Using AI to elevate legal-writing advocacy

Lawyers who ignore AI risk becoming the Blockbuster Video of our profession. Here are practical applications, examples, and prompts for today’s litigators

  Artificial intelligence (“AI”) – particularly large language models (LLMs) such as ChatGPT – has rapidly shifted from novelty to necessity for litigators. Though the potential uses are infinite, this article focuses on one core area: How litigators can collaborate with LLMs to sharpen legal writing and analysis in a practical, everyday way.

  Lawyers who ignore AI risk becoming the Blockbuster Video of our profession – stuck in an older model while everyone else moves on. I am not a computer scientist, and I am not writing from the vantage point of a technologist. I am a lawyer who writes for a living and has discovered the real, concrete benefits of using LLMs as a writing partner with unusual stamina, fast pattern-recognition, and a tireless willingness to suggest, reorganize, and revise.

From rules-based tools to language models: What lawyers are actually using

  AI can feel intimidating when described in computer-science terms – neural nets, embeddings, tokens, vector stores, supervised fine-tuning. But the tools litigators interact with every day – ChatGPT, CoCounsel, Lexis+ AI, Clearbrief, Claude, and others – do not require a technical background to use effectively. What matters is understanding, at a high level, what these systems do and how they differ from the rules-based software we’ve used for decades.  Traditional legal technology largely followed “if/then” conditional rules: If the user selects “contract,” show these fields; if the user inputs a citation, retrieve the matching case. This is rules-based computing – software that behaves like a clerk executing a checklist. It cannot infer, improvise, or reason. Document-assembly tools, court-rule calculators, and basic e-discovery filters all fall into this category.  LLMs are categorically different. Instead of following hard-coded rules, an LLM models patterns in language – how lawyers argue, how courts write, how statutes and rules are structured, how sentences logically follow one another. The model does not “know” law the way a lawyer does. But it is very good at predicting plausible continuations of text, restructuring prose, synthesizing material, detecting inconsistencies, and simulating styles of reasoning.  Most lawyers who say they are “using AI” are really using language models. These systems work entirely in words. Tools like ChatGPT, Claude, Gemini, CoCounsel, and Lexis+ AI are all optimized for reading, generating, reorganizing, and critiquing text. They are not robots, decision-makers, or autonomous agents. They do not possess intent, memory, or judgment in any human sense.  That distinction matters. Lawyers are not delegating legal authority to a machine. They are using a sophisticated text engine that can draft paragraphs, reorganize arguments, summarize material, or expose weak reasoning – but that still requires the lawyer to choose the strategy, determine the law, and decide what is persuasive.

Why LLMs work best with lawyers, not instead of them

  One of the most important truths about language models is that they are built to collaborate. Their architecture requires attorney input, correction, and supervision. LLMs are generative tools: They are excellent at offering possibilities – drafts, structures, themes, alternative framings, and anticipated counterarguments – but they are terrible arbiters of truth.

  They cannot determine whether an argument is preserved, whether a case is distinguishable, or whether a citation is binding. Many lawyers have learned this the hard way by failing to cite-check LLM-generated work before filing and discovering that the “cases” supplied by the model simply do not exist. A language model can produce an opinion that looks and sounds like a real case because it is mimicking patterns in judicial prose – not because it has access to an authoritative database.  LLMs also cannot evaluate credibility, weigh evidence, or assess the practical consequences of a particular theory in the hands of a judge or jury. Those are lawyer functions.  This limitation is what makes the tools useful instead of dangerous, as long as we respect it. Because the model does not “decide,” the lawyer remains fully in command. The lawyer sets the strategy, chooses the authorities, determines the narrative, and exercises judgment. The model accelerates the parts of litigation that require volume, speed, and organizational clarity – drafting, editing, synthesizing, reframing, and pressure-testing.  The dynamic is familiar. A senior lawyer supervises junior lawyers who draft, outline, propose ideas, and flag problems, and the senior lawyer then shapes the final product. With AI, the lawyer steps into that supervisory role with respect to the model itself.  For lawyers, whose work revolves around precision and judgment, understanding the difference between rules-based tools and generative AI is more than academic. When AI is seen as a mysterious black box, it is easy to view it with distrust, especially as sanctions orders make headlines. When it is understood as a language model trained to predict and arrange text, the fear diminishes. The tool is not designed to replace lawyering; it is designed to amplify it.  In practical terms, this understanding encourages lawyers to use LLMs the way they use Westlaw, Lexis, or Casetext: as professional instruments that require expertise, oversight, and thoughtful engagement. Used this way, AI frees attorneys from mechanical tasks and allows them to focus on judgment, strategy, persuasion, and storytelling.

