No corner of legal technology has moved faster than artificial intelligence. In under two years it has gone from novelty to a category that draws billions in investment and sits inside the workflows of most large firms. The headline number tells the story.
In March 2026, the legal AI company Harvey raised at an US$11 billion valuation, up from US$5 billion just nine months earlier. Yet a Stanford study found that even the best legal-research AI still hallucinates in a meaningful share of queries. Both facts are true, and both matter.
This guide sorts the serious tools from the hype, groups them by what lawyers actually do with them, and is honest about the two things the marketing skips: accuracy and confidentiality. Funding figures, valuations and study results are sourced and linked at the foot of the page.
Key takeaways
- For large firms and in-house teams: Harvey is the market leader; Legora is the fastest-rising challenger.
- For legal research: Thomson Reuters CoCounsel and Lexis+ AI, both built on trusted legal databases.
- For contract drafting: Spellbook, which runs inside Microsoft Word, and for high-volume review, Luminance or Robin AI.
- The catch: a 2024 Stanford study found leading legal-research tools hallucinated in roughly 17% to 33% of queries, so a human must verify every AI output. Never treat these tools as a substitute for a lawyer's judgment.
The 2026 legal-AI boom: Harvey's $11 billion year
The market has organised itself into three broad segments, and knowing which one you are shopping in saves both money and disappointment.
- Enterprise assistants for large firms and corporate legal departments: broad platforms that handle research, drafting, review and analysis across a whole organisation. Harvey and Legora lead here.
- Legal research engines that layer AI over established case-law and statute databases: Thomson Reuters CoCounsel and Lexis+ AI.
- Task-specific tools that do one job well, most visibly contract drafting and review inside familiar software: Spellbook, Luminance, Robin AI and others.
Harvey is the defining company of the moment. It began, famously, with a proof of concept about landlord-tenant law and a cold email to OpenAI's Sam Altman, and became one of the OpenAI Startup Fund's first investments. It has since raised more than US$1 billion, reaching an US$11 billion valuation in March 2026, and reports working with the majority of the AmLaw 100, hundreds of in-house teams and clients in 60 countries. Its revenue has grown at a pace almost unheard of in legal technology. The lesson for a buyer is not to be dazzled: Harvey is built and priced for large organisations, and a small firm rarely needs it.
How to choose a legal AI tool
The right tool depends on four questions, in order.
- What task? Research, drafting, contract review, litigation analysis and client intake are different problems solved by different tools. Buy for the job you do most.
- What size are you? Enterprise platforms priced for 400-lawyer firms are wasted on a four-lawyer practice. Pricing across this category varies by more than fifty times.
- How is accuracy handled? Prefer tools that ground answers in verifiable sources and show their citations, and assume you will still need to check the work.
- What happens to your data? Client confidentiality is a professional duty. Only use tools whose terms keep your data private and out of model training, and never paste privileged material into a consumer chatbot.
| Tool | Category | Best for |
|---|---|---|
| Harvey | Enterprise legal AI assistant | AmLaw 100 firms and large in-house teams |
| Legora | Enterprise / collaborative AI | Firms wanting a Harvey alternative |
| Thomson Reuters CoCounsel | AI legal research and drafting | Litigation research and memos |
| Lexis+ AI | AI legal research | Source-grounded research with citations |
| Spellbook | Contract drafting in Word | Transactional lawyers redlining contracts |
| Luminance / Robin AI | AI contract review | High-volume review and due diligence |
| EvenUp | Litigation / personal injury | Personal-injury demand packages |
| ChatGPT / Claude / Copilot | General-purpose AI | General drafting and summarising, with caution |
The leading legal AI tools, by category
Enterprise assistants: Harvey and Legora
Harvey is the broad enterprise platform for large firms and corporate teams, spanning research, drafting and analysis, and increasingly built around AI agents that carry out multi-step tasks. Legora, a European company, has become the fastest-rising challenger, emphasising collaboration between lawyers and AI. Both are enterprise purchases with enterprise pricing and onboarding.
Legal research: CoCounsel and Lexis+ AI
Thomson Reuters CoCounsel, built on the Casetext technology Thomson Reuters acquired in 2023, is strong for litigation research, document review and drafting, and connects to Westlaw's case law. Lexis+ AI from LexisNexis grounds its answers in that publisher's legal database and emphasises linked citations. Both are designed to keep the lawyer in the loop with sources to check.
