How to Cite ChatGPT in APA, MLA, and Chicago
Format-specific citation instructions for ChatGPT and other generative AI tools — covering reference list entries, in-text citations, footnotes, version identification, the reproducibility problem, and what your institution’s academic integrity policy actually requires you to disclose.
The first time most students use ChatGPT for academic work, the citation question comes as an afterthought — if it comes at all. You use the tool, it helps, and then you submit the work without acknowledging it. The problem is that citation and disclosure obligations for AI-generated and AI-assisted content now exist in formal style guides and in institutional academic integrity policies, and not knowing the rules does not protect you from their consequences. APA, MLA, and Chicago all published specific guidance on citing ChatGPT and similar tools within months of ChatGPT’s public release. The rules are not ambiguous — they are just new, unfamiliar, and inconsistently taught. This guide covers every format, every scenario, and every edge case so that the citation question is never an afterthought.
What Citing an AI Chatbot Actually Requires — and Why It Is Different From Any Other Source
Citing a source in academic writing serves two functions: it gives credit to the originator of an idea or piece of information, and it enables a reader to locate and verify that source. Traditional citation formats — for books, journal articles, websites, and interviews — have evolved around these two functions. ChatGPT citation is genuinely different in both respects, and understanding why helps explain the specific choices each style guide makes in handling it.
ChatGPT is not an author in the conventional sense. It does not have original ideas it can claim credit for — it generates statistically probable text based on training data, without consciousness, intention, or verifiable knowledge of the subject it is addressing. At the same time, the organisation that built it — OpenAI — is a real, citable entity, and the model and version it runs are technically identifiable. This makes the “who gets credit” function of citation applicable, though the object of credit shifts from individual human authorship to corporate product development.
The verification function of citation is where AI tools present their most significant challenge. You cannot verify ChatGPT’s output by accessing the same source a reader accesses — because ChatGPT’s responses vary with every session, are influenced by model updates, and are not archived in any publicly retrievable form. Two people entering the same prompt will receive different responses. The same prompt entered at different times will produce different results. This means that any citation of ChatGPT output can, at best, describe what was generated rather than reliably direct a reader to that specific text. The citation becomes more of a disclosure than a traditional academic reference.
The practical implication of all this is that AI citations function primarily as transparency disclosures. When you cite ChatGPT in a paper, you are telling your reader and your instructor that you used this tool, when you used it, what you asked it, and roughly what it generated. You are not directing them to a citable piece of scholarship. All three major style guides have built this reality into their citation formats — which is why those formats look different from a typical article or website citation and why they emphasise the date, the version, and the prompt in ways that standard references do not.
The question of how to cite ChatGPT is separate from the question of whether you should use it as an academic source at all. ChatGPT produces plausible-sounding text, not verified information. It fabricates citations, invents statistics, and confidently states inaccuracies at a rate that makes it unreliable as an evidence source for factual academic claims. Virtually every university library and academic writing guideline specifies that ChatGPT should not be cited as a source for factual claims in the same way peer-reviewed literature is cited.
The appropriate reason to cite ChatGPT in academic writing is when the output itself is what you are discussing, analysing, or using as a primary source for a study of AI behaviour — not as verification of a factual claim. If you used ChatGPT to help generate an outline, edit your prose, or brainstorm ideas, your institution’s academic integrity policy (not a citation format) governs whether and how you must disclose that use. See the section on institutional policies below.
APA 7th Edition — How to Cite ChatGPT in Every Scenario
The American Psychological Association addressed ChatGPT citation directly through its official APA Style Blog guidance published in April 2023. The guidance treats ChatGPT as a software product created by OpenAI and applies the APA format for algorithmic and software outputs, with adaptations for the non-retrievable nature of chatbot responses. This is now the authoritative source for APA ChatGPT citation — check it directly for the most current version, as it may be updated as models and terminology evolve.
The Standard APA Reference List Entry for ChatGPT
APA treats ChatGPT as a piece of software — specifically an algorithmic tool — with OpenAI as the author/publisher. The standard reference format is:
OpenAI. (Year, Month Day). ChatGPT (version identifier) [Large language model]. https://chat.openai.com — Example with real details: OpenAI. (2024, March 15). ChatGPT (GPT-4o) [Large language model]. https://chat.openai.com
Breaking down each element: OpenAI is the author — the organisation that created and operates the tool. The date is the date of the specific conversation you are citing, not the date ChatGPT was released. ChatGPT is the title of the tool. The version identifier in parentheses identifies which model generated the output (GPT-4, GPT-4o, GPT-3.5, etc.). The bracketed descriptor [Large language model] functions like the format descriptor in a standard APA reference (e.g., [Video], [Dataset]). The URL is the platform’s homepage — not a specific conversation link, since ChatGPT conversation URLs are not publicly retrievable without a shared link.
APA Reference List — Three Scenarios
Scenario 1 — General use, current default model:
Scenario 2 — GPT-3.5 (free tier conversations prior to GPT-4o default):
Scenario 3 — When the exact version is unverifiable:
Note on Scenario 3: When you cannot confirm which version was active during your conversation, APA permits omitting the version identifier. However, if you can identify the version through your account settings or by asking ChatGPT directly (“What version are you?”), include it — it strengthens the citation’s transparency.
APA In-Text Citation for ChatGPT
APA uses an author-date system for in-text citations. For ChatGPT, this translates directly: OpenAI is the author, and the date is the date of the conversation. There are no page numbers, paragraph numbers, or section identifiers — because ChatGPT output has none of these.
