How to Cite AI in Academic Papers: APA, MLA, Chicago 2026

• The September 2025 APA update now requires citing specific AI chats when it would help readers, not just the tool generally.
• MLA treats your prompt as the title of the source and places the AI tool in the container field — never as the author.
• Chicago treats AI chats as personal communication (like an email), so they go in footnotes, not bibliographies, unless a public URL exists.
• Up to 93% of AI-generated references can contain errors, fabricated citations, or ghost citations — verification is mandatory.
• Using unverified AI sources can lead to failing grades, manuscript rejections, and even legal sanctions, as seen in the Mata v. Avianca case.

Introduction

You’ve used AI to research, draft, or find sources. Now you need to cite it. Where do you start?

If you’ve ever asked ChatGPT for citations, prompted Claude to summarize a paper, or asked Gemini to generate a bibliography, you now face a problem that didn’t exist five years ago: how do you cite a source that didn’t actually write anything, that didn’t read any books, and that frequently makes things up?

The answer changes depending on which style guide your institution follows. APA, MLA, and Chicago all handle AI differently, and their rules have all changed within the last year. A citation format that worked in 2024 may be outdated in 2026.

This guide walks through every major citation style with concrete examples you can copy, explains the verification process you must use before trusting AI-sourced references, and covers what your university and your professors actually require. By the end, you’ll have a repeatable workflow for citing AI content no matter what style you’re writing in.

What Counts as AI-Generated Content (and What Doesn’t)

Before formatting a single citation, you need to know whether the tool you used actually requires one. The line between “tool” and “source” is blurrier than most students expect.

You Must Cite AI When You Use It For

• Paraphrasing AI output — Even if you reworded the response, if the core idea comes from AI, cite it.
• Quoting AI directly — Verbatim text from any AI tool requires a citation.
• Using AI-generated data, statistics, or findings — If a number or result came from an AI, attribute it.
• Incorporating AI-generated arguments or analysis — Interpretations, summaries, or critiques from AI that appear in your paper.
• Citing AI-generated images — DALL-E, Midjourney, or any image generator used in a figure.
• Using AI-sourced references — If AI suggested a source and you verified and used it, you still need to cite AI.

You Generally Don’t Cite AI When You Use It For

• Brainstorming only — Idea generation without incorporating specific text or ideas into your paper.
• Proofreading or grammar correction — Standard editing tools like Grammarly or Word’s built-in checker.
• Concept understanding — Learning from AI without reproducing its explanation in your writing.
• Translation for personal comprehension — Translating a source text to help you understand it (you’d cite the original source instead).
• Code generation not reproduced in paper — Code used for analysis but not included as text.

The working rule: If AI output appears in your paper, in whole or in part, cite it. When you’re unsure, include a brief disclosure in your methodology or acknowledgments.

APA 7th Edition: Citing AI in Academic Papers

The American Psychological Association issued its most significant AI citation update in September 2025. This is the single most important change students and researchers need to know about.

The September 2025 APA Shift: Specific Chats Matter Now

In 2023, APA recommended citing only the AI tool generally (e.g., ChatGPT) because specific chats weren’t reliably retrievable. Now that every major AI tool provides a shareable URL, APA explicitly recommends citing the specific AI chat when it would be helpful for readers.

“Include a reference and in-text citation for a specific AI chat foremostly when doing so will be helpful for readers.” — APA Style Blog (McAdoo, Denneny, & Lee, 2025)^[1]

This means if you quote, paraphrase, or cite AI-generated output, you should reference the exact conversation, not just the tool.

APA Reference Template (Specific AI Chat)

Reference list entry:

AI Company Name. (Year, Month Day). Title of chat in italics [Generative AI chat]. Tool Name/Model. URL of the chat

Example:

Anthropic. (2025, May 20). Essential grammar topics for high school graduates [Generative AI chat]. Claude Sonnet 4. https://claude.ai/share/329173b2-ec93-4663-ac68-4f65ea4f166d

In-text citations:

• Parenthetical: (Anthropic, 2025)
• Narrative: Anthropic (2025)

APA Reference Template (AI Tool Generally)

Cite the AI tool generally when you used it for editing, brainstorming, translating, or creating visuals — not for specific content.

