How to Cite-Check an Appellate Brief with ChatGPT or Claude

Yes, you can use ChatGPT or Claude to cite-check an appellate brief. Get the full text of each cited opinion, give the agent the brief and one opinion at a time, and ask it to check how the brief uses that authority. Keep the results in a spreadsheet so the review can happen while you draft, not just on filing night. You can do this with an ordinary AI assistant—no special legal AI tool is required.

Here is the step-by-step video. You can also watch it on YouTube.

Set up a working folder

Keep the current brief, downloaded opinions, and cite-check spreadsheet together in one matter folder. Point Claude Cowork, ChatGPT Work, or another file-capable agent at that folder. Treat it as your project: when you start a fresh chat for the next opinion, the agent can find the same brief, source files, and review sheet. Make sure it uses the current brief rather than an earlier draft.

1. Find the opinions

Start with the cases cited in the brief, including citations in the body that may not appear in the table of authorities. You need the full opinion, not just a search-result snippet or an AI summary.

Web search is surprisingly good at finding opinions. A free connector such as CourtListener can retrieve them for your agent; a paid research connection such as CoCounsel can work too. You can also download opinions from Westlaw or simply copy and paste their text into the chat. The important question is whether the agent actually has the complete, correct opinion in front of it. If you cannot find an opinion, flag it for manual investigation rather than treating a failed search as proof that the case does not exist.

2. Review one opinion per fresh chat

For each opinion, open a new chat and give the agent only the brief and that opinion. Ask it to identify every use of the case in the brief and compare those uses with the source. This keeps the task focused and lets you see exactly what the agent read. The folder can contain all your sources, but tell the agent to read only the current brief and its assigned opinion for this review. You can inspect and change the prompt, the source text, and the model rather than relying on an opaque report.

For example:

Read the attached appellate brief and the full text of [case]. Find every place the brief cites this case. Check whether any quotations are accurate, whether the opinion supports the propositions attributed to it, and whether the cited passages and pincites match the source provided. For each issue, identify the location in the brief and the relevant language in the opinion. Tell me what you could not verify; do not guess.

Change those instructions to match what you care about. You might focus on a critical quotation, the court's actual holding, the procedural posture, or a particular pincite. If the source you supplied does not show reporter pagination, the agent cannot verify a reporter pincite from that text alone. Review the underlying sources and resolve flagged questions yourself.

This is a check of how the brief uses the opinion, not a good-law check. Confirm later treatment separately, using an appropriate citator or your own research.

3. Keep a cite-check spreadsheet as you draft

Have ChatGPT or Claude create the spreadsheet in your working folder. For example:

Create a cite-check spreadsheet for the current brief. Give each authority its own row, including authorities cited in the body but missing from the table of authorities. Include the citation, source file or link, review status, questions to resolve, reviewer notes, and who approved it. Do not mark anything as checked yet. Save the spreadsheet in this folder.

After reviewing an opinion in its own chat, ask: “Update the cite-check spreadsheet with this review. Record what you checked, any unresolved issues, and the source you used. Do not mark it approved on my behalf.” A paralegal or attorney can add their own findings, including currentness checks done in Westlaw or Lexis.

The sheet is bookkeeping, not a certificate of correctness. It helps you see what has been checked, what needs another look, and what to revisit after you change an argument. That makes cite checking an ongoing part of drafting instead of a frantic final pass through dozens of authorities.

Want the workflow ready to use?

The DIY method works. But you do not have to build the process, prompts, and spreadsheet management yourself. With DocketDrafter, tell Claude or ChatGPT: “Use DocketDrafter to cite check this brief.” The opinion-by-opinion review workflow and cite-check bookkeeping described above are already built in and maintained by us. You still choose the sources, direct the review, resolve questions, and approve the brief.