— Parenthetical citation (most common): (OpenAI, 2024) — Narrative citation (author name in sentence): According to OpenAI (2024), [paraphrase of output]... — Citing a direct quotation of AI output: "[Exact text from ChatGPT response]" (OpenAI, 2024) — Multiple ChatGPT conversations in same paper: (OpenAI, 2024a) and (OpenAI, 2024b) → Use 'a', 'b', 'c' suffix when multiple conversations on same date
The Prompt and Response Appendix — APA’s Recommended Practice
Because ChatGPT output is not retrievable by a reader, APA’s guidance strongly recommends — and some instructors require — including the full prompt and the complete ChatGPT response as a labelled appendix in your paper. This appendix does not replace the reference list entry; it supplements it by providing the complete context that the citation alone cannot convey.
How to Format an APA ChatGPT Appendix
Label the appendix clearly: “Appendix A: ChatGPT Conversation, [Date].” Include the exact prompt you entered, formatted as a direct quotation or labelled clearly as “Prompt:”. Follow it with the full ChatGPT response, labelled “Response:”. If the conversation involved multiple turns, include each exchange in sequence. Note the version (GPT-4o, GPT-3.5, etc.) at the top of the appendix.
In your paper’s reference list entry, you can add a note directing the reader to the appendix: “See Appendix A for the full prompt and response text.” Some instructors and institutions require the appendix; others treat it as optional but best practice. When in doubt, include it — it demonstrates transparency and demonstrates that you engaged with the tool critically rather than unreflectively.
The appendix also serves a practical purpose for your own writing process: recording what you asked and what the model generated creates a reference trail that makes it easier to accurately characterise the AI output in your paper and to identify where you paraphrased, summarised, or quoted directly from it.
MLA 9th Edition — Citing ChatGPT in Works Cited and In-Text
The Modern Language Association updated its guidance on citing generative AI tools in its official MLA Style resource on citing generative AI, published through the MLA Style Center. MLA’s approach is more integrated than APA’s — it incorporates the AI prompt directly into the Works Cited entry rather than relegating it to an appendix, reflecting MLA’s general emphasis on providing context for source use within the citation itself.
MLA Works Cited Entry Format for ChatGPT
MLA’s core citation template for ChatGPT output is structured around the prompt used to generate the content — treating that prompt as the “title” of the source, with ChatGPT as the container (analogous to a journal or website in standard MLA format). The format follows MLA’s nine-element structure for any source.
"[Descriptive title or abbreviated prompt]." ChatGPT, version, OpenAI, day month year, chat.openai.com. — Example 1: Using an abbreviated prompt as title: "Summarize the causes of the 1929 Wall Street Crash in 200 words." ChatGPT, GPT-4o, OpenAI, 14 Oct. 2024, chat.openai.com. — Example 2: Descriptive title when prompt is very long: "Analysis of Hamlet's soliloquies." ChatGPT, GPT-4o, OpenAI, 3 Nov. 2024, chat.openai.com.
Notice the formatting details that are easy to miss: the “title” (prompt) is in quotation marks, not italics. ChatGPT is italicised as the container/tool name. OpenAI follows as the publisher. The date is in MLA’s day-month-year format with the month abbreviated (Oct., Nov., Jan., etc. — except May, June, and July which are spelled out in full). The URL is the platform homepage, not a specific conversation link.
MLA Works Cited — Multiple Scenarios
Nursing essay — using AI to explain a concept:
Literature analysis — using AI to brainstorm themes:
When version is unknown:
MLA Note on Long Prompts: If your actual prompt was very long, MLA permits using a shortened, descriptive version as the title — something that accurately represents the nature of the query. The full prompt can then be included in a note or appendix. MLA also recommends adding a brief statement in your paper’s text or in a note that AI was used and what for.
MLA In-Text Citation for ChatGPT
MLA’s in-text citations use the first element of the Works Cited entry — which, for ChatGPT sources, is the abbreviated prompt in quotation marks. This is unusual compared to most MLA sources (which use the author’s last name) because ChatGPT entries do not have a human author. MLA’s position is that the prompt is the most meaningful identifier for distinguishing between multiple AI outputs in the same paper.
— Standard parenthetical (abbreviated prompt as identifier): ("Explain the pathophysiology of type 2 diabetes") — Narrative form (integrating the source in the sentence): According to a ChatGPT response generated using the prompt "Explain the pathophysiology of type 2 diabetes," [paraphrase]... — When prompt title would be awkward mid-sentence: When queried about diabetic pathology, ChatGPT (GPT-4o) generated an explanation that [paraphrase]... ("Explain the pathophysiology") — For a direct quotation from AI output: ChatGPT described this as "a condition characterised by insulin resistance and progressive beta-cell dysfunction" ("Explain the pathophysiology of type 2 diabetes").
MLA’s Position on Transparency and Disclosure
Beyond the mechanical citation format, MLA places significant emphasis on transparency about how and why AI was used. The MLA Style Center recommends that writers include a brief explanatory note — either in the text, in a footnote, or in a Works Cited note — describing the role the AI tool played in the work. This is not a replacement for the formal citation, but a supplement to it: it tells the reader whether AI was used for drafting, editing, summarising, translating, or generating ideas, which the citation format alone does not convey.