Reference list entry:

AI Company Name. (Year). Tool Name/Model in Italics and Title Case [Description; e.g., Large language model]. URL

Example:

OpenAI. (2025). ChatGPT [Large language model]. https://chatgpt.com/

Key APA Updates for 2025-2026

• Version numbers are no longer required by default. Use the model name (e.g., Claude Sonnet 4) instead of version numbers. If version information is available, include it unitalicized in parentheses.^[1:1]
• Prompts go in the text, not references. Document AI prompts in the Method section or an appendix, not the reference list.^[1:2]
• Date must include year, month, and day. Not just the year.^[1:3]
• The author is always the company, never the AI. AI cannot take accountability for published claims. Credit OpenAI, Anthropic, Google, or Microsoft as the responsible entity.^[1:4]

Practical APA Examples

Paraphrasing AI output:

The primary factors contributing to urban heat islands include reduced vegetation, dark surface materials, and waste heat from buildings and vehicles (Anthropic, 2025).

Quoting AI directly:

When prompted to “explain the greenhouse effect in simple terms,” ChatGPT produced a definition suitable for middle school students (OpenAI, 2025).

Including a transcript:

The AI-generated methodology response covered mixed-methods approaches, sampling strategies, and ethical considerations (OpenAI, 2025; see Appendix A for complete transcript).

MLA 9th Edition: Citing AI as a Text Source

MLA takes a fundamentally different approach. It does not treat the AI as the author. Instead, MLA uses your prompt as the title of the source and places the AI tool name in the container field.

MLA Reference Template

“Prompt description.” AI Tool Name, model Model Name, Publisher, Date, URL.

Example (text paraphrase):

“Describe the theme of nature in Jane Austen’s Mansfield Park” prompt. ChatGPT, model GPT-4o, OpenAI, 23 Sept. 2024, chatgpt.com/share/66f1b0a0-d704-8000-be9a-85f53c850607.

MLA Examples Across Scenarios

Quoting text:

Nature is depicted frequently throughout Mansfield Park, and it “often mirrors the personalities or inner states of the characters” (“Describe the theme”).

Citing an AI-generated image:

Fig. 1. “Create an expressionist-style image of two people standing on a beach looking at the ocean” prompt, DALL-E, version 3, OpenAI, 23 Sept. 2024, chatgpt.com/share/66f1c3a3-3f90-8000-9750-82c57c4a6592.

Quoting AI-generated creative text (poem):

“The Oak Tree” free verse poem. ChatGPT, model GPT-4o, OpenAI, 23 Sept. 2024, chatgpt.com/share/66f1c740-7500-8000-a38b-6d6045c811f5.

Citing secondary sources found by AI:

MLA recommends clicking through to the original sources AI cites and citing them directly rather than citing the AI summary.^[2] If you cite an AI summary that includes sources without consulting them yourself, acknowledge the secondary sources in your work.

MLA In-Text Citation

MLA uses a shortened version of the prompt in quotation marks:

While traditional definitions of critical thinking emphasize logical reasoning, contemporary scholars incorporate creativity and emotional intelligence into the concept (“Define critical” par. 2).

Chicago Style: AI as Personal Communication

Chicago Manual of Style treats AI-generated content like personal communication, the same category as an email, phone conversation, or text message. The rationale is straightforward: unique chat sessions cannot be perfectly replicated or accessed by other readers.^[3]

Chicago Notes-Bibliography Format

Footnote format:

1. Text generated by ChatGPT, OpenAI, March 7, 2023, chat.openai.com/chat.

Shortened note (for subsequent citations):

2. ChatGPT, “Explain quantum entanglement.”

Bibliography entry: Only include if the chat has a publicly accessible URL (e.g., via ShareGPT or AI Archives). Otherwise, omit from bibliography. The footnote is sufficient.

