Citing AI-Generated Content

You used an AI tool. The output shaped your argument, filled a gap, or helped you articulate something you already understood. Now you’re staring at a blank bibliography and wondering whether this even needs a citation — and if it does, what format that citation takes. The problem is not unique to you. Every major style guide updated its documentation after generative AI tools became mainstream, and none of the existing templates translated neatly into AI interaction formats. The result is a patch of genuinely new guidance that most students haven’t found because it doesn’t sit in the same place as everything else they were taught. This guide consolidates that guidance. It covers every major citation style, every major AI tool, the prompt-as-source problem, the non-reproducibility issue, institutional policy variations, and every scenario where the standard template doesn’t quite apply — with real format examples you can adapt and use today.

Why Citing AI Is Not Optional

The instinct to treat AI output as unattributable — like a calculator result or a spell-checker suggestion — misunderstands what generative AI actually produces. When a large language model generates a paragraph explaining a concept, constructing an argument, or summarising a body of research, it produces text that could be mistaken for original intellectual work by the student. That is the problem citation addresses. Academic attribution is not just a bibliographic housekeeping task. It is the mechanism by which readers can evaluate the provenance and reliability of every claim in a paper. An AI-generated explanation carries different epistemic weight than a peer-reviewed study — and readers deserve to know which is which.

There is also a more direct consequence. When AI-generated content appears in a paper without attribution, most academic integrity frameworks classify it as a form of misrepresentation — specifically, as presenting someone else’s (or something else’s) work as your own. The fact that the “someone else” is a machine rather than a person does not change the structural problem. Academic misconduct procedures do not require intent to be dishonest; they require that the work was presented misleadingly. A citation is your explicit signal that the content has an external source.

The Transparency Principle That Drives Every Style Guide’s Guidance

Every style guide that has published AI citation formats — APA, MLA, Chicago — frames its guidance around one consistent principle: readers must be able to understand what sources were used and evaluate them. That principle creates a problem with AI that doesn’t exist with conventional sources. Traditional sources are reproducible — a reader can retrieve the same article, turn to the same page, and verify the same passage. AI responses are not reproducible. Two people submitting the same prompt to the same AI tool will receive different outputs. The same person submitting the same prompt at different times receives different outputs. This non-reproducibility doesn’t eliminate the citation obligation — it intensifies it, and it shapes the specific formatting conventions that the style guides developed.

What Counts as AI-Generated Content

The citation obligation attaches to content — text, ideas, arguments, data, images, code — that was generated by an AI system and incorporated into academic work. The definition is broader than many students initially assume, because AI tools now operate across every type of academic output. Understanding what falls within scope prevents both under-citing (missing attribution where it’s required) and over-citing (treating every AI interaction as a citation-triggering event).

The boundary between these categories is genuinely contested at many institutions, and reasonable instructors disagree. An AI tool that “improved” a sentence is qualitatively different from one that wrote it from scratch — but the output may look identical to a reader. When in doubt, the conservative approach is to disclose. An unnecessary citation causes no harm; a missing one can cause significant harm to your academic standing.

Information Required for Any AI Citation

Before formatting for a specific style, gather the six core data points that every AI citation format requires in some combination. Having these recorded at the time of use — rather than reconstructed after the fact — is essential, because AI interfaces often don’t preserve conversation history in an accessible way, and the tool may return different outputs if the prompt is re-submitted.

APA 7: Full Citation Format for AI-Generated Content

The American Psychological Association was the first major style authority to publish explicit guidance for citing generative AI. The APA Style blog’s official guidance on citing ChatGPT establishes the framework, which applies to all AI text generators, not just ChatGPT. The core decision APA made was to treat the AI company as the author and the tool as a software source — a category that already existed in APA’s framework.

APA 7 Reference List Format

APA 7 In-Text Citation Format

In-text citations for AI-generated content in APA 7 follow the standard author-date format, treating the developer as the author:

The Non-Retrievability Problem in APA

APA’s guidance explicitly addresses one problem that has no precedent in conventional citation: the AI output you received cannot be retrieved or verified by anyone else. Unlike a journal article — where any reader can find the same text at the same URL — an AI response is unique to your session and your moment of use. APA’s recommended solution is to append the response to the paper as supplementary material, typically in an appendix, and reference the appendix from the in-text citation:

Multiple AI Responses from the Same Tool

If you used the same AI tool on multiple occasions with different prompts, each interaction is a separate source and requires a separate reference list entry. APA distinguishes entries with the date — and if you used the tool on the same date with different prompts, add a lowercase letter suffix (2024a, 2024b) just as with any other multiple-reference-per-year situation:

APA Format for AI Tools Embedded in Other Software

When you use an AI tool embedded within another application — such as Microsoft Copilot in Word, or Grammarly’s generative features — the citation follows the same structure, with the embedding context noted in the description:

MLA 9: Full Citation Format for AI-Generated Content

The Modern Language Association updated its guidance to address generative AI following the widespread adoption of ChatGPT in 2023. MLA’s official guidance on citing generative AI takes a structurally different approach from APA, treating the prompt as the title-equivalent and the AI system as the container — the source within which the response exists. This reflects MLA’s Works Cited template structure, which organises sources by author, title, and container in a specific sequence.