MLA Transparency Note — What to Include
A brief MLA transparency note might read: “I used ChatGPT (GPT-4o) to generate an initial outline for this essay, which I then substantially revised. The AI-generated outline is not reproduced in this paper, but I cite the interaction in the Works Cited list to ensure full disclosure of AI involvement.” Or, more simply: “This paper used ChatGPT to [specific purpose]. All AI-generated text was independently verified against peer-reviewed sources before inclusion.”
The transparency note can appear as a footnote on the first page, in an Author’s Note section at the end of the paper, or as a parenthetical note in the Works Cited entry itself (some instructors specify their preference). When in doubt, ask your instructor before submission — this is an area where institutional expectations still vary considerably, and a note you add voluntarily is always better received than a disclosure you failed to make.
Chicago Style — Notes-Bibliography and Author-Date Formats for ChatGPT
The Chicago Manual of Style (CMOS) addressed AI citation through its Q&A section, and its guidance takes account of the fact that Chicago has two citation systems used in different disciplines: Notes-Bibliography (used in history, arts, and some humanities) and Author-Date (used in some social sciences and natural sciences). The treatment differs slightly between the two, though the core information required is the same.
Chicago Notes-Bibliography Format (Footnotes and Endnotes)
In Notes-Bibliography style, ChatGPT is cited in a footnote or endnote, with a corresponding entry in the bibliography. Chicago’s treatment is the most flexible of the three — it does not prescribe a single rigid format but instead provides principles for what information should be included.
— First full footnote reference: 1. ChatGPT, response to "[Your prompt]," OpenAI, Month Day, Year, https://chat.openai.com. — Example: 1. ChatGPT (GPT-4o), response to "Summarise the key arguments of John Rawls in A Theory of Justice," OpenAI, November 8, 2024, https://chat.openai.com. — Shortened subsequent footnote: 3. ChatGPT, "Summarise the key arguments of John Rawls." — Bibliography entry (alphabetised under O for OpenAI): OpenAI. ChatGPT (GPT-4o). "Summarise the key arguments of John Rawls in A Theory of Justice." Accessed November 8, 2024. https://chat.openai.com.
Chicago Author-Date Format (In-Text and Reference List)
In Author-Date format, Chicago citations work similarly to APA — author name and year in the text, with a full reference list entry. For ChatGPT, OpenAI functions as the author, and the reference list entry follows Chicago’s parenthetical source format.
— In-text citation: (OpenAI 2024) — Reference list entry: OpenAI. 2024. ChatGPT (GPT-4o). Accessed November 8, 2024. https://chat.openai.com. — With specific conversation context: OpenAI. 2024. ChatGPT (GPT-4o), response to "Summarise the key arguments of John Rawls in A Theory of Justice," November 8, 2024. https://chat.openai.com.
Footnote Number → Full Note → Bibliography
Footnote includes: ChatGPT (version), response to “[prompt],” OpenAI, Date, URL. Bibliography entry begins with OpenAI, alphabetised under O. Subsequent footnote references use a shortened form. The full prompt can appear in the footnote or be described in general terms — Chicago’s guidance permits both.
In-Text (OpenAI Year) → Reference List
In-text parenthetical uses (OpenAI Year) — matching APA closely. Reference list entry: OpenAI. Year. ChatGPT (version). Accessed Date. URL. When multiple conversations exist in the same year, disambiguate with letter suffixes: OpenAI 2024a, 2024b. Include “Accessed” date because the content is not archivally retrievable.
Discipline Determines the System
History, arts, literature, and many humanities disciplines use Notes-Bibliography. Natural sciences and some social sciences use Author-Date. If your assignment does not specify, check your department’s preferred Chicago format or ask your instructor. When Chicago is specified without further detail in humanities courses, Notes-Bibliography is almost always intended.
Explanatory Notes Are Encouraged
Chicago style permits — and for AI tools, the Manual effectively encourages — using the flexibility of the footnote system to explain how the AI was used. A footnote citing ChatGPT can include a parenthetical note: “(This response was used to generate an initial outline, subsequently revised substantially by the author.)” This integration of citation and transparency is a natural fit for Chicago’s discursive footnote tradition.
The Reproducibility Problem — Why AI Citation Cannot Function Like a Standard Reference
Every conventional academic citation rests on an implicit assumption: that a reader who follows the reference can access the same source you accessed. A journal article has a DOI and a permanent URL. A book has an ISBN. A website has an archived URL. A court case has a case number. ChatGPT has none of these forms of stable, retrievable identity for its outputs. This is not a technical limitation that will eventually be solved — it is a fundamental feature of how generative language models work. Understanding it explains why the citation formats above include elements that look unusual in a traditional academic reference.
Stochastic Output
ChatGPT generates text probabilistically. Even with identical prompts, model settings, and conditions, outputs vary. There is no canonical version of a response to a given prompt that a reader could retrieve and compare to what you cite.
Model Updates
OpenAI updates its models continuously. The GPT-4o that generated a response in March 2024 is not identical to the GPT-4o of November 2024. Model weights, fine-tuning, and safety training evolve. A prompt entered after a model update may produce substantially different output.
Session Architecture
Conversation history is stored in the user’s account but is not publicly retrievable. Shared conversation links exist but are not permanent archives — they can be deleted by the originating user. There is no centralised repository of ChatGPT outputs equivalent to PubMed or Google Scholar.
These three features combine to make ChatGPT citation an act of documentation rather than referencing. You are documenting what you did, when you did it, and what was generated — not pointing to a stable, independently verifiable source. This is why all three style guides prioritise the date of the conversation, the version of the model, and the prompt used: these elements collectively give a reader the best available approximation of the conditions under which the cited output was generated, even though they cannot recreate that output exactly.
