The four categories #
The term “AI legal assistant” gets applied to four different things. Conflating them is the source of most confused procurement and most confused legal-press coverage.
Category 1: Practice-management-embedded AI. The AI features built into legal practice management platforms. Clio Duo, MyCase IQ, Smokeball Archie, PracticePanther's AI features. The “assistant” helps with matter summarisation, intake-narrative drafting, time-entry generation, and client communication drafts. Used by lawyers; not consumer-facing. (See the PMS buyer's guide for the full vendor breakdown.)
Category 2: General-purpose enterprise AI used as an assistant. ChatGPT Enterprise, Claude for Work, Gemini for Workspace, Microsoft Copilot. Lawyers use these as a generic assistant: drafting, brainstorming, summarising long documents, organising research notes. Not legal-vertical, but with the data-protection addendum executed they are appropriate for client work. The fastest-growing “AI legal assistant” category in 2026.
Category 3: Legal-vertical AI assistants. Tools designed specifically for lawyers: CoCounsel, Vincent, Lexis+ AI, Harvey, Spellbook. Most of these have an “assistant” mode where the user asks a question and the tool produces a research-anchored answer with citations. The closest thing to a true “AI lawyer’s assistant” in 2026.
Category 4: Consumer-facing ‘ask a lawyer’ products. Tools marketed to consumers as a substitute for or complement to attorney consultation: DoNotPay (in its various pivots), Rocket Lawyer, LegalZoom's AI features, various free or freemium chatbot products. These products have a different regulatory profile (UPL, consumer-protection law) and different quality standards.
What each category is good at #
Category 1 (PMS-embedded) is good at: matter summarisation, time-entry generation, client communication drafts, intake-narrative drafting, document tagging. Operationally tightly integrated with the firm's existing workflow; lower marginal cost; modest AI capability ceiling. Best for solo and small-firm work where the assistant lives inside the existing tool.
Category 2 (general-purpose enterprise AI) is good at: long-document analysis, brainstorming, drafting first versions of memos and letters, organising research findings, comparing positions across documents. Highest flexibility; lowest legal-domain specificity; lowest per-seat cost. Best for surrounding analytical work.
Category 3 (legal-vertical AI) is good at: citation-grade research, brief drafting, contract analysis, document review, deposition preparation, cross-referencing primary sources. Highest legal-domain accuracy; highest per-seat cost; most polished workflows. Best for the firm's central legal-output work.
Category 4 (consumer-facing) is good at: simple legal questions where the answer is a well-known statutory or doctrinal point; document templates for routine consumer transactions (NDAs, simple wills, simple leases); helping consumers understand whether they need a lawyer. Quality varies enormously; UPL exposure for the vendor; not appropriate for actual practice.
The UPL line #
The unauthorized practice of law (UPL) line matters more for “AI legal assistant” than for most legal-AI categories because the term is sometimes used in marketing to consumers in ways that imply the AI is a substitute for a lawyer.
The general framework: AI tools used inside a law firm by lawyers and supervised non-lawyers do not raise UPL issues. AI tools used by lawyers to assist their own practice do not raise UPL issues. AI tools used by consumers to substitute for lawyer advice raise UPL issues in many states.
State bars in 2024-2026 have addressed this in opinions and enforcement actions:
- Multiple state bars have taken enforcement action against AI products marketed as “virtual lawyers” or “AI lawyers” without licensed-attorney involvement.
- Bar opinions in California, New York, North Carolina, and other states track the framework: AI as a tool for lawyers is permitted; AI as a substitute for lawyers raises UPL.
- The line between “legal information” (permitted to non-lawyers) and “legal advice” (lawyer-required) is being tested in real time as AI products approach it.
For lawyers using AI inside their practice, the UPL question is not raised. For lawyers considering recommending consumer-facing AI to clients, the recommendation should be informational not directive (the consumer makes the decision).
