Every legal AI tool now claims to do research, drafting and review. Most of them do one of those well, one adequately, and the third not at all, and the marketing does not tell you which. This is a working comparison written for a lawyer or practice manager choosing one, with the caveat that we make one of the products on the list. Where we describe a competitor we have stuck to what they say about themselves publicly, and the table at the end says where we think each one belongs.
First, the test that matters more than the feature list
Take a question your firm has already researched and has a memo for. Run it through the tool. Then try to open every citation the tool gives you. A tool that retrieves the judgment or statute before it answers will let you open all of them. A tool that answers from the model’s memory will give you at least one authority that does not exist, or exists but says something else. Courts in the United States, the United Kingdom, Canada and Australia have all sanctioned lawyers who filed the second kind. Nothing else on this page matters if a tool fails that test.
The three kinds of legal AI tool
It helps to sort the market before comparing products, because the products in different categories are not really competing.
- AI layers for large firms. Sold on enterprise contracts to firms that already have practice management, document management and research subscriptions, and want a single AI surface across all of it. Harvey and Legora are the clearest examples. Strong research and review, no matter management, priced for the top of the market.
- AI inside an incumbent’s product. CoCounsel inside Thomson Reuters, Lexis+ AI inside LexisNexis, Clio Duo inside Clio. Each is good at the thing its parent already did and limited by the parent’s data and jurisdiction footprint.
- Single-job tools. Spellbook for contract drafting in Word, Eve for plaintiff-side case work, Paxton for research and drafting at a lower price point. Narrow, often excellent at the narrow thing, and easy to adopt because they ask for little.
ModulawAI is in a fourth category with fewer occupants: an AI-native practice management system, where research, drafting, review and agents are built into the platform that also holds the matters, time, billing and client portal. The trade-off is obvious. A firm that loves its current practice management system gets less from us than a firm that wants to replace it.
The tools, one by one
Harvey
The best-known legal AI company, backed by OpenAI and sold to the largest firms and in-house teams. Research, drafting, an assistant, a Word add-in and agentic workflows, with a strong emphasis on grounding answers in sources. Harvey is not sold to small firms and does not include practice management. If you are a two-hundred-lawyer firm with settled systems, it is on your shortlist. If you are not, it is probably not available to you. See our Harvey AI alternative page for the fuller comparison.
CoCounsel (Thomson Reuters)
Originally Casetext, now the AI assistant across Westlaw and Practical Law. Its strength is exactly what you would expect: US research grounded in the Westlaw corpus with its citator. Its limits are also what you would expect: the jurisdictions Westlaw covers well, and a price that assumes a Westlaw relationship. Outside the US it is thinner. See CoCounsel alternative.
Lexis+ AI
LexisNexis’s equivalent: conversational research and drafting over the Lexis corpus, with Shepard’s for treatment. Strong where Lexis is strong, including the UK, and with the same structural dependency on a Lexis subscription. A sensible choice for a firm already paying for Lexis that wants the AI layer without a new vendor.
Legora
A European entrant that has grown quickly with large firms, built around collaboration, tabular review across document sets, and a Word add-in. Closest in shape to Harvey, with a strong review product. Again no practice management, and again enterprise sales. See Legora alternative.
Spellbook
A Word add-in for transactional lawyers: drafting, redlining, clause suggestions and playbook review, with an agent that can work across a document. Very good at the one job. It does not research the law behind a clause, does not know which matter the contract belongs to, and reviews one document at a time. See Spellbook alternative.
Paxton
Research, drafting and document analysis aimed at solo and small-firm lawyers in the US, at a price the large-firm products do not offer. US-focused, and a good example of a single-job tool that gets adopted because it asks little of the firm.
Eve
Built for plaintiff-side firms: intake, case evaluation, demand letters and discovery for employment and personal injury practices. If that is your practice it is worth a look. If it is not, it is not for you, which is a virtue in a product.
