> ## Content Index > Fetch the complete content index at: https://www.plaintiffedge.com/llms.txt > Use this file to discover other available public pages before exploring further. # Choosing Your First AI Tool: A Plain-English Guide for Plaintiff's Lawyers - URL: https://www.plaintiffedge.com/choosing-an-ai-tool-plaintiff-lawyers/ - Published: 2026-10-17T01:07:43.000Z - Updated: 2026-10-17T01:07:43.000Z - Description: There are dozens of AI tools, and most claim to be built for lawyers. Here's what they are, what the plans really differ on, and how to pick one. - Author: Travis Siegel - Tags: Start Here, AI Basics, Choosing Tools, Confidentiality *Current as of October 2026\. AI products change fast, so check the links below for anything newer.* If you're like a lot of plaintiff lawyers, you've tried ChatGPT once or twice. You asked it a legal question, got a confident answer, and later found out a case it cited didn't exist. Or you never started, because there are dozens of AI products and every one of them claims to be built for lawyers. Picking a tool is simpler than the marketing makes it look. This post covers what these tools are, who the main players are, why the plan you pay for matters more for lawyers than for almost anyone else, and how to pick one to start with. I use Claude every day in my elder abuse practice, but nothing here requires you to use the same tool I do. This is the first post in a short Start Here series. Later posts in the series cover confidentiality, setting up your AI with standing instructions, and building reusable workflows. ## What these tools actually are The AI tools most lawyers are talking about are built on large language models, or LLMs. An LLM is a program trained on an enormous amount of text so that it can predict, word by word, what a good response to your request looks like. At that scale, prediction turns into something useful: it can summarize a deposition, draft a meet-and-confer letter, or explain a lab value in plain English. Four terms will come up again and again on this site: - **Model.** The engine underneath. Each company makes several models, usually a fast, cheap one and a slower, more capable one. - **App.** The website or program you type into, such as ChatGPT or Claude. The app is the car; the model is the engine. One app usually lets you pick among several models. - **Token.** The unit these tools count text in, roughly three-quarters of a word. Usage limits and prices are measured in tokens, so a 500-page chart uses far more of your allowance than a two-line question. - **Context window.** How much text the model can hold in mind at one time, measured in tokens. If you upload more than fits, the tool either refuses or quietly works from only part of what you gave it. The most important thing to understand is that an LLM predicts plausible text. It is not looking things up in a law library unless the product is specifically built to do that, and even then it can get things wrong. That is why these tools invent case citations. Treat everything an AI tool gives you as a draft from a fast, well-read assistant who has never been admitted to practice anywhere. ## The main options AI tools for lawyers fall into two groups: general-purpose assistants that can help with almost anything, and legal-specific tools built on top of a research database or a particular workflow. ### General-purpose assistants Four companies make the assistants most firms will consider. All four can summarize documents, draft letters, answer questions, and analyze files you upload. The differences are in where they live and how they fit the software your firm already uses. - [**ChatGPT**](https://chatgpt.com/pricing?ref=plaintiffedge.com) **(OpenAI).** The most widely used. A good default if you have never tried anything else. - [**Claude**](https://claude.com/pricing?ref=plaintiffedge.com) **(Anthropic).** What I use every day. I've found it strong with long records and careful drafting, and it lets you save reusable instructions (more on that in a later post). - [**Gemini**](https://gemini.google/subscriptions/?ref=plaintiffedge.com) **(Google).** Built into Gmail, Docs, and Drive. Worth a look if your firm runs on Google Workspace. - [**Microsoft 365 Copilot**](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing?ref=plaintiffedge.com) **(Microsoft).** Built into Word, Outlook, and Teams. Sold as an add-on to a Microsoft 365 business plan. Each of these companies updates its models every few months, and the lead changes hands often. The one that was best for your task last spring may not be best today. That's one more reason to judge them by testing your own work, not by headlines. ### Legal-specific tools The big research platforms now have AI built in: [CoCounsel](https://legal.thomsonreuters.com/en/products/cocounsel-legal?ref=plaintiffedge.com) from Thomson Reuters works with Westlaw, and [Lexis+ AI with Protégé](https://www.lexisnexis.com/en-us/products/lexis-plus-ai.page?ref=plaintiffedge.com) works with Lexis. Because they answer from a database of real cases and statutes, they are much less likely to invent a citation than a general assistant. Less likely is not never. A widely cited Stanford study found that even these tools produced incorrect or unsupported answers on a meaningful share of test questions, so every citation still gets checked. There are also tools built for plaintiff firms, such as [EvenUp](https://www.evenuplaw.com/?ref=plaintiffedge.com) for personal injury demand packages. These can save real time on one specific job, but they are not a substitute for a general assistant you can use on everything else. If your firm already pays for Westlaw or Lexis, find out what AI features your subscription includes before buying anything new. ## Free, paid, and business plans: the difference that matters most For most people, the difference between a free and a paid AI plan is how much you can use it. For lawyers, the bigger difference is what happens to what you type in. Every major vendor offers individual plans (free, and a paid tier around $20 a month) and business plans billed per person. On the individual plans, the vendor may use your conversations to train future models unless you turn that setting off. Both [Anthropic](https://claude.com/pricing?ref=plaintiffedge.com) and [OpenAI](https://chatgpt.com/pricing?ref=plaintiffedge.com) list training on their individual plans as "opt-out," which means it's on until you change it. On their business plans, training on your content is off by default. Google's [Gemini Enterprise](https://gemini.google/subscriptions/?ref=plaintiffedge.com) business plans make the same commitment. Here's how the plan types compare: - **Free ($0).