Collaborating with LLMs to improve legal writing

  Brief writing – whether at the trial or appellate level – is a high-cognitive-load task. It requires synthesizing facts, law, and strategy; sequencing arguments; anticipating counterarguments; and presenting everything with clarity and restraint. Lawyers must simultaneously compose, edit, organize, fact-check, adjust tone, and police internal consistency. LLMs can help across each of these dimensions. Their value is not in “writing the brief for you,” but in augmenting your ability to write more clearly and efficiently.

Understanding how LLMs learn your writing style and voice

  One subtle but powerful advantage of using LLMs consistently is that, over time, the model begins to recognize patterns in your writing style. Although LLMs do not automatically store your documents unless a particular product is configured to do so, they do adapt within a session – and across repeated use when you feed them samples – to the stylistic cues and preferences you provide.  If you repeatedly edit the model’s drafts, correct its tone, clarify preferred phrasing, or supply examples of past briefs, the LLM uses those interactions to refine its future outputs. Over time, it can approximate the hallmarks of your professional voice: whether you tend toward concise or expansive sentences; whether your openings are thematic or procedural; whether you favor a more formal judicial tone or a narrative, story-driven style.

  This creates practical opportunities. You can ask the model to revise text written by an associate “in my usual style,” and it will attempt to replicate the rhythm and tone it has learned from your prior samples. It can help maintain a consistent voice across a long brief drafted in multiple sessions or by multiple authors, smoothing seams so the document reads as a unified whole.  Sample prompt: “Here are two writing samples from prior briefs. Analyze the tone, structure, and stylistic features. Then revise my draft to match that voice. Explain how your revisions align with the identified style.”  Once the model approximates your style, it can also help refine it. It can identify patterns that weaken clarity – overuse of qualifiers, predictable transitions, long lead-in clauses – and propose alternatives that still feel like “you.” In this way, the LLM becomes a kind of adaptive editor: learning from your preferences while nudging your writing toward greater clarity and impact.

Using LLMs to generate early-stage structures and drafts

  LLMs are particularly useful at the beginning of a writing project, when you know the issues but not yet the architecture of the document. With only a rough description of the dispute and your main themes, a model can produce a draft structure that includes section headings, transitions, and placeholders for legal authority.

  For example, when preparing an opposition to summary judgment, you can input a paragraph describing the movant’s arguments and your planned response. The LLM might produce a draft containing an introduction, a summary of the standard of review, arguments ordered from strongest to weakest, and marked spaces for case citations. You remain responsible for inserting authority, correcting inaccuracies, and shaping the argument. But the tool reduces the time spent building scaffolding.  Sample prompt: “Here is a rough outline of the arguments for the plaintiff’s opposition to summary judgment. Turn this into a structured draft with introductions and transitions. Mark where case law should be added, but do not generate citations.”  This accelerates the early phase without ceding judgment to the model. Many lawyers find that dictating ideas into an LLM and letting it convert that speech into a structured draft is far easier than confronting a blank page.

Improving clarity, brevity, and persuasiveness through targeted editing

  Once a draft exists, lawyers often struggle to edit effectively because they are too close to the text. LLMs can serve as neutral editors that identify verbosity, unclear phrasing, or structural clutter. They can perform multiple rounds of revision instantly, making it easy to experiment with tone and structure.  A fact section that reads like a deposition transcript can be streamlined by instructing the model to preserve all record citations while reducing unnecessary detail. The LLM can reorganize sentences, connect related concepts, and strip out words that do not advance the point.  Sample prompt: “Edit this section to make it more concise and readable, keeping all record citations and factual content intact. Improve flow and eliminate repetitive phrasing.”  The output will still need review, but it often provides a much cleaner baseline. LLMs can also generate several variants of the same paragraph – one more formal, one more conversational, one more concise – allowing you to compare styles and sharpen your editorial instincts.

