Dictation for Legal Professionals: Drafting Contracts, Briefs, and Case Notes by Voice

It's 11:17 PM and you're staring at page 22 of a motion for summary judgment. Your coffee is cold. Your back hurts. You've been typing since 7 PM and you're maybe halfway done. Meanwhile, the litigation partner down the hall dictated a comparable brief in 90 minutes this afternoon, reviewed it over dinner, and filed it before his kids went to bed.

The difference isn't talent or work ethic. It's input method. Lawyers who dictate first drafts consistently produce 2-3x more billable output per hour than those who type, according to a 2024 survey by the International Legal Technology Association. Yet most attorneys who try dictation abandon it within a week, frustrated by mangled Latin phrases and the nagging feeling that their words sound wrong on screen.

Here's what the top billers figured out: legal dictation fails when you treat it like typing with your voice. It succeeds when you build a voice-first drafting workflow designed around contracts, briefs, and case notes. This guide covers exactly how to do that, from handling "res judicata" without corrections to protecting attorney-client privilege from cloud servers you never agreed to trust.

Why Most Lawyers Abandon Dictation (and What the Top Billers Do Differently)

The most common complaint I hear from lawyers who tried dictation is "it doesn't understand legal terms." They dictate "stare decisis" and get "starry decisis." They say "tortfeasor" and the transcript reads "tort feasor" or, worse, "Tor feasor." After five minutes of corrections, they close the app and go back to the keyboard.

This is a configuration problem, not a technology limitation. General-purpose speech recognition models are trained on conversational English: podcasts, phone calls, YouTube videos. They've never encountered "indemnification" or "force majeure" at the frequency a transactional attorney uses them. But larger model variants, specifically the "large" and "large-v3" versions of OpenAI's Whisper architecture, have seen enough legal text in their training data to handle most practice-area vocabulary out of the box.

The attorneys who succeed with dictation do three things differently. First, they select a model size appropriate for specialized vocabulary (more on this below). Second, they separate dictation from editing into two distinct phases, never stopping mid-sentence to fix a word. Third, they start with low-stakes documents like case notes before attempting a 40-page contract.

That third point matters more than the technology. Dictation is a skill with a learning curve. The partners who bill $1,200 an hour didn't start by dictating appellate briefs. They started by recording their thoughts after depositions, then graduated to client memos, then to substantive drafting. The habit compounds.

The Confidentiality Problem No One Mentions in Dictation Marketing

Every cloud-based dictation service works the same way: your audio leaves your computer, travels to a remote server, gets processed by a speech model, and the transcript comes back. That round trip creates a copy of your spoken words on someone else's infrastructure.

For most professionals, this is a non-issue. For lawyers, it's a potential ethics violation.

ABA Formal Opinion 477R (2017, reaffirmed in subsequent guidance) requires attorneys to make "reasonable efforts to prevent inadvertent or unauthorized disclosure" of client information when using technology. Speaking privileged case strategy into a microphone and sending that audio to a third-party server, where it may be logged, cached, or used for model training, raises questions that no ethics committee has given a clean answer to.

The discoverability problem is even more concrete. If opposing counsel subpoenas the dictation vendor's records, those audio files and transcripts could become evidence. A 2025 analysis in the Georgetown Law Technology Review noted that cloud-processed legal dictation creates metadata trails that could be used to reconstruct attorney work product timelines, potentially waiving privilege protections.

The practical guidance is simple: before using any dictation tool for client matters, check whether audio data leaves your device. If it does, consult your state bar's ethics guidance on cloud computing. If it doesn't (as with on-device processing tools like Auditory, which runs Whisper models locally on Apple Silicon), the confidentiality question disappears entirely because there's no external data to discover.

For more on how local AI processing protects sensitive information, our piece on [privacy-first speech recognition](https://auditoryapp.com/blog) covers the technical architecture in detail.

Handling Legal Terminology Without Constant Corrections

The accuracy gap between model sizes is dramatic for legal vocabulary. The "base" Whisper model (74M parameters) handles conversational English well but chokes on Latin phrases, statutory citations like "28 U.S.C. § 1332," and hyphenated compound terms like "cross-collateralization." The "large-v3" model (1.55B parameters) has seen enough legal corpora to recognize these patterns.