** Trains on your chats unless you opt out. Fine for learning the tool and for public information. - **Paid individual, about $20 a month** (Claude Pro, ChatGPT Plus, Google AI Pro). Trains on your chats unless you opt out. Fine for heavier use on non-confidential work. - **Business or team, about $20 to $30 per person per month** (Claude Team, ChatGPT Business, Gemini Enterprise). No training by default. Fine for firm work, after you review the terms. - **Enterprise, custom quote.** No training by default. Built for firm-wide use, with stronger admin and data retention controls. Turning off training on an individual plan helps, but it is not the same as a business agreement. Business and enterprise terms usually add admin controls, limits on how long data is kept, and contractual confidentiality commitments. For client medical records in particular, ask the vendor whether it will sign a business associate agreement (BAA). Anthropic, for example, lists a HIPAA-ready offering only on its Enterprise plan. The simple rule while you're learning: if you wouldn't email it to a stranger, don't paste it into a free or individual AI account. Practice on public material such as published opinions, statutes, and government inspection reports. The next post in this series covers confidentiality in more depth, including what California's [State Bar guidance on generative AI](https://www.calbar.ca.gov/sites/default/files/portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf?ref=plaintiffedge.com) says about it. ## Which model to use Once you pick an assistant, you'll notice a menu that lets you choose a model. Every vendor offers the same basic trade-off: fast and light, or slower and more capable. - **Anthropic (Claude).** Haiku is the fast, light model. Sonnet sits in the middle. Opus and Fable are the most capable. - **OpenAI (ChatGPT).** The Instant modes are fast and light. The Thinking or Pro modes are the most capable. - **Google (Gemini).** Flash is fast and light. Pro is the most capable. The version numbers change constantly, so learn the family names instead. The light models are fine for quick work: cleaning up an email, reformatting a list, summarizing a short article. Use the most capable model for anything that requires judgment, such as reviewing a chart for gaps in care, comparing a witness's statements, or drafting something that will go to a court or opposing counsel. For Claude, Opus 5.5 is the top tier model as of this date and what you should use for complex tasks. Fable sits atop Opus in Claude's hierarchy, but Fable is generally unnecessary for almost any task, however complicated. It also uses considerably more tokens than Opus 5.5. The catch is that the more capable models use up your plan's allowance faster. If you hit your limit halfway through the week, that's usually why. I'll cover how to stretch your usage, and when the fast model really is good enough, in a later post. ## How to pick one You don't need the perfect tool. You need one you'll actually use, set up so you can use it safely. Here's the order I'd go in: 1. **Start with what your firm already pays for.** If you run on Microsoft 365, check whether Copilot is available to you. If you're on Google Workspace, check Gemini. If you have Westlaw or Lexis, find out what AI features are already in your subscription. 2. **Run a one-month test on public material.** Pay for the individual plan of one or two general assistants and give each the same three tasks: summarize a published appellate opinion in your practice area, draft a letter from one of your own blank templates, and explain a public document such as a nursing home inspection report from Medicare's [Care Compare](https://www.medicare.gov/care-compare/?ref=plaintiffedge.com) site. Compare the results side by side. 3. **Pick the one whose drafts need the least fixing.** Speed and polish matter less than how much you have to correct. 4. **Move to a business plan before any client information goes in.** Confirm that training is off, find out how long data is kept, and ask about a BAA if you'll be working with medical records. 5. **Keep legal research in a research tool.** Use Westlaw or Lexis AI features for case law, and verify every citation no matter which tool produced it. Most plaintiff firms will be well served by one general assistant plus the research platform they already have. You can add specialized tools later, once you know where your time actually goes. ## Watch out for - **Invented citations.** General assistants will produce case names and pin cites that look real and aren't. Courts have sanctioned lawyers for filing them. - **Default privacy settings.** Individual plans may train on your chats until you turn that off. Check the setting the day you sign up. - **Usage limits.** Paid plans still cap how much you can do in a given window, and long documents eat through the cap quickly. - **Stale advice.** Prices, plan names, and models change every few months, including the ones in this post. Check the vendor's pricing page before you buy. ## The takeaway The tool matters less than how you use it. Pick one general assistant, test it on public material, and move to a business plan before client information ever goes in. Everything else on this site builds from there. ## Resources - [Claude plans and pricing](https://claude.com/pricing?ref=plaintiffedge.com) - [ChatGPT plans and pricing](https://chatgpt.com/pricing?ref=plaintiffedge.com) - [Google AI and Gemini plans](https://gemini.google/subscriptions/?ref=plaintiffedge.com) - [Microsoft 365 Copilot pricing](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing?ref=plaintiffedge.com) - [State Bar of California, Practical Guidance for the Use of Generative AI in the Practice of Law](https://www.calbar.ca.gov/sites/default/files/portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf?ref=plaintiffedge.com) - [Medicare Care Compare](https://www.medicare.gov/care-compare/?ref=plaintiffedge.com) (public nursing home data to practice on) **Next in the Start Here series:** Confidentiality first: what can and can't go into an AI tool. If this saved you some time, subscribe and I'll send a new post each week. --- *This post is for informational purposes only and is not legal advice. Reading it, or contacting me about it, does not create an attorney-client relationship or any other privileged or confidential relationship. AI tools, laws, and ethics rules change quickly and differ by jurisdiction. Verify everything here independently and make sure your use of AI meets the professional and ethical standards that apply to you.* **About the author.** Travis Siegel is a plaintiff's attorney in California who represents victims of elder abuse and neglect and their families in cases against nursing homes and assisted living facilities. He uses AI daily in his practice and writes Plaintiff Edge to share what works. More about his practice at [neglectattorney.com](https://neglectattorney.com/?ref=plaintiffedge.com).