Spotting inconsistencies, gaps, and structural problems (including reverse outlining)

  Because briefs are typically drafted over days or weeks, inconsistencies are almost inevitable. You might frame an issue one way in the introduction and another way in the argument. A casual statement early in the brief might later conflict with the chosen standard of review. These inconsistencies undermine credibility more than most writers realize.  LLMs are effective at scanning the entire draft and identifying mismatches in reasoning or tone. They can flag contradictory statements, missing logical steps, factual assertions unsupported by citations, or transitions that do not follow from what came before.  Sample prompt: “Read this draft and identify any internal inconsistencies, circular reasoning, unsupported assertions, or unclear transitions. Explain the issue and propose revisions.”  Beyond this global review, LLMs are especially powerful for reverse outlining. Lawyers often outline before writing, but once the drafting process begins, the document can drift away from that original plan. By the time you have a full draft, it may be hard to tell whether the brief still follows a logical, persuasive structure.  A reverse outline is an outline you create after the draft is written; it reveals what you actually wrote, not what you meant to write. When you ask an LLM to generate a reverse outline, it reads the entire document and distills each paragraph or section into a single sentence capturing its main point. This produces a high-level map of your argument as it currently exists.  Patterns and problems become obvious. A section meant to establish the standard of review may slip into argument. A paragraph intended to support a legal proposition may introduce new facts nowhere else discussed. A theme introduced in the opening may disappear until the final page. The outline may reveal that your strongest argument is buried in the middle instead of leading the way.

  Sample prompt: “Produce a reverse outline of this draft by summarizing the main point of each paragraph or section. After outlining it, identify gaps, redundancies, or unclear sequencing. Tell me where the logic weakens or where additional explanation is needed.”  Because the model is not attached to any sentence, it highlights issues that are easy to miss when you have lived with the draft for weeks. Reverse outlining with an LLM also helps diagnose incomplete narratives: If the outline shows a leap from one idea to another, that is often a sign that a premise is missing or that you relied on an assumption not actually expressed on the page.  In effect, the LLM becomes a structural analyst – a role that traditionally required a second lawyer with fresh eyes. Now, you can obtain structural feedback on demand, without waiting for a colleague’s time.

Using LLMs as writing mentors: syntax, style, and judgment

  Perhaps the most underrated use of LLMs in legal writing is as a mentor – a tool that can give real-time feedback on the craft of writing. Even experienced lawyers benefit from reminders about sentence construction, rhythm, emphasis, and clarity.  You can paste in a paragraph and ask the model not just to revise it, but to explain why its version may be stronger.  Sample prompt: “Rewrite this paragraph in a clearer and more persuasive way, and explain why your version is stronger.”  The model might explain that it moved the key point to the beginning of the sentence, converted passive constructions to active where appropriate, removed buried verbs, or used parallel structure to clarify comparisons. This is the kind of granular feedback most lawyers stop receiving early in their careers, if they ever received it at all.  You can also use LLMs to refine specific skills. For issue statements and framing, you might ask the model to rewrite your issue statement several different ways – doctrinal, policy-oriented, narrative, narrow, broad – and explain the persuasive implications of each.  Sample prompt: “Rewrite my issue statement five different ways, each adopting a different persuasive strategy. Explain which version is most consistent with the deferential standard of review.”  To surface unstated premises, you can have the model identify assumptions you are making but not articulating.  Sample prompt: “Identify any assumptions I am making but not stating explicitly. Point out where my argument jumps from one idea to another without explaining the reasoning.”  To enforce paragraph discipline, ask it to find paragraphs that try to do too much and suggest how to divide them.

Sample prompt: “Identify paragraphs where I combine more than one idea or argumentative function. Suggest how to divide or restructure them for clarity.”  For judicial readability and tone, the model can flag sentences that may be hard for a busy judge to process or places where your tone becomes too sharp, too tentative, or inconsistent with the seriousness of the case.  Sample prompts: 

“Identify any parts of this section that may create unnecessary processing difficulty for a judicial reader. Rewrite them for clarity and explain why your revision improves readability.”  “Analyze the tone of this section and identify sentences that conflict with the tone appropriate for judicial writing. Suggest more suitable alternatives.”  And to uncover recurring style habits and tics, you can ask the model to identify repeated phrases, overused qualifiers, or rhythm patterns that make your writing feel monotonous.  Sample prompt: “Identify recurring stylistic habits or patterns in my writing that weaken clarity or variety. Provide examples and suggest alternatives.”  This “mentor mode” is especially valuable for solos and small-firm lawyers who may not have colleagues available to give detailed feedback on writing quality. It offers ongoing professional development in the one skill most central to litigation: clear written advocacy.