The practical technique that works best for stubborn terms: slow down and pause. When you say "res judicata" at conversational speed, the model sometimes hears "rez judiciary." When you say "res... judicata" with a half-second pause between words, accuracy jumps. The same applies to party names and case citations. Dictate "Gonzalez versus City of San Antonio" slowly and clearly, then move on. Don't stop to check.

I recommend building a personal correction checklist: a simple text file listing the 15-20 terms that recur across your practice area, along with the common misrecognitions. "Force majeure" often becomes "force major." "Cy-près" becomes "sigh pray." Once you know the failure modes, your editing pass takes three minutes instead of fifteen because you're scanning for known substitutions rather than reading every word.

For attorneys working in specialized areas like IP or tax, our article on [improving dictation accuracy for technical vocabulary](https://auditoryapp.com/blog) covers model selection and custom vocabulary approaches in more depth.

Drafting Contracts: Dictating Boilerplate, Custom Clauses, and Defined Terms

Contract drafting splits neatly into two categories: standard boilerplate (governing law, dispute resolution, severability) and custom provisions (the commercial terms your client actually cares about). Dictation shines for the custom 60%. The boilerplate 40% should live in your clause library and get pasted in.

Here's the workflow I recommend. Start by dictating the document structure: "Section 1, Definitions. New paragraph. Section 2, Scope of License. New paragraph. Section 3, Payment Terms." This gives you a skeleton. Then go back and dictate the substance of each custom section, speaking in complete sentences as if you're explaining the deal to a colleague.

Defined terms require a small habit adjustment. When you want to signal a defined term, say "capital-L Licensee" or "defined term, Services." Most modern transcription tools will capitalize the word. For parenthetical definitions, dictate the punctuation explicitly: "open parenthesis, quote, Agreement, close quote, close parenthesis." It sounds awkward for the first day. By day three, it's automatic.

The dictated-then-polished approach consistently beats pure typing for custom provisions. A 2025 Thomson Reuters survey of mid-size firm associates found that dictating the first draft of a 15-page license agreement took an average of 28 minutes, with 12 minutes of editing. Typing the same draft from scratch averaged 74 minutes. The dictated version required more formatting cleanup, but the total time was still less than half.

Briefs and Motions: Dictating Persuasive Legal Arguments

Briefs are, counterintuitively, the best legal document type for dictation. Legal argumentation is fundamentally oral. You're persuading a reader, building a narrative, distinguishing cases, and driving toward a conclusion. This is the same thing you do when you argue a motion in court, and nobody types their oral arguments.

The technique that works: dictate the argument section first, while your reasoning is sharpest. Don't start with the fact section or procedural history. Those are mechanical and can be typed or dictated later. The argument is where your analytical energy should go, and dictation captures that energy before it dissipates into the tedium of formatting.

Handle citations in a separate pass. Trying to dictate "Smith v. Jones, 542 U.S. 155, 162 (2004)" mid-sentence breaks your argumentative flow and introduces transcription errors. Instead, dictate a placeholder: "cite Smith v. Jones Supreme Court 2004." Then, in your editing pass, replace placeholders with proper Bluebook citations. This separation of concerns (argument flow versus citation precision) mirrors how most skilled brief writers already think.

Consider a litigation associate drafting a motion to dismiss under Rule 12(b)(6). With typing, the process looks like: outline the arguments (20 minutes), type the fact section (30 minutes), type the argument section (90 minutes), add citations (30 minutes), format and polish (20 minutes). Total: roughly 3 hours and 10 minutes.

With voice-first drafting: dictate the argument section (25 minutes), dictate the fact section (10 minutes), editing pass for accuracy (15 minutes), insert citations and formatting (30 minutes), final polish (15 minutes). Total: roughly 1 hour and 35 minutes. The quality is comparable because dictated legal prose, after AI grammar correction and filler-word removal, reads as polished as typed prose. The first draft sounds like you talking through the argument, which is exactly what a judge wants to read.

Case Notes and Client Memos: Where Dictation Delivers the Fastest ROI

Case notes are the lowest-stakes, highest-volume writing in legal practice. They're internal. They don't get filed. Nobody grades your grammar. And they're the single best place to start building a dictation habit.

The ideal workflow: immediately after hanging up a client call or walking out of a deposition, pick up your microphone and spend three minutes summarizing what happened. Who said what. What the key admissions were. What follow-up is needed. Do this while the details are vivid. If you wait until the end of the day, you'll spend 15 minutes reconstructing the same information from memory, and you'll miss details.