MLA 9 Works Cited Format

MLA 9 In-Text Citation

MLA in-text citations use a shortened title in quotation marks — specifically, the opening words of the prompt as they appear in the Works Cited entry. This maintains the standard MLA author-title in-text format, adapted for the absence of a human author:

MLA’s Additional Transparency Requirement

MLA goes further than APA in one specific direction: it explicitly advises writers to reproduce or describe the AI-generated content in the paper — either directly in the text or in an appendix — and to note how the AI output was used. This reflects MLA’s roots in humanities scholarship, where the provenance and interpretation of sources is treated as intellectually significant in itself. An MLA citation for an AI source should ideally be accompanied by a note explaining what role the AI content played in the paper:

Multiple Works Cited Entries for the Same Tool (MLA)

When multiple AI responses from the same tool appear in the same paper, each gets its own Works Cited entry, distinguished by the different prompt text. MLA does not use the year-letter suffix system (that is APA’s approach). Instead, each entry is unique because the prompt — which functions as the title — is different for each:

Chicago Style: Notes-Bibliography and Author-Date

Chicago style operates through two systems: notes-bibliography (used primarily in humanities) and author-date (used in social sciences, some natural sciences). Both systems have been adapted to accommodate AI citations, drawing on Chicago’s existing treatment of software and digital tools. The adaptation is less formally codified than APA or MLA — the Chicago Manual of Style handles AI within its evolving guidance on electronic sources — but the logic is consistent across both systems.

Chicago Notes-Bibliography Format

Chicago Author-Date Format

Chicago’s Approach to AI in Humanities Disciplines

Because notes-bibliography Chicago is dominant in history, literature, philosophy, and related humanities disciplines — fields where methodological transparency has deep roots — many instructors in these areas expect more than a formal citation. They expect a discursive note explaining the nature of the AI interaction: what was asked, why it was used, and how the output relates to the argument. This is consistent with Chicago’s notes format, which accommodates longer discursive footnotes. A well-constructed Chicago footnote for an AI source might read:

Tool-by-Tool Citation Reference: ChatGPT, Claude, Gemini & More

The citation formats above apply universally, but each major AI tool has slightly different interface characteristics that affect how you gather the information for the citation. The comparison below gives the developer name, model format, version identification, URL, and any tool-specific citation considerations for each major platform.

The Prompt-as-Source Problem

Every citation format treats the prompt — the question or instruction you gave the AI — as a critical element of the reference. This creates a genuinely novel situation in academic writing. No prior citation format required the reader to know what question you asked before they could understand the answer you received. With AI, the prompt is not incidental context; it is the fundamental determinant of the output. A different prompt produces a different response. The same prompt on a different day may produce a different response. Documenting the prompt is documenting the source.

Verbatim vs Descriptive Prompts in Citations

Style guides differ on whether to reproduce the prompt verbatim or describe it:

Multi-Turn Conversations as Sources

Many AI interactions involve multiple exchanges — follow-up questions, requests for clarification, or instructions to expand on an earlier response. Citing a response that emerged from a multi-turn conversation is more complex than citing a single prompt-response pair, because the output was shaped by the full conversation history, not just the final prompt.

Non-Reproducibility and the Archive Problem

Every citation format assumes that a sufficiently motivated reader could retrieve the source and verify its content. AI responses break this assumption completely. When you cite a journal article from 2019, the reader can retrieve the same article and read the same text. When you cite an AI response from November 2024, the reader cannot retrieve it. Even if they use the same tool, the same model version, and the exact same prompt, they will receive a different response. The output is stochastic — shaped by the model’s probabilistic text generation — and the same surface prompt generates different outputs across instances.

Practical Archiving Solutions

Because non-reproducibility is a structural feature of AI sources rather than an edge case, both APA and MLA recommend proactive documentation. The practical approach to archiving is straightforward: copy the entire relevant AI response into a separate document at the time of use, noting the date, the tool, the version, and the exact prompt. This archived text becomes the appendix content referenced in your citation. It also protects you in an integrity context — if questions arise about how you used AI tools, you have a contemporaneous record of what the tool generated versus what you wrote independently.

