How to Cite AI in Legal Writing

Download This Article as a PDF pdf download
Loading ...
How to Cite Exhibits in a Legal Brief​

Contents: AI for Law Firms: A Comprehensive Guide

Not all AI is built for law

Get the guide

To cite AI in legal writing, first determine whether the AI output itself needs to be cited or whether you simply need to disclose that AI assisted your work. Then follow the citation rules that apply to the document and preserve enough information for a reader to identify the output.

AI is now a routine part of legal work, but citing it isn’t as straightforward as citing a case or statute. There’s no single citation format that applies to every AI-generated response, and the right approach depends on why the output appears in your work.

For example, you may need to cite an AI response because you’re quoting it, analyzing its accuracy, or showing what information someone relied on. That’s different from disclosing that an AI for law firms tool helped you research or draft a document.

This guide explains when lawyers should cite AI, what information an AI citation for lawyers should contain, and how current Bluebook and APA approaches differ. It also covers the risks of relying on AI-generated legal material and why every underlying authority still needs to be verified.

Key takeaways

  • Citing AI and disclosing AI use are different. A citation points to a specific AI output, whereas a disclosure explains that AI assisted the work.
  • There’s no single citation format for AI-generated content in legal writing. Follow the rules that govern the document, whether those come from a court, citation manual, journal, employer, or client.
  • A useful AI citation should make the output identifiable and reviewable. Record the tool, date, and provide enough context to help readers understand what generated the response, including the prompt where relevant.
  • Clio’s 2026 Legal Trends for Solo and Small Law Firms report found that 71% of solo practitioners and 75% of small firms use AI. As adoption grows, firms need clear rules for verification and responsible use.
  • Verdict: Cite AI when the output itself matters to the analysis. Disclose AI-assisted work when the applicable rules require it, and always verify the underlying legal authority before relying on it.

What are legal citations?

Legal citations identify the sources that support statements in legal documents such as briefs, motions, and judicial opinions. A complete citation allows the reader to locate the source and assess whether it supports the proposition being made.

The exact format depends on the citation style and type of source. For a reported case, a citation commonly includes:

  • Names: The names of the lead parties involved in the case
  • Volume: The reporter volume number
  • Reporter: The name of the reporter
  • Page number: The first page of the decision
  • Court: The relevant court abbreviation
  • Date: The year the case was decided

For example, the Library of Congress gives the following case citation format:

Stearns v. Ticketmaster Corp., 655 F.3d 1013 (9th Cir. 2011)

The general structure is:

Appellant Name v. Appellee Name, [Volume Number] [Reporter Name and Series] [Page Number] ([Court] [Year])

Do lawyers need to cite AI?

Lawyers should cite AI output when the response itself is relevant to the point being made. That may be because the output is being quoted, examined, or used to show what information someone relied on.

For example, an AI response may be relevant when a lawyer is analyzing model bias or showing how a client was influenced by generated information. In those situations, the AI output is part of the evidence or discussion, so the citation helps the reader identify and assess it.

AI shouldn’t be cited as legal authority when the underlying source is available. If ChatGPT or Claude refer to a case or statute, retrieve the original source and cite that instead. Lawyers should also follow any AI-specific rules imposed by the relevant court, publication, employer, or client.

How are lawyers using AI for legal writing in 2026?

Lawyers are using AI tools across the legal-writing process, from early research and matter analysis through to drafting and citation checking. AI can speed up this work, but lawyers remain responsible for the accuracy of the final document and the legal judgment behind it.

Clio’s 2026 Legal Trends for Mid-Sized Law Firms report found that 86% of mid-sized firms had adopted AI. As use becomes more routine, firms need clear processes for reviewing AI-assisted work and deciding when disclosure or citation is required.

Here are a few specific ways that lawyers are using AI for legal writing:

Drafting and revision

Lawyers can use AI to create an initial draft, improve an existing one, generate or edit templates, and more. The output should still be reviewed against the matter record and applicable law before it is used. This is why using a jurisdiction- and matter-aware legal AI tool that has access to your documents and their context can be a game-changer for saving time.

Research and matter analysis

AI can help lawyers identify relevant issues, review large matter files, and find authority more quickly. It can also organize the facts of a case so lawyers have a clearer starting point for their analysis.

Citation support

AI can suggest citation formats or flag missing information, but generated citations shouldn’t be accepted at face value. Lawyers should open the underlying source and confirm that it is current and supports the proposition being made.

Clio Work brings these parts of the workflow together. It can analyze matter documents, research relevant law, and support drafting in context. Every research output links back to its source, while Clio’s built-in citator checks whether cited authority is still current. Lawyers can then review the underlying material before relying on it in their work.

