ChatGPT vs Claude vs Copilot for Lawyers

For most law firms the choice comes down to where lawyers already work. Microsoft 365 Copilot lives inside Word, Outlook and Teams, so it fits firms standardised on Microsoft 365. ChatGPT Business or Enterprise and Claude Team or Enterprise are standalone workspaces that are strong at drafting, summarising and working through long documents. All three need a lawyer to check every citation. None of them runs the firm's back-office workflows across Clio, NetDocuments or iManage unattended.

The firm plans, side by side

Individual plans (ChatGPT Plus, Claude Pro, consumer Microsoft 365 plans) are not what a firm should put client work into. The comparison that matters is between the business tiers each vendor sells to organisations. Prices and wording below are the vendors' own, as of September 2026; follow the links for the current terms.

What each vendor says about your data

Confidentiality is the first question a firm asks, and it is one we will not answer on the vendors' behalf. Here is what each says, in its own words, with the page it says it on. Your IT, risk and general counsel teams should read the underlying terms, not a summary.

• OpenAI (ChatGPT Business and Enterprise). Its enterprise privacy page says: "We do not train our models on your data by default" and that OpenAI "encrypts all data at rest (AES-256) and in transit." On retention it says workspace admins control how long data is kept, and deleted conversations are removed within 30 days unless the law requires otherwise (OpenAI enterprise privacy).
• Anthropic (Claude Team and Enterprise). The pricing page lists "No model training on your content by default" for Team, and the Enterprise page says: "Your prompts, data, and results are not used to train our models by default." Enterprise adds audit logs and data retention controls (Claude Enterprise).
• Microsoft (Microsoft 365 Copilot). The Copilot page says: "Be confident your data is encrypted at rest and in transit, isolated between tenants, and not used to train foundation models." Copilot works over the documents and email a user can already access in Microsoft 365 (Microsoft 365 Copilot).

Vendor terms are only half of the question. The ABA's Formal Opinion 512 (July 2024) says lawyers using generative AI must understand the tool's capabilities and limits, protect client information, communicate with clients about its use, and charge reasonable fees. It recommends informed client consent before client confidences go into a generative AI tool, and says boilerplate consent in an engagement letter is not enough. Your state bar may have its own guidance.

What each is good at for legal work

Drafting

All three produce a usable first draft of a demand letter, a client update, a clause or an internal memo, and all three improve sharply when you give them the firm's own precedent to work from. Copilot has the advantage of drafting where the document already lives: in Word, next to the firm's template, and in Outlook, inside the thread you are replying to. ChatGPT and Claude are separate workspaces, so the precedent and the output travel by upload and copy-paste unless you connect them to your document store.

Summarising and reviewing

Summarising a deposition transcript, a long email chain or a set of contract amendments is where these tools save the most time. Claude and ChatGPT handle long uploads well and let you ask follow-up questions across a set of files in a project. Copilot summarises what is already in the tenant, such as a Teams meeting you attended or a SharePoint folder you have access to, without moving the files anywhere else.

Research, with a caveat

General-purpose models are not citators. They can help frame an issue, suggest search terms, explain a doctrine in plain language and summarise authorities you provide, but a case name, pin cite or quotation they generate has to be verified against a primary source before it goes anywhere. Courts have sanctioned lawyers for filing briefs with citations an AI tool invented. If research is the main use case, legal research products built on a citator are the better fit; our guide to legal AI tools covers them.

A practical rule many firms adopt: the assistant can propose, a lawyer must dispose. Every citation is pulled and read, every factual statement about the matter is checked against the file, and the output is treated as a draft from a junior colleague, not as work product.

ChatGPT for lawyers

ChatGPT Business is sold to teams of 2 to 200 on standard or premium seats, and Enterprise is sold through OpenAI's sales team. OpenAI lists connections to Google Workspace, Microsoft 365, Slack and others, plus "workspace agents for customized workflows" ( OpenAI pricing). For a firm, the draw is a general-purpose assistant with custom GPTs: a firm can build one that knows its style guide or its intake checklist and share it across the practice group.

Choose ChatGPT if your lawyers want one flexible assistant for drafting, analysis and brainstorming, you are not standardised on Microsoft 365, and you want to build shared assistants for recurring tasks.

