The syllabus finally says AI is allowed. Great. Then comes the quieter question nobody explained: allowed how, and credited how? A permission slip is not a citation format, and the gap between them is where honest students still trip.

Plenty of courses have moved past the blanket ban. Your professor now says you may use a chatbot to brainstorm, outline, or check grammar, as long as you are open about it. That openness is the catch. “Be transparent” sounds simple until you sit down to write the paper and realize you have no idea whether AI goes in your reference list, in a footnote, in a sentence at the top, or all three. Getting it wrong when you were trying to do the right thing feels worse than getting caught cutting corners.

Here’s the Deal: when a professor allows AI and asks you to disclose it, you owe two separate things. First, a short disclosure that says what you used AI for, in plain language, usually one or two sentences near your methods or in a note. Second, a formal citation if AI generated specific text or ideas you are leaning on, formatted in your required style. APA treats the AI as the author of the output and asks you to describe your use in the text. MLA does not list AI as an author and instead cites the prompt. Whichever style you use, save your prompts and the responses, because that record is what protects you if anyone asks later.

This is the trap that catches careful students. A permission to use AI is not a permission to stay silent about it. Most instructors who open the door still expect a paper trail, and many now write disclosure requirements right into the assignment. If your professor allowed the tool but said nothing about crediting it, treat that as an oversight rather than a green light, and ask. A one-line email removes all doubt, and it signals that you take the rule seriously.

The reason disclosure matters even when use is permitted is trust. A grader wants to know which thinking is yours and which came from a machine, so they can assess your actual learning. Hiding a permitted tool undercuts that, and it can read as dishonest even though the use itself was fine. If your course rules are still fuzzy on the details, our guide to the exact questions to ask about a vague AI policy gives you a script for pinning it down before you submit.

The two styles most students use handle AI in noticeably different ways, so check which one your course requires before you format anything.

APA style. The American Psychological Association treats the generated text like a personal, non-retrievable source and puts the maker in the author slot. A reference entry lists the company as author, the year, the model name and version as the title, the descriptor “Large language model,” and the URL. In the body, you cite it like any other work and, importantly, you describe how you used it in your text or a methods note. APA also suggests keeping the full prompt and response in an appendix when the exchange is long or central to your argument. Their official guidance on citing ChatGPT walks through the exact format with examples.

MLA style. The Modern Language Association goes a different route. It does not name the AI as an author, because it does not consider a tool capable of authorship. Instead you build the citation around the prompt you used, treated as the title, with the tool as the container, plus the version, the company, the date, and the URL. MLA also asks you to acknowledge in your prose that the material was generated and to note that outputs cannot be reproduced exactly. The MLA Style Center page on citing generative AI lays out the template and several worked cases.

If your professor named a different style, such as Chicago, the underlying logic is the same: identify the tool, the version, the date, and your prompt, and describe your use. When you are unsure, the honest, specific note in your own words matters more than nailing every punctuation mark of the format.

Citation and disclosure are not the same task. The citation is the formatted entry; the disclosure is the plain-English statement of what you did. Most graders care more about the second one.

For a research paper or lab report, put a short disclosure in your methods section: one or two sentences naming the tool, the version, and exactly what you used it for, whether that was outlining, generating a first draft you then rewrote, or checking your grammar. For a humanities essay with no methods section, a footnote on the first page or a brief line in an acknowledgments note does the same job. Keep it specific and unglamorous. “I used ChatGPT to brainstorm counterarguments in section two, then wrote and verified the final text myself” tells a grader everything they need and nothing they do not.

Resist the urge to bury it or dress it up. A vague “AI was used in the preparation of this document” invites more suspicion than a precise sentence, because it hides the part the grader actually wants to see.

Here is the habit that separates students who can defend their work from students who just hope nobody asks. Save your chat sessions as you go. Copy the prompts and the responses into a document, or export the conversation, and file it with your draft.

Two things make this worth the five minutes. First, if an instructor or an integrity office ever questions your process, a timestamped log of what you asked and what came back is the strongest evidence you have that your use matched the policy. That protection matters because automated detectors are unreliable, and original human writing gets flagged more often than people expect, as we covered in our piece on what to do when an AI detector flags your original writing. Second, a saved prompt record makes your own citation easier to write, since the date, the model version, and the exact wording are all sitting right there.

One caution while you are pasting things in: be mindful of what you feed a public chatbot, especially unpublished research or anything personal. Our AI tool data-sharing checklist covers what to keep out of the box.

This is the error to burn into memory, because it looks like diligence and lands like fabrication. Do not cite ChatGPT, or any chatbot, as the source for a fact, statistic, or quotation that actually came from a real study.

Say you ask an AI for background on a topic and it hands you a tidy claim with a number attached. If you cite the chatbot for that number, you have done two bad things at once. You have credited a machine for a fact it likely pulled from somewhere real, and you may have passed along a figure the model invented, since chatbots produce confident, wrong citations routinely. The correct move is to treat AI output as a starting point, find the original source it points toward, verify the claim there, and cite that real source instead. You cite the AI only for text or ideas the AI itself generated, such as an outline or a phrasing it produced, never for external facts it merely repeated.

The same discipline applies to rewriting. If you run your own sentences through a tool to smooth them out, know the line between a grammar fix and a full rewrite, which we break down in our comparison of paraphrasers versus grammar checkers for students. A light polish rarely needs a citation; a machine-generated paragraph does.

Assume yes, and confirm with a short email. Permission to use a tool is not the same as permission to hide it, and most instructors who allow AI still expect a disclosure. A one-line question about their preferred format protects you and shows good faith.

It depends on your style and your use. APA does add a formal reference entry and asks you to describe your use in the text. MLA builds a citation around your prompt. In every case, add a plain-language disclosure sentence saying what you used AI for, since that is what graders read first.

List the company as author, the year, the model name and version as the title, the label “Large language model,” and the URL, then cite it in text like any source. APA also recommends describing your use and, for long or central exchanges, including the full prompt and response in an appendix.

No, and this is the common trap. Chatbots repeat or invent facts, so citing one for a statistic credits the wrong source and risks passing along a fabricated number. Track down the original study, verify the figure there, and cite that. Cite the AI only for text it generated itself.

Save your prompts and the responses with a date, filed alongside your draft. Add a short note on what you used the tool for. That timestamped record is your best evidence that your use matched the policy, and it makes writing the citation itself much faster.

• Permission to use AI is not permission to stay quiet; disclose it even when the tool is allowed.
• Write a plain, specific disclosure sentence in your methods, a footnote, or an acknowledgments note, saying exactly what you used AI for.
• Cite in your required style: APA names the maker as author and asks you to describe your use; MLA cites the prompt and does not list AI as author.
• Save your prompts and responses with dates, both as proof and to make the citation easier to write.
• Never cite a chatbot for a fact or statistic from a real study; find and cite the original source instead.

An allowed-AI policy is a gift and a small test at the same time. Your professor is trusting you to use a powerful tool honestly and to show your work, and the students who come out ahead are the ones who treat that trust as seriously as any other rule. Write the one-sentence disclosure. Format the citation in the style you were assigned. Keep the prompt log. It takes ten extra minutes, and it turns “I used AI” from a confession into a credential. For more study guides, browse our learning and courses section.