What are the main risks of using AI for legal writing?

The main risks are inaccurate legal information, confidentiality problems, and overreliance on generated output. Tools such as ChatGPT and Claude can support legal writing, but lawyers still need to verify the law and remain responsible for the final work.

AI can generate inaccurate or fabricated legal information

General-purpose AI tools can produce convincing answers that contain mistakes. That can include invented cases, incorrect quotations, or authorities that do not support the proposition being made.

Any case, statute, or other authority suggested by AI should therefore be checked against a trusted legal source before it appears in legal writing.

Information might not be up to date 

There’s no longer one meaningful “knowledge cutoff” that applies to ChatGPT as a whole, and the same principle applies to Claude. Capabilities vary by model and configuration, while some tools can retrieve more up-to-date information via web searches or connected sources.

Plus, even if the information it pulls is from a recent source, this doesn’t guarantee it’s authoritative or correct. Lawyers should open the source themselves and confirm that the law is still valid.

Confidential information needs careful handling

Entering client or matter information into a general-purpose consumer AI tool can create confidentiality and data-handling risks. Before using ChatGPT, Claude, or another AI platform for legal work, lawyers should understand how the provider handles submitted information and whether the firm has approved the tool for that purpose.

Legal-specific AI tools provide additional safeguards, but lawyers still need to follow their professional obligations and internal policies.

See how Clio protects firm and client data across its AI features.

AI output can reflect bias

AI systems learn patterns from large datasets, which means their output can reproduce bias or make assumptions that are inappropriate for the matter. Lawyers should treat generated analysis critically rather than assuming that a confident answer is neutral.

This becomes especially important when an AI response informs legal strategy or an assessment of a client’s position.

Intellectual-property questions remain unsettled

Using AI-generated material can also raise questions about copyright and attribution. Contractual rights granted by an AI provider do not necessarily resolve whether a particular output is copyrightable or whether it resembles protected material.

Lawyers should consider the applicable law and the provider’s current terms when AI-generated text forms a meaningful part of the final work.

Professional duties still apply

Using AI does not change a lawyer’s existing professional obligations. ABA Formal Opinion 512 makes clear that duties such as competence and confidentiality continue to apply when lawyers use generative AI, alongside any additional requirements imposed by courts or local rules.

These reasons are why AI should support legal writing rather than replacing lawyers’ review. As a lawyer, you remain responsible for whatever is ultimately submitted, sent to a client, or relied upon in a matter.

Which citation styles apply to AI in legal writing?

There’s no single citation style that governs AI-generated content in every legal document. The correct approach depends on the court, publication, institution, or other rules that apply to the work.

For legal writing in the U.S. and Canada, common citation systems include:

  • The Bluebook: Widely used for court filings, law reviews, and other legal writing.
  • ALWD Guide to Legal Citation: Another legal-specific citation manual used in legal education and practice.
  • Court-specific citation rules: Some courts publish their own requirements, which take precedence when they apply.
  • The Indigo Book: A free legal citation guide that follows many familiar legal citation conventions.
  • The Canadian Guide to Uniform Legal Citation: Commonly used for legal citation in Canada.

General academic styles can also apply when legal writing is produced outside a traditional legal context. These include APA, MLA, and Chicago style.

For AI-generated material, always check the latest guidance for the citation system that you’re using. AI citation practices are still evolving, so a court or journal may impose additional requirements beyond the general style manual.

Not all AI is built for law

General-purpose tools weren't designed for confidential client work or ethics rules. Get the free guide to see what to look for in legal AI you can actually trust.

Get the guide

What information should an AI citation include?

An AI citation should include enough information for the reader to identify the tool and understand the output being cited.

Depending on the citation style and context, that may include:

Element What to record Example
Tool and provider Name of the tool and the company behind it ChatGPT by OpenAI; Claude by Anthropic
Output or chat title The title of the conversation, if one exists “Intellectual Property Case Research”
Date generated Date the response was produced August 25, 2026
Model or version Record when visible and relevant GPT-5.1; Claude Opus 4.5
Prompt The wording used, when it helps the reader evaluate the output “Summarize recent U.S. cases concerning IP and generative AI.”
Stable link or record Share link, archived transcript, appendix, or exhibit Share URL or Exhibit A

If the AI output can’t be shared because it contains confidential information, preserve an appropriate internal record instead and follow the rules that apply to the matter.

How do you cite ChatGPT in Bluebook format?

There’s no single Bluebook format that applies to every ChatGPT citation. Therefore, a practical approach is to treat the output as an online source and include enough information for the reader to identify the response and understand how it was generated.