Clio Duo
Clio’s AI assistant inside its practice management product: summaries, drafting help, time entry suggestions and questions about the matter. Useful for Clio firms and a sign of where practice management is heading. It is not a research tool grounded in primary law, and its reach is Clio’s data. See ModulawAI vs Clio.
ModulawAI
Our product. Research grounded in retrieved primary law for the US, UK, Canada, EU, Nigeria, Ghana, Kenya and South Africa with linked citations and treatment; drafting in Word as track changes from the firm’s own precedents; tabular review across document sets; background agents with approvals; and the practice management around it, matters, calendar, time, billing, client portal and CRM, priced per seat with usage credits from solo to enterprise. The obvious weakness: if you want to keep your practice management system, you are paying for modules you will not use, and our archive of older law reports in some jurisdictions is shallower than the incumbents’. The obvious strength: no other tool on this list covers the four African jurisdictions, and none of the others puts the AI inside the file.
Side by side
| Tool | Category | Research grounded in sources | Drafting in Word | Bulk document review | Practice management | Sold to |
|---|---|---|---|---|---|---|
| Harvey | AI layer | Yes | Yes | Yes | No | Large firms, in-house |
| CoCounsel | Incumbent AI | Yes, Westlaw corpus | Yes | Yes | No | Westlaw customers |
| Lexis+ AI | Incumbent AI | Yes, Lexis corpus | Yes | Partial | No | Lexis customers |
| Legora | AI layer | Yes | Yes | Yes | No | Large firms |
| Spellbook | Single job | Limited | Yes, core product | One document at a time | No | Transactional lawyers |
| Paxton | Single job | Yes, US | Partial | Partial | No | Solo and small firms, US |
| Eve | Single job | Partial | Partial | Yes, case files | Partial, plaintiff workflow | Plaintiff firms |
| Clio Duo | Incumbent AI | No | Partial | No | Yes, Clio | Clio customers |
| ModulawAI | AI-native platform | Yes, eight jurisdictions | Yes | Yes | Yes | Solo to enterprise, incl. Africa |
Based on publicly available information at the time of writing. Vendor capabilities and pricing change frequently; verify with each vendor before deciding.
Who is funding this, and who they build for
Funding rounds are not a quality signal, but they are an excellent predictor of who a company will sell to, which matters more to a buyer than the leaderboard. In March 2026 Harvey raised $200 million at an $11 billion valuation co-led by GIC and Sequoia, then in September 2026 raised $550 million at $15.5 billion co-led by Diffusion and Lightspeed, alongside acquiring the AI security company Guardrails AI. Legora raised a $550 million Series D led by Accel in March 2026 at $5.55 billion, added a $50 million extension in April taking it to about $5.6 billion with Nvidia’s venture arm participating, and was reported in August 2026 to be raising again above $10 billion. Its total funding stands at roughly $816 million.
Capital at that scale gets deployed against enterprise contracts, and the products reflect it: strong security posture, published benchmarks, agentic research, and an implementation process that assumes a legal operations function on the other side. If you are a large firm, that is exactly what you want. If you are a twelve-lawyer practice, you are not the buyer those roadmaps are written for, and the pricing will tell you so.
The incumbents are in a different position again. CoCounsel and Lexis+ AI are distribution plays as much as product plays: they reach customers who already pay for Westlaw or Lexis, and their research quality is bounded by the corpus their parent already licensed. Clio Duo reaches firms already running Clio. None of this is a criticism. It does mean that for most tools, the question of whether it fits you is largely answered by what you already buy.
The column this table does not have: can it build you a tool?
Every product above produces the same class of output. The most advanced version of it is workflow automation. Harvey’s Agent Builder lets a firm assemble a multi-step process in natural language, with conditionals, classification and role based permissions, and Harvey reported more than 25,000 customer built workflows by February 2026, with firms adding roughly 300 a week (harvey.ai). Legora’s Workflows chain drafting, tabular review, research and translation into a single agentic run. These are genuinely good features. They are also pipelines: documents go in, and documents, tables and answers come out.