Using LLMs to enhance responsive legal writing

  Responsive writing – oppositions, replies, and sur-replies – is particularly challenging because it requires engaging with someone else’s framing. Opposing counsel may distort the facts, misstate the law, or arrange arguments in a way that obscures weaknesses. LLMs can help you quickly understand and organize what you’re responding to and then craft a more focused rebuttal.  You can start by asking the model for a neutral outline of opposing counsel’s brief. That alone clarifies the structure you are facing. From there, the model can identify logical weaknesses, factual overstatements, or gaps in authority and suggest ways to respond.  Sample prompt: “Provide a neutral outline of the arguments in this opposition, then identify weaknesses, mischaracterizations of fact, or unsupported assertions. Explain why each is vulnerable and suggest rebuttal themes.”  It is also helpful to have the model generate the strongest, most coherent version of your opponent’s position – the “steel-man” version – so that your reply addresses what the judge is most likely to see, not a watered-down caricature.  Sample prompt: “Rewrite the opposing argument as persuasively as possible and explain what rhetorical techniques make it stronger.”  Seeing the argument at its best forces you to raise the level of your reply. You can then ask the model to help prioritize which points truly require a response and which are distractions that can be safely ignored without conceding anything of substance.  Finally, LLMs can help you avoid overstepping in reply briefs by flagging arguments that look more like new issues than true rebuttal.  Sample prompt: “Compare my draft reply to the original motion and opposition. Identify any arguments in the reply that a court might view as new rather than responsive, and suggest how to reframe them as proper rebuttal.”  Used this way, the LLM supports a more disciplined, strategic approach to responsive writing rather than encouraging point-by-point trench warfare.

Using LLMs to support fact mastery and record navigation

  Legal writing lives or dies on the facts, yet mastering a complex record is one of the most time-consuming and mentally exhausting tasks in litigation. LLMs can dramatically accelerate this process by helping lawyers extract, organize, and interrogate factual material before they ever begin drafting. When presented with deposition transcripts, hearing transcripts, discovery responses, or document sets, a model can summarize key points, identify contradictions, build chronologies, and highlight testimony relevant to particular legal standards.  This is not a substitute for personally reviewing the record – nothing replaces a lawyer’s judgment – but it allows the lawyer to start with a structured understanding rather than a mountain of unprocessed information. A deposition that might take hours to digest can be distilled into a clear, thematic overview that allows the lawyer to focus attention on nuance rather than initial organization.  LLMs are particularly effective at creating timelines, comparing different witnesses’ versions of events, and flagging factual assertions that appear inconsistent or unsupported. When preparing for summary judgment, the model can map disputed facts to corresponding evidence or testimony. When preparing for trial, it can isolate credibility-related statements or testimony that may open the door to impeachment.  Sample prompt: “Create a timeline of material facts from this transcript. Identify inconsistencies, credibility issues, and testimony that contradicts the opposing party’s narrative. Note any statements relevant to the elements of negligence.”  Used thoughtfully, AI becomes a factual base camp: you begin your writing with clarity, structure, and a sense of where your strongest themes lie.

Using LLMs to strengthen oral argument preparation

  LLMs are not only helpful in writing briefs – they can meaningfully improve oral argument preparation. A model can simulate questioning styles, adopt the tone of a skeptical panel, or reframe your opponent’s position in a way that forces you to articulate sharper answers. This is especially useful when preparing for appellate argument or complex trial-level hearings such as Daubert motions, motions in limine, or dispositive motions.  Because LLMs excel at identifying logical vulnerabilities, they can generate the kinds of “bench-hostile” questions that lawyers may not ask themselves. These questions often target the precise weaknesses judges tend to focus on: preservation, standard of review, remedy, causal links, or doctrinal inconsistencies.  Sample prompt: “Act as a skeptical three-judge appellate panel hearing this case. Ask fifteen rapid-fire questions focusing on my weakest arguments, preservation issues, and remedy. After the questions, identify themes I should emphasize in rebuttal.”  For trial lawyers preparing evidentiary arguments, a model can test whether the logic of an evidentiary objection is internally consistent or whether a different framing might present a cleaner path to exclusion or limitation.  LLMs can also generate succinct, judge-facing versions of your argument – the short, clear formulations that become invaluable when thinking on your feet. In this sense, the model becomes a rehearsal partner: tireless, adaptive, and capable of adopting any posture you need it to.