The practical consequence for your writing is that ChatGPT citations should always be accompanied by enough contextual information — either in the text itself, in a note, or in an appendix — that a reader understands what the output said. You cannot simply cite (OpenAI, 2024) and expect the reader to follow the reference as they would follow a journal article citation. When you use or quote AI output, describe it fully in your text, because the citation alone cannot serve the verification function that makes academic citation meaningful.
Identifying Which Version of ChatGPT You Used — and Why It Matters
All three citation formats require or recommend including the version of ChatGPT that generated the output. This is not bureaucratic pedantry — different versions of the model have different capabilities, training data cutoffs, and behavioural profiles. GPT-3.5, GPT-4, GPT-4 Turbo, and GPT-4o produce meaningfully different outputs for many prompts. Specifying the version is a form of methodological transparency: it tells the reader what level of capability was involved in generating the output you are citing.
model parameter in your API request. Options include gpt-4o, gpt-4-turbo, gpt-4, gpt-3.5-turbo.The most common citation problem with ChatGPT is not knowing which version was used or when — because the student did not record this information at the time and is now trying to reconstruct it. The solution is simple: before you begin any ChatGPT session you might cite, record the date, the version shown in the interface, and save or copy the complete conversation. ChatGPT conversations are stored in your account history, but they can be deleted, and account access can be lost.
A practical workflow: create a document titled “ChatGPT Research Log” for each paper or project. For each AI session, note: date, version, full prompt(s), key content of the response. This log becomes your appendix material if required and your verification resource if your citation is questioned. It takes ninety seconds to create at the start of each session and prevents reconstruction problems later.
Citing Other Generative AI Tools — Gemini, Claude, Copilot, Perplexity, and Bard
ChatGPT is the most widely used AI chatbot in academic contexts, but it is not the only one that students cite. Google Gemini (formerly Bard), Anthropic’s Claude, Microsoft Copilot, Perplexity AI, and Meta AI are all tools that appear in academic citation questions. The citation principles are identical across all of them — the format elements change to reflect each tool’s developer and naming — but the specific details differ enough to warrant covering each.
Google Gemini
Developer: Google LLC (or Google DeepMind). Tool name: Gemini. APA: Google. (Year, Month Day). Gemini (model version) [Large language model]. https://gemini.google.com
MLA: “Prompt.” Gemini, version, Google, day month year, gemini.google.com.
Anthropic Claude
Developer: Anthropic. Tool name: Claude. APA: Anthropic. (Year, Month Day). Claude (claude-3-5-sonnet, or version) [Large language model]. https://claude.ai
MLA: “Prompt.” Claude, claude-3-5-sonnet, Anthropic, day month year, claude.ai.
Microsoft Copilot
Developer: Microsoft. Tool name: Microsoft Copilot (formerly Bing Chat). APA: Microsoft. (Year, Month Day). Microsoft Copilot [Large language model]. https://copilot.microsoft.com
Note: Copilot is GPT-4-based but developer is Microsoft, not OpenAI.
Perplexity AI
Developer: Perplexity AI Inc. Perplexity retrieves live web sources and generates summaries — it is functionally different from ChatGPT, closer to a search engine. Cite the specific sources it retrieved where possible rather than Perplexity itself for factual claims.
Google Bard
Bard was rebranded as Gemini in February 2024. Papers written before that date citing Bard should use: Google. (Year, Month Day). Bard [Large language model]. https://bard.google.com — Do not retroactively change these to Gemini citations.
Meta AI / Llama
Developer: Meta Platforms, Inc. Tool name: Meta AI. APA: Meta. (Year, Month Day). Meta AI (Llama version) [Large language model]. https://www.meta.ai — Academic use of Meta AI is currently less common but follows the same citation principles.
The unifying principle across all of these is: developer organisation as author, tool name as title, version in parentheses where available, [Large language model] as descriptor, date of conversation, URL of platform. This template applies to any AI chatbot or language model tool, current or future. When a new AI tool appears and no style-specific guidance exists yet, applying this template produces a citation that is internally consistent with the established guidance for ChatGPT and analogous tools.
When You Are Required to Cite AI — and When Disclosure Is the Right Mechanism
A persistent confusion in student AI usage is conflating two distinct obligations: the obligation to cite AI as a source (when AI output appears in your work), and the obligation to disclose AI use (when AI assisted your process but no AI text appears in the final submission). These are different requirements governed by different policies, and handling them correctly requires understanding the distinction.
Cite When: AI Output Appears in Your Work
If you quote, paraphrase, or closely reproduce ChatGPT-generated text in your submission — even rewritten or substantially revised — a formal citation using your required style (APA, MLA, Chicago) is required. The citation acknowledges that the intellectual content originated from an AI tool rather than from your independent analysis or a published human author.
Disclose When: AI Assisted Your Process
If you used ChatGPT to generate an outline, brainstorm ideas, improve your grammar, or suggest search terms — but no AI-generated text appears in your submission — a process disclosure note is required by many institutions rather than a formal citation. Check your course policy. Disclosure without citation is often appropriate here.
Neither Is an Option: AI Prohibited Entirely
If your institution or assignment prohibits AI use entirely, neither citing nor disclosing covers a violation — the use itself is the issue. Review your syllabus, assignment brief, and institutional academic integrity policy before using any AI tool. “I cited it properly” is not a defence for using a prohibited tool.