If the prompt hasn’t appeared in your text, include it in the note:

1. ChatGPT, response to “Explain how to make pizza dough from common household ingredients,” OpenAI, March 7, 2023.

Chicago Author-Date Format

In-text citation:

Recent advances in natural language processing have enabled more context-aware AI assistants (Claude, February 15, 2025).

Reference list entry: Only include if a stable public URL exists. Otherwise, document in text only.

Key Chicago Principles

• AI chats go in footnotes, not bibliographies, unless a public URL is available.
• If you edited the AI-generated text, note that in the text or footnote.
• Chicago follows COPE’s position statement that AI cannot be an author.^[3:1]

The Hallucination Problem: Verifying AI Sources

This is the most critical section for academic integrity. Every student using AI in research needs to understand what happens when AI fabricates citations.

What Is AI Hallucination?

AI hallucination refers to the phenomenon where AI models generate plausible-looking but fabricated information, including fake citations, nonexistent authors, incorrect DOIs, and made-up journal names. The models don’t retrieve factual information; they predict text patterns that sound correct.^[4]

The scale is staggering:

• 25–40% of AI-generated references are completely fabricated^[4:1]
• Only 26.5% of AI-generated references are fully accurate^[4:2]
• Up to 93% of AI citations in academic work contain errors or are entirely invented^[4:3]
• Nature (April 2026) reported that “tens of thousands of publications from 2025 may include invalid references generated by AI,” based on an audit finding rates of fabricated citations climbed steeply since 2023.^[5]
• At NeurIPS 2025, approximately 100 hallucinated citations made it through peer review before being flagged.^[4:4]

The Four Types of AI Hallucination

AI-generated references don’t just fail; they fail in distinct, identifiable patterns.

Type 1: Fully Invented
A fake title, author, and DOI — nothing exists.

Example: “Smith, J. (2024). The cognitive effects of prolonged AI use. Journal of Educational Psychology, 45(3), 112–125.” → No such article exists.

Type 2: Chimera
A real author or journal paired with a fake paper title.

Example: “Chen, L. (2024). AI hallucination patterns in legal reasoning. Harvard Law Review, 138(4).” → Chen publishes real work, but this paper doesn’t exist.

Type 3: Distorted
A real paper with incorrect year, volume, DOI, or page numbers.

Example: A paper cited with a DOI that resolves to a completely different article.

Type 4: Ghost
A real paper used inappropriately for unrelated claims — the source exists but was never actually cited by or relevant to the original paper.

Example: Claiming a medical study proves a conclusion the study never supported.

Real-World Consequences: The Mata v. Avianca Case

The most famous example of hallucinated citations making it into official documents is the 2023 Mata v. Avianca case, which set a legal precedent for AI use consequences.^[4:5]

The facts: Attorney Steven A. Schwartz and his colleague Peter LoDuca filed a brief in Mata v. Avianca (Southern District of New York) that cited six case names fabricated by ChatGPT. The AI generated plausible-sounding legal precedents, including Varghese v. China Southern Airlines Co., Ltd., complete with fabricated judicial opinions and quotes. When opposing counsel and the judge could not find these cases, Judge P. Kevin Castel questioned their existence. Schwartz then used ChatGPT again to verify the cases, and the AI falsely assured him they were real and could be found on reputable legal databases.

The ruling: Judge Castel ordered Schwartz, LoDuca, and their firm (Levidow, Levidow & Oberman) to each pay a $5,000 fine for failing to verify sources and standing by the fake cases even after their existence was questioned.^[4:6]

Why this matters to students: Universities classify fabricated citations as academic misconduct regardless of intent. One student citing unverified AI sources could face a failing grade, an honor code violation, or a course failure.^[4:7]

The 4-Step Verification Workflow

Never paste an AI-generated reference into your paper without verifying it. Follow this systematic process:

Step 1: Resolve the DOI

Copy the DOI into https://doi.org. If it returns a 404 error or redirects to an unrelated article, the citation is fake. Even a working DOI can be fabricated, so check the metadata after resolving.