Using LLMs to help trial lawyers preserve issues more effectively

  One of the most valuable (and

underappreciated) uses of LLMs for trial lawyers is in identifying and preventing issue-preservation problems that later complicate or foreclose appellate review. Many forfeiture problems arise not because trial counsel misunderstood the law, but because the issue was not framed with specificity, not raised at the correct procedural moment, or not supported by an adequate record.  LLMs can assist by clarifying the procedural prerequisites for a particular issue. A lawyer can ask the model to outline the steps required to preserve a constitutional objection, challenge expert methodology, or raise a jury-instruction claim. The model can also help draft concise objections or targeted motions that articulate the basis for the issue cleanly and avoid generic or boilerplate formulations.  Sample prompt: “Explain the steps necessary to preserve a challenge to an expert’s methodology under California law at the trial level. Then draft three sample objections – concise, specific, and grounded in the record.”  LLMs can also analyze upcoming motions to identify where preservation risks lie. A tentative ruling might hint at an evidentiary ruling that requires clarification or offer a rationale that, if not addressed immediately, will be difficult to challenge later. Asking the model to review the ruling and identify potential forfeiture traps can help trial counsel ensure the record is complete.

Illustrative hypotheticals: How lawyers might use LLMs in real practice

  To make these ideas more concrete, consider a few examples drawn from common litigation scenarios.  Scenario 1: Imagine receiving a sprawling, unfocused opposition to your motion at 5:00 p.m. on a Friday. Instead of spending hours sorting through rhetorical flourishes and factual noise, you upload the brief and ask the LLM for a neutral outline of the arguments, followed by a list of weaknesses, contradictions, and misstatements of law. Within minutes, you have a roadmap for the reply you must file on Monday. 

Scenario 2: Or consider preparing a Daubert challenge with limited time. You can give the model the expert report and ask it to identify methodological weaknesses, inconsistencies with the expert’s deposition, or gaps in the evidentiary foundation. The model produces a list of potential attack points, each of which you can refine using actual law and evidence.  Scenario 3: You are drafting a motion in limine but sense that the argument feels thin. You ask the LLM to articulate the best version of opposing counsel’s likely response. The exercise clarifies the weaknesses in your own framing and helps you rewrite the motion to neutralize those points before they are even raised.  Scenario 4: Picture yourself preparing jury instructions or special verdict forms. An LLM can compare your proposed instructions to the pleadings, jury questions, and evidence, highlighting omissions or inconsistencies that could cause reversible error. In minutes, you have a list of issues that might otherwise surface only on appeal.  These hypotheticals demonstrate the same point: when used strategically, LLMs augment – not replace – the lawyer’s judgment. The human lawyer remains the storyteller, strategist, and decision-maker. The model helps illuminate paths the lawyer might not see as quickly under deadline pressure.

Conclusion

  LLMs are not magic, and they are not junior partners waiting to take over your practice. They are sophisticated language tools that, when used thoughtfully, can make the hard work of legal writing faster, clearer, and more disciplined. They help you see your own drafts with fresh eyes – exposing gaps in logic, weak framing, tonal missteps, and structural problems – while offering alternatives you can accept, reject, or refine.  The key is to treat the model as a collaborator rather than a decision-maker: You remain responsible for the law, the record, and the strategy; the LLM helps you express that strategy in the most effective way. For litigators willing to engage with the technology on those terms, AI becomes less a threat and more a force multiplier – one that can sharpen advocacy, improve craft, and ultimately serve clients more effectively.

Janet Gusdorff is a California Certified Appellate Law Specialist and the founder of Gusdorff Law, P.C. She represents plaintiffs in complex, high-stakes civil appeals and partners with trial lawyers to navigate post-trial strategy, preserve issues, and elevate written advocacy. She can be reached at This email address is being protected from spambots. You need JavaScript enabled to view it..