The most important thing to understand is that proper citation does not legitimise AI use that is otherwise prohibited. Citation formats are mechanical tools for acknowledging sources — they say nothing about whether the use of a source was academically permissible. An assignment that prohibits AI assistance is not satisfied by properly formatting the ChatGPT citation. The citation question and the permissions question are entirely separate.
That said, where AI use is permitted — which is increasingly the case across many types of assignments at many institutions — correct and complete citation and disclosure is the professional and ethical standard. The ethical use of AI tools in university settings involves not just avoiding prohibited uses but actively documenting permitted ones so that your work’s provenance is clear and your academic integrity is unambiguous.
Paraphrasing Versus Quoting ChatGPT Output — What Changes in the Citation
Academic writing rarely reproduces sources verbatim — the norm is paraphrase, summary, and synthesis, with direct quotation reserved for passages where the exact wording matters. The same norms apply to AI-generated content, but with an additional complication: because ChatGPT output has no stable, verifiable text, the distinction between “paraphrase” and “direct quotation” is less analytically meaningful than it is for human-authored sources. Nevertheless, the conventions of each style guide apply.
1. “Exact text,” ChatGPT (GPT-4o), response to “[prompt],” OpenAI, Date, URL. Quotation marks around the AI text in your body copy.A practical note on direct quotation of AI output: most instructors and style guides implicitly discourage it unless the AI’s specific wording is analytically necessary. AI-generated prose does not have the authority of a subject expert, so quoting it verbatim offers no scholarly advantage over paraphrase and may actually undermine your paper by suggesting you are treating AI output as an authority comparable to a published source. Use direct quotation of ChatGPT output sparingly, and only when the specific language matters to your argument.
How Institutional Academic Integrity Policies Affect Your Citation Obligations
Citation format (APA, MLA, Chicago) is standardised across institutions. Academic integrity policy is not. Each university, college, and course instructor sets their own rules about whether, how, and to what extent AI tools may be used — and what disclosure or citation is required for any AI involvement. This means that correctly formatting your ChatGPT citation in APA style does not guarantee you have met your institution’s specific requirements. You need both.
Step 1 — Check Your Course Syllabus First
Most instructors now include an explicit AI use policy in their course syllabi. Look for sections titled “AI Tools,” “Generative AI,” “Academic Integrity,” or “Permitted Resources.” The syllabus policy overrides general institutional policy for your specific assignment. “The university allows AI” is not a defence if your course syllabus prohibits it.
Step 2 — Review the Institutional Academic Integrity Policy
Your university’s academic integrity or academic honesty policy documents the institutional position on AI tools. Many institutions updated these policies after November 2022 and again after GPT-4’s release. Find the current version in your student handbook or the institution’s policy database — not the version from 2021 that predates ChatGPT.
Step 3 — Ask When Policy Is Ambiguous
If your syllabus does not mention AI, or if it says something vague like “all work must be your own,” ask your instructor directly before using AI tools. Email them with specific questions: “May I use ChatGPT for [specific purpose]? If so, how should I disclose or cite this use?” A written response creates a record and clarifies expectations before submission.
Step 4 — Apply Both: Citation Format + Institutional Disclosure
Where AI use is permitted and disclosure is required, apply the appropriate citation format (APA, MLA, or Chicago as specified for your assignment) plus any additional disclosure your institution requires — which may include a statement in a cover sheet, a note in the paper, or a separate AI Use Declaration form. Some institutions have developed standardised AI declaration forms that supplement but do not replace the formal citation.
Step 5 — Keep Records After Submission
After submission, retain your ChatGPT conversation logs, your citation documentation, and any instructor communication about AI use for the duration of your course (and beyond — academic integrity investigations can be initiated weeks or months after submission). If your use was properly permitted and disclosed, this record protects you. If you receive any query about AI use, having this documentation makes a significant difference.
For students at institutions with particularly detailed or recently updated AI policies — which now includes most major universities — the academic integrity resource provides context for understanding how these policies apply to different types of academic work. The intersection of AI citation and academic integrity is one of the fastest-moving areas of academic policy, and staying current with your institution’s specific position is a practical necessity.
Special Cases — Shared ChatGPT Links, API Outputs, ChatGPT Plugins, and Custom GPTs
Beyond standard ChatGPT conversations, several specific use contexts create citation complications that the standard format templates do not cover directly. Shared conversation links, API-generated outputs, responses generated through third-party plugins or integrations, and custom GPTs all warrant specific consideration.
When a ChatGPT Conversation Has a Public URL
ChatGPT allows users to generate shareable links to specific conversations. When such a link exists and is included in your citation, it replaces the generic https://chat.openai.com as the URL. Include it as: https://chat.openai.com/share/[conversation-id]. Note that these links can be deactivated by the user who created them, making them less stable than a standard URL — consider including the conversation in an appendix regardless.
APA: OpenAI. (2024, March 3). ChatGPT (GPT-4o) [Large language model]. https://chat.openai.com/share/[id]
MLA and Chicago: include the specific URL in place of the generic platform URL.
ChatGPT Output Generated via the OpenAI API
If you accessed GPT-4o or GPT-3.5 through the OpenAI API rather than through the chat interface, the citation is substantively the same — the developer is OpenAI, the model is specified by the API model parameter you used, and the date is the date of the API call. Include the model string as the version identifier: GPT-4o, gpt-4-turbo-preview, gpt-3.5-turbo, etc. Note in a disclosure that the output was generated programmatically via API call rather than through the standard chat interface.