A potential citation could include:

Provider, Title of Chat or Output, Product Name (full date generated), stable URL or archived record (explanatory parenthetical describing the prompt).

For example:

OpenAI, Intellectual Property Case Research, ChatGPT (August 25, 2026), [stable share URL or archived transcript] (responding to the prompt “Summarize recent U.S. cases concerning intellectual property and generative AI.”).

Including the prompt can be especially useful when the wording or reliability of the AI response is part of the discussion. Georgetown Law Journal guidance similarly recommends preserving the prompt when it helps readers understand the context of the output.

If the conversation can’t be shared, preserve the output in an appendix, exhibit, or other appropriate record where the applicable rules permit. A generic ChatGPT homepage is less useful because it does not allow the reader to retrieve the specific conversation being cited.

Before filing or publishing, check the citation against the current Bluebook edition and any rules imposed by the relevant court or journal.

How do you cite ChatGPT in APA format?

APA treats ChatGPT differently depending on whether you’re citing the AI tool itself or a specific, retrievable conversation.

For the tool itself, APA’s current reference structure is:

OpenAI. (Year). ChatGPT [Large language model]. URL

For example:

OpenAI. (2026). ChatGPT [Large language model]. https://chatgpt.com/

However, if you’re citing a specific ChatGPT conversation that readers can access, APA uses a different format that includes the date, title of the chat, product name, and share URL.

A shared chat citation can follow this structure:

Author. (Year, Month Day). Title of chat [Generative AI chat]. ChatGPT. URL

Use the exact title and date of the conversation, and only provide a share link if doing so doesn’t expose confidential or privileged information.

APA updated its guidance for generative AI references in 2025, so avoid older formats that describe ChatGPT using wording such as “Nov 3 version.” Check the current APA guidance before publishing or submitting the work, particularly if you’re citing a specific AI conversation rather than the tool generally.

How should lawyers handle AI citations in 2026?

Lawyers should treat AI citation as part of a broader verification and governance process. First decide whether the AI output itself belongs in the work. Then follow the controlling citation and disclosure rules, preserve a reviewable record, protect confidential information, and verify every underlying authority.

Using a legal-specific platform makes verification easier by connecting the output to the matter and the law. Clio Work provides cited legal research, checks whether authority is current, analyzes matter files, and supports drafting in context. Lawyers remain responsible for reviewing the sources, reasoning, and final document.

Book a demo to see Clio Work in action.

What's the difference between citing AI and disclosing the use of AI?

Citing AI is different from disclosing it. A citation refers to the factual content or authority presented by an AI tool’s output, used in cases like referencing information a client relied on, demonstrating bias in LLM output, or using AI content as an example. Disclosure is about declaring that AI was used in the drafting process.

How do you cite ChatGPT in Bluebook format?

A potential Bluebook citation includes the author, chat title, main page title, date generated, URL, and an explanatory parenthetical with the prompt. Example: OpenAI, Intellectual Property Case Research, ChatGPT (November 3, 2023), https://chat.openai.com/ (responding to the prompt “Can you provide summaries of recent cases related to intellectual property and AI in the United States?”).

How do you cite ChatGPT in APA format?

A potential APA citation follows the format: OpenAI. (Year). ChatGPT (Month Day version) [Large language model]. https://chat.openai.com. For example: OpenAI. (2023). ChatGPT (Nov 3 version) [Large language model]. https://chat.openai.com.

What information should a ChatGPT citation include?

A ChatGPT citation should include the author (typically OpenAI), the chat or output title, the date the text was generated, the URL, and ideally an explanatory parenthetical containing the prompt that generated the content. Including the prompt is particularly useful when the validity of the AI content is part of the discussion.

Should the AI prompt be included in the citation?

Yes, when possible. Best practice is to include the prompt that generated the content in an explanatory parenthetical. Doing so is especially important when the validity of the AI content is a point of contention in the discussion, since it gives the reader context to evaluate the output’s reliability and origin.

What are the main risks of using ChatGPT for legal writing?

There are four key risks: a limited scope (ChatGPT’s training data has a cutoff date, so it can’t offer fully current information), lack of reliability (no guarantee that generated facts or cases are real), unresolved copyright infringement concerns (especially around ownership of output), and ethical concerns related to client confidentiality and bias.

Can ChatGPT generate accurate legal citations on its own?

No, not reliably. While ChatGPT can draft legal citations, it lacks proper text formatting capabilities and can sometimes generate inaccurate or fabricated citations. Any AI-generated citations need to be fact-checked and formatted manually before they’re relied upon in legal writing.

Not all AI is built for law

General-purpose tools weren't designed for confidential client work or ethics rules. Get the free guide to see what to look for in legal AI you can actually trust.

Get the guide