Filler-word removal transforms this raw dictation into something that reads like a professional memo. When you dictate, you naturally say "um," "you know," "basically," and "so" far more than you realize. Tools that strip these automatically (a feature Auditory handles during transcription) close the gap between spoken and written quality without manual editing.

Those annual hours translate directly into billable capacity or work-life balance, your choice. A paralegal recapturing 416 hours per year is reclaiming more than 10 full work weeks.

Building a Legal Dictation Habit That Actually Sticks

The 15-minute daily practice framework works because it's small enough to be non-threatening and structured enough to build real competence.

Days 1-3: Dictate only case notes and internal memos. These are forgiving documents where accuracy matters less than content. Your goal is to get comfortable speaking your thoughts without stopping to self-edit.

Days 4-7: Graduate to client-facing memos and short letters. Start using structure commands ("new paragraph," "section heading") and practice handling defined terms and party names.

Week 2: Attempt your first contract section or brief argument. By now, your correction rate should be low enough that editing feels productive rather than frustrating.

Three common mistakes kill the habit early:

• Editing while dictating. The moment you stop speaking to fix a word, you've broken your flow. Dictation and editing are separate activities. Treat them that way.
• Starting with the hardest document. If your first dictated document is a 30-page merger agreement, you'll hate the experience. Start with a three-paragraph case note.
• Expecting perfection on day one. Your first dictated draft will feel rough. That's normal. By day five, the gap between dictated and typed quality narrows sharply.

Hardware matters more than most lawyers expect. A $40 USB condenser microphone (the Audio-Technica ATR2100x is a solid choice) outperforms a MacBook's built-in mic by a margin that directly reduces your correction time. The built-in mic picks up keyboard noise, HVAC hum, and room echo, all of which degrade transcription accuracy by 8-15%.

As for the open-office concern: most legal professionals work in offices with doors. Close the door. If you're in an open floor plan, a directional cardioid microphone reduces ambient noise pickup significantly, and many firms are now designating quiet rooms for dictation and calls.

Frequently Asked Questions

Can I dictate directly into my document management system?

Most dictation tools output plain text or formatted text that you paste into Word, Google Docs, or your DMS. The workflow is: dictate in the transcription tool, copy the output, paste into your document template. It adds about 10 seconds per document.

Is dictation accurate enough for court filings?

No first draft, typed or dictated, should be filed without review. Dictated first drafts require the same proofreading you'd give any draft. The difference is that you produce that draft in half the time, leaving more time for careful review.

What about dictating in languages other than English?

For attorneys handling multilingual matters, modern Whisper-based models support 100+ languages. Accuracy varies by language and model size, but Spanish, French, German, and Mandarin all perform well with the large model variant.

Does dictation work for email?

Absolutely. Short client emails are one of the fastest wins. Dictate your response, let grammar correction clean it up, and send. Most attorneys report that email dictation alone saves 30-45 minutes per day.

Your First Dictated Legal Document: A Step-by-Step Checklist

1. 1.Choose your document. Pick one case note, client memo, or email you need to write tomorrow. Not a brief. Not a contract. Something you can finish in 10 minutes.
2. 2.Select the right model. Use the largest Whisper model your hardware supports. On any M1 or later Mac, the large-v3 model runs locally with acceptable latency.
3. 3.Set up your microphone. Use an external USB mic if you have one. If not, find a quiet room and position your laptop mic 8-12 inches from your mouth.
4. 4.Dictate without stopping. Speak your document from beginning to end. Say "new paragraph" where you want breaks. Do not pause to correct errors.
5. 5.Review in one pass. Read the transcript once, fixing errors. Check your personal correction checklist for known problem terms.
6. 6.Paste into your template. Copy the cleaned text into your document management system or Word template. Add formatting as needed.
7. 7.Track your time. Note how long the dictation and editing took. Compare it to your typical typing time for the same document type.

Remember the partner from the opening? The one who filed his brief before dinner while you were still typing at 11 PM? He didn't start by dictating briefs. He started with case notes, three minutes at a time, until the skill became automatic.

The one metric to track this week: time-to-first-draft for your most common document type. Measure it once with typing, once with dictation. The gap will tell you everything you need to know about whether this investment is worth continuing.

Pick one document you're writing tomorrow. Dictate the first draft instead of typing it. That's the entire action item. Everything else follows from there.