Contracts: Spellbook, Luminance and Robin AI
Spellbook is the leading choice for transactional lawyers because it runs inside Microsoft Word, drafting and redlining contracts where the work already happens, and reports having analysed millions of contracts. Luminance and Robin AI focus on higher-volume contract review, negotiation and due diligence for busy in-house and deal teams.
Litigation and specialist tools
Specialist tools are emerging in every practice area. EvenUp, for example, builds demand packages for personal-injury firms. General-purpose assistants such as ChatGPT, Claude and Microsoft Copilot are widely used for first drafts and summaries, but demand the most caution on both accuracy and confidentiality.
Newer and specialist AI tools worth watching
Legal AI has a long tail of newer and practice-specific tools, from research assistants such as Paxton AI and Alexi to intake, timekeeping and in-house workflow tools, competing on price, a narrow focus or a particular workflow. For a small or specialised firm, one of these can deliver much of the value of an enterprise platform at a fraction of the cost. Trial them on your real work, and hold them to the same accuracy and confidentiality standards as the market leaders.
The hallucination problem, and why verification is non-negotiable
Marketing for legal AI often implies the tools are reliable enough to trust. The best independent evidence says otherwise.
A 2024 Stanford study, "Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools", found hallucination rates of about 17% for Lexis+ AI and 33% for Westlaw's AI-assisted research, compared with 43% for general-purpose GPT-4.
In other words, even purpose-built legal research tools grounded in real databases returned incorrect or misgrounded answers in roughly one in six to one in three queries, per the Stanford HAI summary of the research. The tools have improved since, but the rule has not changed: a qualified human must verify every citation, quotation and conclusion before it goes near a client or a court. Lawyers have already been sanctioned for filing AI-invented cases. Treat AI as a fast, fallible junior, never as the final word.
Confidentiality and ethics
A lawyer's duty of confidentiality does not pause for convenience. Before using any AI tool with client information, confirm three things: that the provider's terms keep your data private and exclude it from model training, that the tool meets your firm's security standards, and that your use complies with the guidance your bar or law society has issued on AI. Purpose-built legal tools generally offer enterprise privacy terms; free consumer chatbots usually do not, and privileged material should never be pasted into them.
What it costs
Pricing ranges enormously, from modest monthly subscriptions for task-specific tools to six-figure and seven-figure enterprise contracts for the large platforms. The single most common mistake is over-buying: matching a small firm's routine work to a platform designed for a global one. Start from the task and the firm size, trial before you commit, and price the specific tool that does the job you actually do.
Frequently asked questions
What is the best AI tool for lawyers?
It depends on the task. Harvey leads for large firms and in-house teams, Thomson Reuters CoCounsel and Lexis+ AI lead for legal research, and Spellbook leads for contract drafting inside Word. There is no single best tool for every lawyer.
Can lawyers rely on AI for legal research?
Not without verification. A 2024 Stanford study found leading legal-research AI tools hallucinated in roughly 17% to 33% of queries. AI can accelerate first-pass research, but a qualified lawyer must check every citation and conclusion before relying on it.
Is it safe to put client information into AI tools?
Only into tools whose terms keep your data private and out of model training, and only in line with your bar or law society's guidance. Never paste privileged or confidential material into a free consumer chatbot.
How much do legal AI tools cost?
Pricing varies by more than fifty times, from modest monthly subscriptions for task-specific tools to six-figure enterprise contracts for platforms such as Harvey. Match the tool to your firm size and the task rather than over-buying.
What is Harvey AI?
Harvey is the leading enterprise legal AI platform, used by much of the AmLaw 100 and hundreds of in-house teams. It reached an US$11 billion valuation in 2026 and offers research, drafting and analysis, but is built and priced for large organisations.
The bottom line
Legal AI is now genuinely useful, and genuinely fallible. Buy for the task and your firm size, insist on privacy terms that protect client data, and verify every output. For more practical resources see our free legal tools and our guide to the best legal practice management software, and if you are building or joining a firm, browse current legal jobs and internships on LegalAlphabet.
Sources and further reading
- Harvey, growth round at an $11 billion valuation (2026).
- Stanford HAI, AI on Trial: Legal Models Hallucinate in 1 out of 6 (or more) queries (2024).
- Stanford RegLab, Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools.
Last reviewed July 2026. This is one of the fastest-moving areas of legal technology; funding, features and accuracy claims change quickly, so verify current details with each provider.