Claude for lawyers

Claude Team is priced per seat with standard and premium tiers; Enterprise adds SCIM, role-based access, audit logs and custom data retention (Claude pricing). Anthropic lists connectors for Microsoft 365, Gmail, Google Drive and Slack on the Enterprise page. The legal tasks that suit it best are long-document ones: reading a full agreement set and answering questions across it, or redrafting in a consistent voice.

Choose Claude if your heaviest use is reading and drafting against long documents, and you want admin controls such as audit logs and retention settings on the Enterprise plan.

Firms that also use Claude Code to build their own automations should read is Claude Code reliable for law firms? before putting it near matter data.

Microsoft 365 Copilot for lawyers

Copilot is an add-on to a qualifying Microsoft 365 plan, and it works inside the apps most firms already use all day. Microsoft says it can search work data "including LOB applications and services using 100+ connectors" and includes agents grounded in work data (Microsoft 365 Copilot). For a firm on Microsoft 365, the draw is that nothing new has to be approved as a separate data store: Copilot reasons over what each user can already see.

That same design means Copilot is only as tidy as your permissions. If a SharePoint site is open to the whole firm, Copilot can surface it to the whole firm, so an access review is sensible before rollout.

Choose Copilot if your firm runs on Word, Outlook and Teams and you want AI where the work already is, under the Microsoft agreement you already have. See is Copilot reliable for law firms? for how it behaves on legal work specifically.

What none of them do

All three are assistants: they help a person, inside a conversation or a document, and they wait to be asked. That is the right design for drafting and analysis. It is a different job from the repetitive business-side work that fills a firm's operations team's day:

• A new matter email arrives; the conflict check runs; the matter is opened in Clio; the engagement letter goes out through DocuSign.
• Every attachment that lands in a shared inbox is identified, named to the firm's convention and filed to the right workspace in NetDocuments or iManage.
• Before bills go out, time entries in Elite 3E are checked against each client's billing guidelines and flagged for the billing partner.

Each of those runs on a trigger, touches three or four systems, has to happen the same way every time, and needs a record of what ran. The assistants can connect to some tools, and each vendor is adding agent features, but running a scheduled, logged workflow across a practice management system and a document management system is not what they are built for. We map exactly where the connections stop in what Copilot and Claude can't connect to, with system-by-system detail for iManage and NetDocuments.

Where Caddi fits

Caddi is for the business side of the firm, not for legal analysis. Lawyers keep ChatGPT, Claude or Copilot for drafting and review; Caddi runs the intake, filing, billing and records work around the matter.

• Discover finds the repetitive work in the tools your team already uses and ranks it by the hours it takes.
• Automate: the operations person shows the workflow once, and Caddi builds it as deterministic code. No IT project.
• Govern: every run is logged and reviewable, so the firm can show what ran, when and on what data.

AI builds it, verified code runs it, and any model step inside a run is scoped, logged, and reviewable. Caddi is SOC 2 compliant. Plans start with a free Individual plan; Team is $250 per user per month (pricing).

Choose something else if what you need is help writing and reviewing legal documents. That is an assistant's job, and one of the three above, or a legal-specific tool, will serve you better.

How to choose, by firm size

Solo and small firms. Cost and simplicity decide it. If you already pay for Microsoft 365, the Business Standard or Business Premium bundles with Copilot keep everything in one agreement. If not, a handful of ChatGPT Business or Claude Team seats is the quickest start. Either way, set a written rule for what may and may not be entered.

Midsize firms. Admin controls start to matter: SSO, retention settings and a way to see usage. Run a short pilot with two practice groups on two tools, on real (consented or non-confidential) work, and compare the time saved and the review burden.

Large firms. Most end up with more than one tool: an enterprise assistant for everyone, a legal-specific platform for research and review, and a workflow layer for operations. The governance question is the same across all three: who can use it, on what data, and what record is kept.

Related comparisons

• Caddi vs ChatGPT
• Caddi vs Claude
• Caddi vs Microsoft 365 Copilot
• Copilot can't connect to Elite 3E?
• Legal AI tools for law firms
• AI automation for law firms