
How to cite AI in a seminar paper (without losing your mind)
You were working on a paper with a chatbot, it threw out an idea you had not thought of, you checked it, and it turned out to be right. Now the idea is in your paper. What goes in the citation? "Chatbot, 2026"? Not quite. This guide explains it in a few simple steps.
First of all: ask your lecturer
Before using AI in academic work, make sure your course lecturer, and your university, accept that kind of use. Some courses allow it, some limit it and some forbid it. The course guidelines always come before this guide.
When do you need to cite at all?
Not every use of AI needs a citation. If the tool helped you phrase things, fix mistakes, translate, summarise an article or find sources, in most cases you do not need to cite it. One sentence at the start or end of the paper, saying what you used it for, is enough. For example:
"AI tools were used in this paper for language editing and for finding sources. All claims were checked by me."
You need a citation when the tool gave you a new idea: a claim, a link between two things, a comparison you had not thought of. You checked that idea, and now it is part of your argument.
The golden rule: don't cite the chat. Cite what you checked.
A chatbot is not a source. It cannot back up a claim, and it may answer the same question differently tomorrow. So you cannot write "(Chatbot, 2026)" as if it were a book. What makes the idea reliable is your checking, and that is what you cite.
The method in three steps
1. A marker in the text. Where the idea appears, add a short marker: AI-1, then AI-2 for the next finding, and so on.
2. A short list at the end of the paper. After your usual reference list, start a list called "AI-Assisted Findings". Each finding gets one line: your name, the date, a sentence describing the finding, and which tool you used.
Note: before a finding goes on the list, you must check the AI's claim in the sources themselves. AI can be wrong with complete confidence, and can even invent sources that do not exist. Only a claim you have checked and found to be true can be cited.
3. A verification table. After the list comes a short table of the sources in which you checked the finding, numbered V1, V2 and so on. In each row, write the source and exactly what you checked in it.
That's it. No link to the conversation, no prompt you sent, and no story about how it happened.
A full example
Noah is writing a seminar paper on the poet Thomas Hale. While working with a chatbot, it tells him that the poem "Stump", which Hale published in 1931, uses the same image of a burnt tree that appears in a letter he sent his brother that year. Noah does not stop there. He opens the Collected Poems and reads the poem, then opens the volume of letters and finds the letter, and sees that it really is true. (The poet, the books and the finding in this example are invented.)
In the text:
The image of the burnt tree was not born in the poem. It appeared first in a private letter to his brother (Clark, 2026, AI-1).
At the end of the paper, after the references:
AI-Assisted Findings
AI-1. Clark, N. (2026, November 10). Thomas Hale's poem "Stump" and his 1931 letter to his brother use the same image of a burnt tree [AI-assisted finding; Claude, Anthropic]. Verified: V1, V2.
Verification Sources
| No. | Source | What was checked |
|---|---|---|
| V1 | Hale, T. (1980). Collected poems. Harbor Press. | The poem "Stump", p. 41 |
| V2 | Hale, T. (1975). Letters, 1925–1940. Harbor Press. | The letter to his brother, p. 88 |
If one of these books is also in your main reference list, that is fine. It can appear in both places.
What if the paper uses a different style?
The principle is the same in every style. Only the marker in the text changes. The verification table stays the same, and the entries are written in the style of the paper.
| Style | Marker in the text |
|---|---|
| APA | (Levi, 2026, AI-1) |
| MLA | (Levi, AI-1) |
| Chicago | A footnote that mentions AI-1 |
| Harvard | (Levi 2026, AI-1) |
| IEEE | [AI-1] |
| Vancouver | (AI-1) |
Why does this work? A moment of zoology
When someone discovers a new species of animal, describing it is not enough. One specimen has to be deposited in a museum, with a catalogue number, so that anyone can come and check. Nature made the animal, but the discovery belongs to whoever recognised it and recorded it. In our case the AI threw out the idea, and the discovery was yours: you recognised that it was new and you checked it. The verification table is the museum. That is why the method is called Type-Specimen Citation.
A quick check before you hand in
- Every marker in the text has a line in the list of findings.
- Every finding has at least one verification source.
- A finding you could not verify stays out of the paper, or goes in as a hypothesis, not as a fact.
- If you discover that someone has already written this idea, it is no longer an "AI finding". Cite whoever wrote it first, as usual, in your references.
What if my lecturer doesn't know the method?
Maybe not yet. It is a new proposal from OpenPlag Labs, and not yet an official standard. If your lecturer has approved AI use and there are no instructions on how to cite it, you can use the method and add one sentence at the top of the findings list explaining it. Most lecturers will be glad to see that you checked.
Want to go deeper? The full draft of the method is open for discussion on Academia. And if you need a refresher on the basics, we have a guide to the common citation styles and a guide to disclosing AI use.
Want to see this on your own paper? Run a check. Free during the launch period.
Sign up See a sample reportGet the blog by e-mail
One new post a week, straight to your inbox. No spam, and you can leave with one click.
We will send you a confirmation e-mail. Your address is used only to send the blog. Privacy policy