your institution and your target journal. Their policies differ and are still changing, so check the ones that apply to you and declare your use when they ask for it. One rule is the same everywhere: an AI tool cannot be an author. Authorship requires accountability, the ability to stand behind the work and answer for it, and software cannot take that responsibility. This is the same accountability criterion as in Plagiarism, Authorship & Citation. If AI helped substantially, describe how in a short note in your methods or acknowledgements. It does not get a byline.
Run the decisions
Below are eight everyday tasks. For each one, decide whether it is a go (fine as described), a verify / disclose (usable, but only with a specific safeguard), or off-limits. Your choice reveals the reasoning.
🤖 Can I use AI for this?
Sort each task: go ahead, use it but verify, or off-limits.
Same tool, three verdicts. The verdict depends on what the task involves. If the tool helps with your thinking or writing and you still own and check the result, it is usually fine. If it supplies facts, numbers, or citations that you would otherwise pass on unchecked, verify them first. If the task involves participant data, the reproducibility of your results, or authorship, it is off-limits, because those are duties you cannot hand to a machine.
Why it matters: AI causes trouble when it lets you skip the parts of research that are your responsibility: reading your sources, reproducing your statistics, protecting your participants, and standing behind your words. If you check everything it produces and disclose your use as your policy requires, you get the time savings without neglecting any of those duties.
Common questions
Is it unfair to use AI to improve my English if I am not a native speaker?
Academic integrity rules are not aimed at using a tool to say your own ideas more clearly, so this is not something to worry about. The line is the same as for a proofreader or a writing center: the ideas, the analysis and the responsibility for every claim stay yours, and the tool changes only the wording. Polishing the grammar of a paragraph you wrote is fine. Asking for a paragraph on a topic and then adapting it is not, however much you edit it afterwards. Two cautions apply. Check that the polished version still says what you meant, because these tools change meaning as easily as grammar and sound equally confident either way. And find out what your institution asks you to declare. For language editing it is often nothing, but check the policy rather than assume.
Why does AI invent references that don't exist?
A large language model generates text by predicting plausible next words. It does not look references up in a database of real papers. To the model a reference is a pattern (author, year, journal, title). It can produce one that looks convincing but matches no real paper, with real-sounding authors, a real journal and a plausible title. This is often called hallucination. It is a built-in consequence of how the tool works, and it happens regularly. Treat every AI-suggested reference as unverified: find the actual paper, confirm that it exists and says what you are citing it for, and cite that. Never paste an AI reference list into your work unchecked.
Do I have to disclose that I used AI in my research?
It depends on the policy of your institution and of your target journal, so find out what they require and follow it. The rules are still changing, and they differ. Some ask you to declare any AI assistance, including editing, others only substantive use, and many now require a statement describing how AI tools were used. Two things hold everywhere: AI cannot be listed as an author, because it cannot take responsibility for the work, and you remain fully accountable for everything in the document. When in doubt, disclose. A short note on how you used AI takes little effort and protects you. Hiding substantive use that you were required to declare is a breach of academic integrity.