1. Per-seat pricing
Per-seat pricing charges a fixed fee per user per month or per year. This is the most common model in legal software generally, and many legal AI platforms have adopted it. The appeal is simplicity: you know your cost upfront and it scales linearly with team size.
Typical per-seat pricing for legal AI platforms ranges from $50 to $500 per user per month, depending on the platform, the features included, and the tier. Some platforms offer tiered plans where basic research costs less than a full suite including drafting, document analysis, and matter management.
Advantages. Predictable budgeting. No surprise bills. Easy to calculate ROI per attorney. Encourages unlimited use since there is no per-query cost, which means attorneys are more likely to use the tool thoroughly.
Disadvantages. Can be expensive for large teams with uneven usage. A firm with 50 attorneys where only 20 use the tool regularly is paying for 30 unused seats. Some per-seat models include usage limits that make the "unlimited" framing misleading.
What to check. Are there usage caps within the per-seat price? Can you add or remove seats monthly? Is there a minimum seat requirement? Are admin and read-only users counted as seats?
2. Usage-based and per-query pricing
Usage-based pricing charges based on what you consume: queries run, tokens processed, documents analyzed, or some combination. Per-query pricing is a specific variant where each search or analysis task has a defined cost.
This model can appear cheaper than per-seat pricing on paper, especially for small teams or occasional users. A solo practitioner who runs ten AI research queries a month might pay less under a usage model than under a per-seat model.
Advantages. You only pay for what you use. Low barrier to entry. Can be cost-effective for occasional or part-time users.
Disadvantages. Creates a psychological cost barrier. Attorneys hesitate to run additional queries because each one costs money. This discourages the iterative research approach that produces the best results. Usage costs are unpredictable, making budgeting difficult. Heavy users may find that usage-based pricing costs more than a flat per-seat rate.
What to check. How are queries defined? Does a follow-up question count as a new query? How are tokens calculated? What counts as a "document" for document analysis pricing? Is there a monthly spending cap option?
3. Enterprise and custom pricing
Many legal AI platforms, especially those targeting Am Law 200 firms and corporate legal departments, do not publish pricing at all. Instead, they require a sales conversation and provide custom quotes based on firm size, expected usage, and negotiating leverage.
Enterprise pricing typically involves annual contracts with significant commitments. The vendor customizes the pricing based on your specific situation, which can result in a better deal for large buyers or a worse deal for firms without negotiating leverage.
Advantages. Custom configurations for large deployments. May include dedicated support, custom integrations, and SLA guarantees. Volume discounts for large teams.
Disadvantages. No price transparency. Lengthy sales processes. Difficult to compare with other options. Often requires annual commitments. The custom nature of each deal means you cannot easily benchmark your price against what other firms are paying.
What to check. What is the minimum commitment period? What happens if usage exceeds the quoted amount? Can you scale down without penalty? Are implementation and training costs included or separate?
5. Why transparent pricing matters
Transparent pricing is not just a preference. It is a signal of how the vendor views their relationship with you. A vendor that publishes pricing is saying: we believe our product is worth this price and we are willing to let you compare it directly with the alternatives.
Hidden pricing creates information asymmetry that benefits the vendor. The sales team knows what other firms pay; you do not. They know their margin; you do not. They can anchor the conversation to your current spending rather than to the actual value of the product. This dynamic wastes your time and often results in you paying more than you should.
For law firms, transparent pricing also simplifies internal approval processes. When you can point to a published price list, the budget conversation with firm management is straightforward. When pricing is custom and undisclosed, the approval process involves more stakeholders and takes longer.
Irys publishes all pricing on its website. There are no hidden tiers, no enterprise-only features, and no per-query surcharges. Every firm gets the same price for the same product. This approach reflects a belief that the product should sell on its merits, not on the sales team's negotiating skill. See the Irys pricing page for current rates.
