Legal AI Price Index 2026: what 29 tools actually cost
Twenty-two of the twenty-nine vendors below publish no list price at all. That is the first useful fact about this market: for most of it, the only way to learn the cost is to enter a sales process. This page collects what is knowable, grouped by how each vendor charges rather than alphabetically, because the charging model tells you more about the fit than the number does.
Figures marked reported come from marketplace and procurement data, not from the vendor. Treat them as a way to tell a $200 a month decision from a $200,000 a year one. They are not quotes.
Published list prices
Seven of the twenty-nine state a price you can read without talking to anyone. In a market this opaque, that is itself a differentiator. Ordered by entry point.
| Tool | Price | Charged by |
|---|---|---|
| Termly | $10 per site per month, billed annually | Site |
| PandaDoc | $19 per user per month | User |
| GenieAI | Free plan, then $75/mo (Pro), $320/mo (Business, 5 users), from $600/mo (Enterprise) | Feature tier |
| Sypro | GBP 149 per project per month, projects under GBP 3m | Project |
| Lawhive | From ~GBP 199 per matter | Matter |
| Luminance | List ~$41,340 a year for 1,000 documents, unlimited users | Document volume |
| Mistral AI | $0.02 to $6.00 per million tokens | Usage |
Per user, per month
The model most teams can budget for. Watch for seat minimums, which are where the advertised figure and the invoice part company.
| Tool | Reported price |
|---|---|
| Wordsmith | ~$450 per user per month |
| Spellbook | ~$99 to $199 per user per month, ~$350 on enterprise plans with a 10-seat minimum |
| Summize | ~GBP 99 per user per month |
| Parley Pro | From ~$79 per user per month |
| LegalFly | ~$75 per user per month plus a platform fee |
| Quill | ~$22 to $50 per user per month |
Per seat, per year
| Tool | Reported price |
|---|---|
| Hebbia | ~$3,000 to $10,000 per seat per year |
| Legora | ~$3,000 per seat per year |
| Oracle Aconex | From ~$3,000 per user per year |
Annual contracts, quoted
Platform purchases. The spread inside each range is driven by seat count, contract volume and which modules are licensed, so the low end rarely describes what a mid-sized buyer pays.
| Tool | Reported annual contract |
|---|---|
| Harvey AI | ~$50,000 to $290,000 |
| Ironclad | ~$30,000 to $250,000, marketplace median near $40,000 across 363 recorded purchases |
| Leah | ~$50,000 to $200,000 |
| Lexion | ~$15,000 to $150,000 |
| Evisort | ~$30,000 to $150,000 |
| Juro | ~$15,000 to $130,000, marketplace medians around $31,000 to $34,500 |
| Procore | ~$15,000 to $80,000 |
| Apryse | From ~$1,500 |
Usage, volume and everything else
| Tool | Reported price | Charged by |
|---|---|---|
| Relativity | ~$11 to $30 per GB per month | Data volume |
| Contractbook | ~$39 to $599 per month | Plan tier |
| Contract Companion | ~$674 for organisations under ten employees | Organisation size |
| Leegality | ~INR 2 to 3 lakh a year across tiers | Tier |
| Looplex | Usage-based, ~R$19.90 per user above the included tier | Documents and workflows |
What the spread tells you
The gap between the cheapest and the dearest tool here is roughly four orders of magnitude, and most of it is not explained by capability. It is explained by who the vendor sells to. Tools sold to law firms and enterprise legal departments carry an implementation project and price accordingly. Tools sold to commercial teams price per seat because the buyer has no procurement cycle to survive.
So the useful question when comparing two figures is not which is cheaper. It is whether the cheaper one is priced that way because it does less, or because it is sold to someone who would never sign a $50,000 contract.
Method and corrections
Published prices are quoted from the vendor's own pricing page as at the date above. Reported ranges come from marketplace and procurement datasets and are labelled as reported everywhere they appear. Where we could not source a figure we leave the entry out rather than estimate one. LegalOn is absent from this page for that reason.
Prices move. If a figure here is wrong, or if you have a quote that contradicts a range, we would rather correct it than leave it. Full method is on how we test.