Step 2: Search the Exact Title

Search the article’s full title in quotation marks on Google Scholar, CrossRef, or PubMed. If the article doesn’t appear in two or three databases, it likely doesn’t exist.

Step 3: Verify Author and Journal

Check the journal’s ISSN and scope. If a paper about “AI hallucination in dental studies” appears in a journal focused on mechanical engineering, you’ve found a chimera. Look up the author on ORCID or Google Scholar to confirm the paper appears in their publication history.

Step 4: Read the Source

Even if a citation passes all metadata checks, it might misattribute claims. Open the actual paper and verify it supports your argument. This step catches ghost citations: real papers cited for conclusions they never made.^[4:8]

Where Hallucinations Are Most Likely

AI-generated citations are most unreliable in:

• Niche or specialized topics
• Recent publications (2024–2026)
• Book chapters and book sections
• Legal precedents and medical journals
• Non-English-language sources

When AI is a Tool vs. a Source

This distinction matters more than most students realize. Using AI as a tool for brainstorming, editing, or understanding is fundamentally different from using it as a content source.

When AI is a Tool (Disclose, Don’t Cite)

Cite the AI tool generally or include a disclosure statement in your methodology:

• Editing and proofreading — Used AI to refine grammar, clarity, or flow
• Brainstorming thesis statements — AI generated ideas you developed independently
• Concept explanation — Used AI to understand a difficult topic
• Translation — Used AI to translate text for comprehension (cite the original source)
• Research outlining — Used AI to structure a paper you wrote

When AI is a Source (Cite in Reference List)

• Quoting or paraphrasing AI text — Directly cite specific AI output
• Using AI-generated data — Numbers, statistics, or findings from AI
• Including AI-generated visuals — Figures, diagrams, or images created by AI
• Citing AI-suggested secondary sources — Even verified, acknowledge the AI path

APA’s Guidance on Both Scenarios

APA explicitly separates these two use cases. When you used AI for editing or analysis of your own writing, the citation goes in the Method section or author note:

“I used ChatGPT (OpenAI, 2025) to edit my paper and provide feedback on clarity and flow.”^[1:5]

When you incorporated AI-generated content, cite the specific chat with full reference format.^[1:6]

Citation Templates for Every Scenario

Use this table as a quick reference when you’re formatting citations. The examples show the same AI interaction cited across all three major styles.

Scenario 1: Paraphrasing AI Text

Scenario 2: Quoting AI Text

Scenario 3: AI-Generated Image as Figure

Scenario 4: AI as Research Assistant

Scenario 5: Secondary Sources Found by AI

Key principle: Always cite the original, verified source, not the AI, when the AI acted as a pointer to a real publication. If you used AI to paraphrase content from that source, cite the original. If you incorporated AI’s original explanation, cite the AI.

Best Practices for Transparent AI Use

Citing AI properly is only part of ethical AI use. Follow these practical guidelines:

1. Verify Every AI-Sourced Reference

• Never paste AI-generated citations directly into your paper. This is the single most important habit you can develop.
• Use doi.org to check every DOI.
• Search exact titles in Google Scholar, CrossRef, and PubMed.
• Open the actual paper and confirm it supports your claim.^[4:9]

2. Document Your AI Prompts

Save screenshots or transcripts of every AI interaction. This creates an audit trail that demonstrates transparency and protects you from accusations of academic dishonesty.

3. Check Your Syllabus and University Policy

Many universities have specific AI policies that go beyond citation requirements. You’ll need to check whether your institution requires:

• A formal AI disclosure statement
• Documentation of exactly how AI was used
• Both disclosure and formal citation

4. Save Chat Transcripts

If you used AI to generate substantial text, save the conversation. Many professors expect you to include a transcript in an appendix or supplement.