Tool-Augmented and Custom-Instructed Outputs
Custom GPTs and plugin-augmented ChatGPT sessions may produce outputs meaningfully shaped by additional tools, instructions, or retrieved data — not purely by the base language model. For custom GPTs, include the custom GPT name in the citation: ChatGPT (GPT-4o, [Custom GPT name]). For plugin-augmented responses, note the plugin used in your citation or appendix, since the plugin’s data retrieval may affect the content of the response in ways that the base model alone would not.
When GPT Powers a Third-Party Tool
Many third-party academic tools — Grammarly, Notion AI, Jasper, and others — are built on OpenAI’s API and use GPT under the hood. When citing output from such tools, cite the third-party tool (Grammarly, Notion, etc.) as the tool you used, not OpenAI or ChatGPT — the user interface and any platform-specific fine-tuning make the third-party tool the more accurate citation subject. Developer is the third-party company, not OpenAI.
Quick-Reference Format Comparison Tables — APA, MLA, and Chicago Side by Side
For rapid reference during the writing process, the tables below compare the three citation systems element by element. Use these alongside the full format guidance above — they are meant as a memory aid rather than a replacement for the detailed examples and context provided in each section.
| Citation Element | APA 7th Edition | MLA 9th Edition | Chicago 17th Edition |
|---|---|---|---|
| Author / Creator | OpenAI (organisation name) | No author — prompt title first | ChatGPT or OpenAI (varies by format) |
| Date | Year, Month Day of conversation | Day Month Year of conversation | Month Day, Year of conversation |
| Title | ChatGPT (version) [Large language model] | “Prompt text” in quotation marks | ChatGPT (version) — no quotes, no italics in footnote |
| Publisher | Not listed separately (OpenAI is author) | OpenAI (after ChatGPT tool name) | OpenAI |
| URL | https://chat.openai.com | chat.openai.com (no https in MLA) | https://chat.openai.com |
| In-text format | (OpenAI, 2024) | (“Abbreviated prompt”) | Footnote number¹ or (OpenAI 2024) |
| Version specification | In parentheses after ChatGPT: (GPT-4o) | After comma following ChatGPT: GPT-4o | In parentheses after ChatGPT: (GPT-4o) |
| Prompt inclusion | Appendix (recommended, sometimes required) | In Works Cited entry as “title” | In footnote (brief description or full prompt) |
| Multiple AI conversations | Suffix: OpenAI (2024a), (2024b) | Different prompt titles distinguish entries | Different footnote numbers; bibliography distinguishes by date or prompt |
| Disclosure requirement | Citation + appendix | Citation + transparency note | Citation + optional explanatory note in footnote |
| Scenario | APA Example | MLA Example | Chicago Example |
|---|---|---|---|
| GPT-4o, November 2024 | OpenAI. (2024, Nov. 10). ChatGPT (GPT-4o) [Large language model]. https://chat.openai.com |
"Prompt text." ChatGPT, GPT-4o, OpenAI, 10 Nov. 2024, chat.openai.com. |
ChatGPT (GPT-4o), response to "[prompt]," OpenAI, November 10, 2024, https://chat.openai.com. |
| Version unknown | OpenAI. (2024, June 1). ChatGPT [Large language model]. https://chat.openai.com |
"Prompt text." ChatGPT, OpenAI, 1 June 2024, chat.openai.com. |
ChatGPT, response to "[prompt]," OpenAI, June 1, 2024, https://chat.openai.com. |
| Google Gemini | Google. (2024, Aug. 15). Gemini (Gemini 1.5 Pro) [Large language model]. https://gemini.google.com |
"Prompt." Gemini, Gemini 1.5 Pro, Google, 15 Aug. 2024, gemini.google.com. |
Gemini (Gemini 1.5 Pro), response to "[prompt]," Google, August 15, 2024, https://gemini.google.com. |
| Anthropic Claude | Anthropic. (2024, Oct. 3). Claude (claude-3-5-sonnet) [Large language model]. https://claude.ai |
"Prompt." Claude, claude-3-5-sonnet, Anthropic, 3 Oct. 2024, claude.ai. |
Claude (claude-3-5-sonnet), response to "[prompt]," Anthropic, October 3, 2024, https://claude.ai. |
Common Formatting Errors That Undermine AI Citations — and How to Avoid Each One
Getting the substance of an AI citation right — the correct date, the correct version, the correct author — is necessary but not sufficient. Formatting errors that are consistent across all citation types can undermine an otherwise correct citation. The errors below are the most frequently observed in student work and are straightforward to prevent once identified.
Using “ChatGPT” as the Author Instead of “OpenAI”
In APA and Chicago, the author is the organisation that created and operates the tool — OpenAI — not the tool’s name. The tool name appears in the title position. “ChatGPT” as author is incorrect in both formats. In MLA, where the “author” slot is occupied by the prompt title, OpenAI appears later as the publisher — again, “ChatGPT” does not appear in the author position.
Using the Release Date Rather Than the Conversation Date
ChatGPT was publicly released on November 30, 2022. A surprisingly common error is using this date — or November 2022, or 2023 as “the year ChatGPT launched” — instead of the actual date of your specific conversation. The date in all three citation formats refers to when you generated the output, not when the model was released. This matters because the model changes over time, and the conversation date is part of the citation’s contextual transparency.