6. A framework for comparing total cost
To compare legal AI platforms on cost, you need to look beyond the headline price. Here is a practical framework for calculating total cost of ownership.
Calculate the annual subscription cost. For per-seat pricing, multiply the per-seat cost by the number of users. For usage-based pricing, estimate monthly usage based on current research and drafting volume, then multiply by twelve. Add any annual minimum commitments.
Add implementation costs. Include onboarding fees, data migration costs, and any integration development. These are one-time costs but they can be significant and should be amortized over the expected contract period.
Factor in time savings. Estimate the hours saved per attorney per month. Multiply by the blended billing rate or internal cost rate. This is the value side of the equation and it often dwarfs the subscription cost. A tool that saves each attorney five hours per month at a $300 blended rate generates $1,500 per month in recovered capacity, per person.
Consider displacement. If the AI platform replaces or reduces an existing subscription like Westlaw or Lexis, the net cost is the AI platform price minus the reduction in the existing subscription. Some firms find that the net cost of adopting AI is negative after accounting for displaced tools.
Evaluate the risk of switching. Factor in contract lock-in periods, the cost of migrating away if the product does not work, and the retraining cost for your team. Platforms with monthly commitments and easy data export carry less switching risk than those with multi-year contracts and proprietary data formats.
7. Vendor-by-vendor pricing table
Each row reflects what the vendor shows on its own pricing or product pages, checked Oct 2026. "Not published" means no list price appeared on the pages checked. We do not repeat third-party figures that cannot be traced to a vendor page. Where a vendor has its own pricing guide, the vendor name links to it.
| Vendor | Pricing model | Published list price | Free trial | Source | Checked |
|---|---|---|---|---|---|
| Irys One | Flat monthly per seat | $299 per month per seat (Irys One) | 7 days, no credit card required | Irys pricing page | Oct 2026 |
| Harvey | Sales-led custom quote | Not on their site | Not published | Vendor's own pages | Oct 2026 |
| Legora | Sales-led custom quote; premium tier described as consumption-based | Not on their site | Not published | Vendor's own pages | Oct 2026 |
| CoCounsel | Plans listed by firm size; quote or sales call | Not on their site | Yes, by request (CoCounsel Essentials) | Vendor's own pages | Oct 2026 |
| Spellbook | Per license, set by team size; two named plans | Not on their site | 7 days | Vendor's own pages | Oct 2026 |
| Paxton | Per user, monthly or yearly; custom Enterprise | $499 per user per month ($2,999 per year) | 7 days, credit card required | Vendor's own pages | Oct 2026 |
| Laine AI | Flat monthly plans by user count, with an included usage allowance; usage beyond the plan billed at a per-use rate | From $350 per month (Solo plan, 1 user) | 7 days | Vendor's own pages | Oct 2026 |
| GC AI | Flat monthly Individual plan; Team priced on request; Enterprise custom | $500 per month (Individual plan) | 14 days, after a demo | Vendor's own pages | Oct 2026 |
| Lexis+ AI (now Lexis+ with Protégé) | Customized quote | Not on their site | 2 days | Vendor's own pages | Oct 2026 |
| Westlaw | Quote-based subscription (Westlaw Advantage) | From $256.75 per month (single circuit) to $399.75 per month (all states and federal) | 7 days, no credit card needed | Vendor's own pages | Oct 2026 |
| Clio Work | Quoted per user; sold on its own or packaged with Clio Manage | Not on their site | Yes, length not published | Vendor's own pages | Oct 2026 |
Clio quotes Clio Work per user rather than listing a price. The one Clio figure that is published is the Clio Manage Starter plan, at $49 per user per month, which is the practice-management subscription underneath rather than the AI. Confirm any figure with the vendor in writing before you buy.
Frequently asked questions
How much does legal AI software cost per user?
Published per-seat legal AI prices generally run from roughly $20 a month for a light, single-feature tier up to several hundred dollars a month for a full research and drafting platform. Irys One is $299 per seat per month, published on one page. Most of the larger vendors publish no number at all and quote per firm after a sales call, so a per-user figure for those only exists once you have asked for it.