“Free AI legal assistant”: a misleading category #
“Free AI legal assistant” is among the most-searched variations of the term. The honest assessment of what is actually available at zero marginal cost in 2026:
- Free consumer-tier ChatGPT, Claude, Gemini. Not appropriate for lawyer use on client documents (the privacy policy permits training; the Heppner ruling treats consumer-tier AI exchanges as outside privilege). Acceptable for general legal-research questions that do not involve client information.
- Free consumer legal-help tools. Many UPL-cautious; quality varies; come with disclaimers that limit reliability. Some genuinely useful for routine consumer questions; not a substitute for legal advice.
- Free trials of paid tools. Most legal-vertical AI vendors offer 7-30 day free trials. Useful for evaluation; not a long-term solution.
For a lawyer asking “what's the best free AI legal assistant?” the honest answer is: nothing free is appropriate for client work. The minimum viable subscription cost (general-purpose enterprise AI at $25-$60/month) is small enough that any practising lawyer should pay for it.
For a consumer asking the same question: free consumer legal-help tools can answer routine questions but cannot replace lawyer advice for anything material. The risk of relying on free AI for an actual legal problem is the consumer's risk, not the AI vendor's risk.
How to choose (for lawyers) #
For lawyers asking “which AI legal assistant should I use?” the answer is almost always a stack rather than a single tool. The minimum viable stack:
- Practice management with embedded AI (Category 1). Whatever your PMS ships with. Marginal cost negligible if you're already on the PMS.
- One general-purpose enterprise AI subscription (Category 2). Claude for Work, ChatGPT Enterprise, or Gemini for Workspace. ~$25-$60/month. Verify the data-protection addendum.
- Optional: a legal-vertical AI (Category 3). CoCounsel, Vincent, Lexis+ AI, Westlaw Edge AI, or Harvey, depending on practice and budget. ~$80-$600+/month. (See the vendor comparison.)
The stack covers the categories of work most lawyers do. Adding more is rarely necessary; adding less leaves capability gaps that show up as inefficiencies or as missed cite-checks.
Category 4 (consumer-facing) is not part of a lawyer's stack. It is something the lawyer might recommend to a client for a specific purpose, with the usual disclaimers.
Frequently asked.
What is an AI legal assistant?
The term covers four different product categories: practice-management-embedded AI (Clio Duo, MyCase IQ), general-purpose enterprise AI used by lawyers (ChatGPT Enterprise, Claude for Work), legal-vertical research AI (CoCounsel, Vincent, Lexis+ AI, Harvey), and consumer-facing legal-help tools (DoNotPay, Rocket Lawyer's AI features). Different tools, different uses, different appropriate buyers.
Is there a free AI legal assistant?
Not one appropriate for lawyer use on client documents. Free consumer-tier ChatGPT, Claude, and Gemini are not Rule 1.6-appropriate. Free trials of paid legal-vertical tools are evaluation-only. The minimum viable cost ($25-$60/month for one enterprise AI) is small enough that any practising lawyer should subscribe.
Are AI legal assistants UPL?
Not when used by lawyers in their own practice. The UPL question arises with consumer-facing AI products marketed as substitutes for attorney advice. Multiple state bars have taken enforcement action against products in that category. Tools used inside law firms are permitted; tools marketed to consumers as ‘virtual lawyers’ raise UPL issues.
Which AI legal assistant should I use?
For solo and small-firm lawyers: practice-management-embedded AI plus one general-purpose enterprise AI ($130-$210/month total). Add a legal-vertical AI (CoCounsel, Vincent, Lexis+ AI) for citation-grade research if the firm's work mix justifies the additional spend.
Dan Hughes is the founder of IXSOR. Twenty years in the UK, twenty-five in North Carolina. Ex-BBC. Ex-Apple. Lifelong technologist. And, most importantly: not an attorney. He writes about legal AI from the operational and infrastructure side, where the rules meet the machines. Reach: [email protected]. More by Dan →