Omitting the [Large language model] Descriptor in APA
APA citation format requires a bracketed descriptor for non-standard source types — just as [Video], [Podcast episode], or [Dataset] appear in other APA references. For ChatGPT, the correct descriptor is [Large language model]. Omitting this leaves the citation without its type identification, which APA requires. Some students substitute [AI], [Chatbot], or [Software] — these are not incorrect, but [Large language model] is the APA-preferred descriptor per the official guidance.
Incorrectly Italicising in MLA
In MLA Works Cited, ChatGPT (the tool/container name) is italicised, but the prompt title in quotation marks is not italicised. A common error is italicising the prompt text or placing both in quotation marks. The formatting convention mirrors the MLA treatment of an article (in quotation marks) published in a journal (italicised) — here, the prompt is the “article” and ChatGPT is the “journal/container.”
Not Specifying the Version When It Is Available
All three style guides indicate that the model version should be included when known. Students often omit it either because they did not record it at the time or because they assume it is optional. When you can identify the version (and you usually can, either from the interface or by asking the model directly), include it. A citation without a version identifier is technically acceptable but less transparent than one that includes it — and transparency is the point of the citation.
Treating AI Citation as Optional When AI Output Appears in the Paper
Some students cite AI in a footnote or acknowledgement but omit it from the reference list or Works Cited, treating it as a process tool rather than a source. If AI-generated content — even paraphrased — appears in your paper, the citation belongs in the reference list/Works Cited/bibliography. The in-text citation alone, or a footnote alone, is not a complete citation in APA or MLA. Chicago Notes-Bibliography requires both the footnote and the bibliography entry.
Using a Specific Conversation URL Without Verifying Its Permanence
ChatGPT shared conversation links can be deactivated. If you include a specific conversation URL (https://chat.openai.com/share/[id]) in your citation, verify it is still active at the time of submission and note in your appendix that the link may not remain active permanently. For submissions where permanence matters — dissertations, published research — the generic platform URL plus a full appendix is more reliable than a specific conversation link.
Integrating AI Citations into a Complete Reference List — Format, Order, and Hanging Indent
Beyond the content of the citation itself, the mechanics of how AI citations integrate into your reference list, Works Cited, or bibliography are important for formatting correctness. These mechanical details are the same for AI citations as for any other source type within each style — but students sometimes format AI citations differently from the rest of their reference list, either because they treated it as a special case or because they copied a format from the web that differs from their institution’s requirements.
APA Reference List — Formatting Rules for AI Citations
APA reference lists use a hanging indent: the first line of each entry is flush left, and subsequent lines are indented 0.5 inches. This applies to ChatGPT citations exactly as it does to journal article or book citations. Entries are alphabetised by author’s last name or, for organisations, the first significant word of the organisation name — so OpenAI citations are alphabetised under “O”. The reference list is double-spaced throughout, with no additional space between entries (the double spacing provides sufficient visual separation). “References” is the centred heading, not “Bibliography” or “Works Cited.”
If you have multiple OpenAI citations in the same reference list — for example, one GPT-4o conversation and one GPT-3.5 conversation — order them chronologically within the same author: OpenAI (2023, September 4) comes before OpenAI (2024, March 12). Where two conversations share the same date, use the (a), (b) suffix system in both the reference list entry and the in-text citation.
Where Does the ChatGPT Entry Fall in the Alphabetical List?
MLA Works Cited is alphabetised by the first word of each entry. For ChatGPT citations, the first word is the first word of the prompt title (in quotation marks) — not OpenAI, not ChatGPT. So a prompt beginning “Analyse the…” is alphabetised under “A”. A prompt beginning “Summarise…” is alphabetised under “S”. This differs from how most sources are alphabetised (by author surname) and can make ChatGPT entries appear in unexpected positions within a Works Cited list.
Works Cited also uses hanging indent — flush first line, indented continuation lines. “Works Cited” is the centred heading. Double-spacing throughout.
Footnotes Are Indented; Bibliography Entries Use Hanging Indent
Chicago footnotes are formatted with the first line indented and subsequent lines flush left — the reverse of the hanging indent used in bibliography entries and in APA/MLA reference lists. The Chicago bibliography uses the hanging indent (flush first line, indented continuation). Both use single spacing within entries and double spacing between them in most academic submission formats. “Bibliography” or “Works Cited” (for Author-Date) is the standard heading; “References” is also acceptable in Author-Date format.
Distinct AI citation frameworks — APA, MLA, and Chicago each require different formatting, different primary elements, and different disclosure practices
The format of your citation depends entirely on which style your assignment requires — not on any universal AI citation standard. APA emphasises the developer-organisation as author. MLA centres the prompt as the primary identifier. Chicago provides the most narrative flexibility through its footnote system. Knowing which style applies to your assignment is the first step; the format follows from that.
Citation Errors Shouldn’t Cost You Marks on Good Research
Whether it’s getting the AI citation format right, building a complete bibliography, or ensuring your paper meets your institution’s academic integrity requirements — our team of subject specialists and citation experts can review, correct, and advise on citation formatting across APA, MLA, Chicago, Harvard, and Vancouver.
Discipline-Specific Guidance — Which Format Is Used Where, and What Your Department Expects
The citation format required for your assignment is determined by your discipline and institution, not by your choice or by any universal standard. Knowing which format is standard in your field prevents the confusion of applying the wrong style’s rules. The overview below covers the most common discipline-format pairings in higher education and the specific AI citation implications for each.