None of the nine products in the table produces a running application. ModulawAI does, through Canvas, an interactive artifact builder that turns a plain description into working software. Describe a damages calculator, a client intake form, a limitation period checker, a settlement scenario board or a small internal dashboard, and you get a real page you can open and use, usually in a few minutes, with no code written by the lawyer.
The honest comparison here is not with a legal product at all. Claude Artifacts and ChatGPT Canvas do this, and they do it well. The difference is context: a general assistant does not know your matters, your precedent library, your jurisdiction or who is allowed to see which file, so anything it builds starts from a blank page and whatever you paste into it. If you already use Claude or ChatGPT for this and it works for you, that is a reasonable answer, and it is worth saying so in a list that is otherwise about legal vendors.
The limits are worth putting next to the claim. Canvas apps run in a sealed sandbox on an opaque origin with no network access at all. That is deliberate, because it means an application assembled by a model cannot reach your matter file, your firm’s data, your session cookies or any outside service. It also means the app is not wired into your data: its inputs are typed in. And these are internal tools today. Publishing one to a public web address for clients to use is on the roadmap and is not shipped. Read the capability as a way for a fee earner to get bespoke software in an afternoon instead of never, which is a narrower claim than it first sounds and still one nothing else on this list makes.
Where adoption actually is
Clio’s 2025 Legal Trends Report, published 16 October 2025, found 79% of legal professionals using AI tools in their daily work, against 19% in 2023. The ABA’s 2024 Legal Technology Survey Report, from 512 attorneys in private practice, put firm-level adoption at 30%, up from 11%, with 46% at firms of 100 or more attorneys and 18% among solos. Thomson Reuters estimates AI will free about 240 hours per professional per year, worth roughly $19,000 each. Every one of those numbers comes from an organisation that sells legal technology or research, this article included, so weigh them accordingly. The direction is not seriously disputed; the magnitudes are marketing until your own firm measures them.
The counterweight is the sanctions record. Damien Charlotin’s AI Hallucination Cases database logged 2,041 decisions involving fabricated AI content as at 14 September 2026 across more than 40 jurisdictions, 1,395 of them in the United States. ABA Formal Opinion 512 of 29 July 2024 and the SRA’s warning notice of 17 August 2026 both put the responsibility squarely on the lawyer regardless of the tool, and the SRA is explicit that reliance on an AI output would not be a suitable defence.
How to choose in an afternoon
- Decide whether you are keeping your practice management system. If yes, you want an AI layer or an incumbent’s AI. If no, or if you do not have one, you want an AI-native platform.
- List the jurisdictions you actually advise on. Cross off every tool that does not hold primary sources for them.
- Run the citation test above on whatever is left, with a question you have already answered.
- Add up the real cost per seat: the AI tool plus everything it does not replace.
- Ask each vendor four security questions: where is the data held, is it used for training, who can see which matters, and is every AI action logged with a reviewer. Refuse to upload anything until you have written answers.
If ModulawAI survives your shortlist, book a demo and bring the memo. We would rather lose the comparison on a real question than win it on a slide.
Sources
- Clio, 2025 Legal Trends Report (16 October 2025), for AI adoption and realisation figures.
- American Bar Association, 2024 Legal Technology Survey Report, released March 2025, 512 respondents in private practice.
- Thomson Reuters Institute, Future of Professionals 2025, for the 240 hours and $19,000 per professional estimates.
- Harvey on its September 2026 round and the Guardrails AI acquisition; CNBC on the March 2026 round.
- Legora on its Series D; TechCrunch on the extension and NVentures participation.
- Spellbook pricing, quoted rather than published; per-seat ranges from third-party reviews including Lawyerist.
- Damien Charlotin, AI Hallucination Cases database, 2,041 cases as at 14 September 2026.
- ABA Formal Opinion 512 (29 July 2024) and the SRA warning notice on the misuse of AI (17 August 2026).
Most of the tools above answer legal questions. Running the firm is a different job: see our guide to choosing a legal CRM for how intake, matters, time and billing fit together.