How we compare legal AI

Start with the work, then choose the tool

• Source access and passage support for the jurisdiction you practise in.
• Document workflows, corrections, Word output and continuity between colleagues.
• The actual plan, processing locations, contractual controls and total cost.
• A repeatable trial using approved material, named reviewers and acceptance criteria.

What these recommendations establish

We distinguish a vendor’s advertised capabilities from our judgment about fit. Missing public evidence is a question to ask, not proof that a capability is absent. We do not assign numerical product scores or claim an independent hands-on benchmark.

How to read the Weblaw material

Weblaw’s public challenge describes a limited test by its organizer. The slide columns do not establish a reliable mapping between vendors and results. We retain the observations with anonymous column labels and do not convert them into a named-vendor ranking.

Check the configuration before buying

Plans, licensed collections, integrations and processing arrangements can change. Ask the vendor to confirm the current scope in writing, and repeat the trial on your own workflow. Source-check dates describe the evidence used by this edition.