Discipline-Specific Guidance for Citing AI

The same citation format templates apply across disciplines, but the context in which AI citations appear — and the additional explanatory obligations that accompany them — varies significantly by field. What counts as appropriate use and appropriate documentation in a psychology empirical paper is different from what’s expected in a literary analysis, a nursing care study, or a legal memorandum.

Institutional and Instructor Policies: What Overrides Style Guides

The citation formats documented here are the authoritative standards from the style guide organisations — APA, MLA, and Chicago. But in the context of your specific academic submission, institutional and instructor policies hold higher authority than style guide recommendations. A citation format tells you how to attribute AI content if you use it. An institutional policy tells you whether you may use it at all. These are separate questions, and confusing them has consequences.

When the Syllabus Conflicts With Style Guide Defaults

You may encounter a syllabus that says “use APA format” while also saying “AI-generated content is not permitted in any form.” These two instructions are not in conflict — the first tells you how to format your sources, and the second tells you that AI is not a permissible source. In this situation, no AI citation format applies because AI content should not appear in the paper at all. The style guide’s citation template for AI does not grant permission for use; it merely documents how attribution should be structured when use is permitted.

Citing AI-Generated Images and Code

Text generation tools attract the most attention in academic AI citation discussions, but image generation tools and coding assistants are equally subject to attribution requirements when their outputs appear in academic submissions. The core framework is the same — identify the tool, the developer, the prompt, the date, and the access point — but the location and format of the citation adapts to the type of content.

Citing AI-Generated Images

When an AI-generated image appears in a paper as a figure, the attribution appears in the figure caption rather than exclusively in the reference list. The figure caption should identify the image as AI-generated, name the tool and company, describe the prompt, and provide the date:

Citing AI-Generated Code

AI-generated code — produced by tools like GitHub Copilot, Claude, ChatGPT, or other coding assistants — requires attribution within the code itself (as comments) and, when the code constitutes a submitted deliverable, in the submission’s reference list or bibliography. The attribution-in-comments approach is the standard in professional software development contexts; the reference list entry is the academic submission standard.

A critical point for code submissions: many computer science programmes treat AI-generated code exactly as they treat AI-generated text — as potentially prohibited without explicit permission. The additional complication with code is that AI coding tools are embedded in development environments and operate continuously during coding sessions, making it difficult to clearly delineate where AI-assisted code ends and original code begins. If you used a coding assistant, document which specific functions or sections it generated, rather than applying a blanket “AI-assisted” label to the entire submission.

Common Citation Errors and Corrections

AI citation practice is new enough that formatting errors are widespread — in student submissions, in published articles, and even in style guide examples that predated the clarifications published after 2023. The errors below are the most frequently documented in academic writing feedback, each with its root cause and exact correction.

Pre-Submission Checklist for AI Citations

Before submitting any paper that includes AI-generated content, verify every item in this checklist. Each point addresses a specific, verifiable criterion that determines citation compliance.

Frequently Asked Questions About Citing AI-Generated Content

Keeping Your AI Citations Current as Guidance Evolves

The guidance in this article reflects what APA, MLA, and Chicago have formally published as of 2024–2025. This is a genuinely evolving area. All three style guides have updated their AI citation guidance at least once since initially publishing it, and further updates are expected as AI tools become more diverse, as model versioning becomes more complex, and as institutional policies solidify from their current transitional state. The safe approach is to treat this guide as a current-state foundation while checking the style guide’s own official publication for any updates at the time of your submission.

For papers where the citation requirements are unusually complex — a dissertation using AI tools with institutional-specific disclosure requirements, a journal submission with author guidelines that modify default APA treatment of AI sources, or a multi-source paper where AI is one of many sources requiring careful integration — professional formatting support removes the risk of getting these details wrong. Our dissertation and thesis writing service handles format-to-institution compliance for graduate-level work. Our proofreading and editing service audits every reference in a paper against the required style and catches AI citation errors before submission.

The underlying principle that will not change — regardless of how the format templates evolve — is that transparency about source provenance is non-negotiable in academic writing. AI-generated content is no exception to that principle. Knowing precisely how to document that transparency, in the format your course, journal, or institution requires, is the competency this guide aims to build. For further reading on related formatting and academic integrity topics, our guides on citation and referencing across all major style guides, the ethical use of AI in university settings, and writing effective academic introductions extend the context developed here.