Default citation style by academic discipline — percentages indicate approximate prevalence within each field. Always confirm the required style with your specific institution and course assignment brief; these are disciplinary norms, not universal rules.
Law deserves special attention in the context of AI citation because legal citation uses specialist systems — OSCOLA (UK), Bluebook (US), AGLC (Australia) — rather than APA, MLA, or Chicago. None of these legal citation systems had developed comprehensive AI citation guidance by early 2025. The most widely adopted approach is to cite AI tools in law assignments using either a footnote description in OSCOLA style or an analogous format that includes the key elements: developer, tool name, version, date of conversation, and URL. Where ChatGPT is used in legal research or writing, many law faculties additionally require explicit disclosure to the instructor regardless of the citation format used. For detailed guidance on law-specific citation and academic work, the law assignment help and legal writing services resources cover discipline-specific requirements in more detail.
How AI Citation Fits Into a Research Writing Workflow — From Draft to Submission
The practical integration of AI citation into a complete academic writing workflow is worth tracing from beginning to end, because citation errors most often result from inadequate record-keeping at the point of AI use rather than misunderstanding of the format rules at the point of citation. Knowing the format is necessary but not sufficient — you also need a workflow that generates the information the citation requires.
Before Using AI
Check your syllabus and institutional policy. Confirm AI use is permitted for your specific task. If uncertain, email your instructor. Create a citation log document before opening ChatGPT.
During AI Use
Record: date, version (check interface), full prompt(s), key response content. Copy complete exchanges to your log. Note what you will use and how — for drafting, editing, brainstorming, etc.
While Writing
Integrate AI output with clear attribution (in-text citation) wherever AI content appears. Note in your writing whether you are quoting or paraphrasing AI output. Prepare appendix content from your log.
Before Submitting
Verify all AI citations against the correct format (APA/MLA/Chicago). Check version identifiers. Confirm Works Cited / Reference List entries are complete. Add appendix and any required disclosure statement.
For students using AI tools as part of a broader research process — searching for sources, synthesising literature, structuring arguments — the citation question intersects with the larger question of how to use AI responsibly in academic work. The ethical use of AI in academic settings resource addresses the wider framework, while this guide focuses specifically on the citation mechanics. For guidance on academic research skills more broadly, including how to find, evaluate, and cite peer-reviewed sources alongside or instead of AI-generated content, the research paper writing services and literature review writing services provide expert support at every level of the research process.
Frequently Asked Questions About Citing ChatGPT
In APA 7th edition, the reference list entry is: OpenAI. (Year, Month Day). ChatGPT (version) [Large language model]. https://chat.openai.com
For example: OpenAI. (2024, October 14). ChatGPT (GPT-4o) [Large language model]. https://chat.openai.com
The in-text citation is (OpenAI, 2024) for a parenthetical citation or OpenAI (2024) for a narrative citation. No page number is needed. APA recommends including the full prompt and response in an appendix. If the version is unknown, omit it from the parentheses. The date is the date of your specific conversation, not ChatGPT’s public release date. This format comes from the official APA Style Blog guidance on citing ChatGPT.
MLA 9th edition Works Cited format: "Abbreviated prompt text." ChatGPT, GPT-4o, OpenAI, day month year, chat.openai.com.
For example: "Explain the causes of World War I in 200 words." ChatGPT, GPT-4o, OpenAI, 14 Oct. 2024, chat.openai.com.
The in-text citation uses the abbreviated prompt in quotation marks: ("Explain the causes of World War I"). MLA italicises ChatGPT as the container (tool name), places the prompt in quotation marks as the title, lists OpenAI as the publisher after the version, and uses MLA’s day-month-year date format. Official guidance is available through the MLA Style Center guidance on citing generative AI.
Chicago Notes-Bibliography footnote: ChatGPT (GPT-4o), response to "[your prompt]," OpenAI, Month Day, Year, https://chat.openai.com.
Chicago bibliography entry: OpenAI. ChatGPT (GPT-4o). "Prompt description." Accessed Month Day, Year. https://chat.openai.com.
Chicago Author-Date in-text: (OpenAI 2024) — no comma between author and year in Chicago style (unlike APA). The reference list entry is: OpenAI. 2024. ChatGPT (GPT-4o). Accessed [date]. https://chat.openai.com. Chicago’s flexibility in the footnote system allows including a brief description of how the AI was used within the note itself.
OpenAI. (Year, Month Day). ChatGPT (version) [Large language model]. https://chat.openai.com/share/[conversation-id]. Note that shared links can be deactivated by the creator — they are less stable than permanent archived sources. In your citation or a note, indicate that the conversation was shared by another user, not conducted by you. If you conducted the conversation yourself, include the full exchange in an appendix rather than relying on the shared link’s permanence.(OpenAI, 2024a) and (OpenAI, 2024b) — with the corresponding suffix in the reference list entries. In MLA: different prompt texts in the Works Cited entries naturally distinguish multiple AI conversations; in-text, use the abbreviated prompt that matches the relevant Works Cited entry. In Chicago: each conversation gets its own footnote; multiple conversations from the same session can reference the same footnote with “ibid.” if consecutive, or use a shortened citation form. In all three styles, the appendix should separately label and present each distinct conversation if multiple exchanges are included.Extend your writing and citation skills: citation and referencing guide · research paper writing · proofreading and editing · literature reviews · ethical AI use in university settings · academic integrity policy guidance · paper formatting service · dissertation support · critical analysis papers · law assignment help · nursing assignment help · academic writing services