5. Be Honest with Your Professors

If you used AI extensively, disclose it upfront. Transparency builds trust. Professors generally prefer honest disclosure over discovering AI use through plagiarism detection tools.

What Your Professor Should Know

Proper citation of AI content isn’t just about formatting rules. It’s about academic integrity, transparency, and protecting yourself from accusations of dishonesty.

Why Citation Matters for Academic Integrity

When you cite AI, you’re doing three things:

1. Acknowledging the source — Readers can see where ideas, text, and data originated.
2. Demonstrating transparency — Your reader understands your research workflow.
3. Protecting yourself — Citing AI explicitly frames it as a tool, not as plagiarism.

How to Talk to Your Professor About AI Use

If you’re unsure whether your professor allows AI, here’s how to handle it:

• Ask directly. Most professors will clarify their stance. A simple email (“Should I cite AI if I use it for research?”) is respectful and professional.
• Disclose broadly when uncertain. It’s always safer to disclose more than less.
• Check your syllabus first. Many courses now include an AI policy section in the syllabus.

Policy Compliance Tips

• Read your university’s AI policy. Requirements vary widely across institutions.^[6]
• Check journal policies. If you’re writing for publication, publisher AI guidelines may differ from your university’s requirements.^[5:1]
• Follow the stricter policy. If your professor requires disclosure and a journal requires it too, comply with both.
• Don’t assume silence means permission. In an increasing number of university policies, “no policy stated” means you should disclose AI use as a precaution.

Conclusion

Citing AI-generated content in academic papers isn’t complicated, but it is different from citing traditional sources and it changes every few months as style guides update. The core principles remain consistent:

• Cite AI as a source when you use its output; disclose it as a tool when you use it for editing or brainstorming.
• APA now requires specific chat citations; MLA treats prompts as titles; Chicago treats AI chats as personal communication.
• Always verify AI-generated references. The 26.5% accuracy rate means the majority of AI citations will contain errors.
• Document your AI use. Screenshots, transcripts, and honest disclosure protect you from academic integrity violations.

The skills you’re learning here aren’t just about getting your current paper right. They’re about building habits that will serve you through every academic and professional paper you write. In an academic landscape where AI adoption is accelerating, transparency isn’t optional; it’s the foundation of credible scholarship.

References

Related Resources

• How to Cite AI Tools (ChatGPT, Claude) in Academic Papers — A companion guide on citing AI tools themselves, not the content they generate.
• How to Use AI Tools Ethically in Academic Writing — Navigating AI assistance while maintaining academic integrity.
• How to Cite Podcasts and YouTube Videos — Citation formats for multimedia sources across APA, MLA, and Chicago styles.

FAQ

Do I need to cite AI if I only used it for brainstorming?

Generally no, but check your institution’s policy. Most guidelines require citation only when you incorporate AI output into your paper. If you used AI for idea generation without using any specific text or ideas it produced, no citation is typically needed.

What if my AI tool doesn’t provide a shareable URL?

For APA, use the tool’s general URL. For MLA, include the tool URL you accessed. For Chicago, treat it as personal communication — include the chat date and tool name in the footnote only.

Can I cite AI as the author?

No. Neither APA, MLA, Chicago, nor Harvard treats AI as an author. AI cannot take accountability for published claims. Credit the AI developer (OpenAI, Anthropic, Google, Microsoft) as the responsible entity.

How do I cite AI-generated references I verified myself?

Cite the verified source directly, not the AI. If you relied on AI to find the source, disclose this in your methodology section but cite the actual publication in your reference list.

Citing AI properly can be confusing. Our academic writers ensure every source is correctly formatted — including AI-generated content. Get help today. Order assistance now.

This guide synthesizes updated guidance from the APA Style Blog (September 2025), MLA Style Center (2025), Chicago Manual of Style (2025), peer-reviewed research on AI hallucination (2026), and university AI citation policies. All external sources were verified as of June 2026.