What pricing models do legal AI vendors actually use?
Four, and most vendors use more than one. Per seat charges a flat amount for each named user. Plan tiers bundle a seat count and a feature set into a named package. Usage-based pricing charges per query, per document, per matter or per token, often on top of a smaller base fee. Enterprise or custom pricing replaces all of it with a negotiated quote. The model matters more than the headline number, because only the first two give you a bill you can predict before the month starts.
What is the difference between flat-rate and per-seat pricing?
Per-seat pricing multiplies by headcount: ten attorneys cost ten times one attorney, and the amount per person stays the same however hard they work. Flat-rate pricing charges one amount for the firm or for a plan tier regardless of how many people use it. The two are often confused because a per-seat price can still be flat in the sense that it does not vary with usage. That second sense is the one that decides whether your bill is predictable.
Which legal AI tools have predictable pricing rather than variable usage charges?
A price is predictable when the vendor publishes it, states what is included, and charges no overage for normal work. Irys One meets that test at $299 per seat per month with no per-query or per-token charge on the advertised plan. Vendors that quote only after a sales call cannot be assessed for predictability from the outside, because the terms that create variability, usage allowances, overage rates and annual uplifts, live in the contract rather than on the website. The table below marks which vendors publish a figure and which do not.
What hidden costs should I watch for in a legal AI contract?
Seven recur: a seat minimum that sets a floor well above the number of people who will use the tool; implementation or onboarding fees charged once at signing; training and change-management billed separately; integration or connector fees for your document management system; overage rates for queries, documents or tokens beyond an allowance; support tiers where a usable response time costs extra; and an annual uplift clause that raises the rate automatically at renewal. Ask for all seven in writing before you compare two quotes, because any one of them can move the real total by a third.
How do I compare total cost of ownership beyond the per-seat rate?
Build a three-year number, not a monthly one. Take the seat rate times the seat minimum times twelve, add implementation, training and integration fees in year one, add the expected overage based on your own volume rather than the vendor's example, then apply the annual uplift to years two and three. Do that for each vendor on identical assumptions. The exercise usually reorders the shortlist, because the cheapest headline rate and the cheapest three-year total are frequently different vendors.
What does legal AI cost for a 50-person firm?
Multiply the per-seat rate by the number of people who will genuinely use it, not by headcount. At $299 per seat, fifty seats is $299 times fifty per month, but most firms that size start with the ten to fifteen attorneys doing the drafting and research and add seats as use spreads. Watch for a seat minimum in that range, since a vendor requiring fifty seats from day one removes the option of starting small.
What do enterprise legal AI tiers usually include?
Single sign-on, administrative controls over who can see which matters, audit logging, a named customer success contact, security review support, and a negotiated data-handling agreement. They are bought for governance rather than for extra capability, and the per-seat rate is normally higher than the self-serve tier for the same underlying product. If none of those six items is something your firm is required to have, the enterprise tier may be paying for procurement features rather than legal work.
How should I price a team with a few heavy users and many occasional users?
This is the case where the pricing model decides the winner. Usage-based pricing favors it, because the occasional users cost almost nothing. Per-seat pricing penalizes it, because a person who logs in twice a month costs the same as a daily user. Two practical options: buy seats only for the heavy users and leave the occasional ones off the platform until their use justifies a seat, or look for a lighter published tier for the occasional group. Count your actual heavy users before you ask for a quote, because vendors size the proposal from headcount if you let them.
Where can I find transparent published pricing for legal AI?
On the vendor's own pricing page, and the test is whether a number appears there without a form. Of the platforms in the table below, a minority publish a rate; the rest route you to a demo request. The table records what each vendor published and the month it was checked, so you can see which is which before you spend a call on it. Confirm any figure with the vendor in writing before you buy, because published pages change without notice.
