Methodology
How AI for Legal finds and checks AI case studies, what their sources support, and how to report a correction.
AI for Legal is a source-linked directory of documented AI implementations in legal. Every entry points to a published source and describes work at an organization named by that source. When the source withholds the organization's name, we preserve that anonymity. We do not present hypothetical results as completed deployments. This page explains how we choose entries, what each source supports, and what we have not independently verified.
Where entries come from
Each case study starts with a traceable published source. That may be a vendor case study, press release, regulatory filing, conference talk, or news report documenting the deployment. Each entry names and links to its source. We show a source publication date only when the source provides one; we do not guess a date or source type.
How we decide what to publish
Before an entry is published, automated checks look for an identifiable organization, a concrete problem and solution, and a source that readers can follow. Entries that are too thin, uncertain, or unsourced stay out. These checks are not a human fact-check or an independent audit. If an entry has been verified separately, we say so; otherwise, the source remains the basis for its claims.
How we handle figures and dates
Metrics appear as the cited source reports them. We do not manufacture ROI figures or fill in missing deployment timeframes. When a source reports no metric or timeframe, the entry says “Not reported by source.” When a source does not support another field, the page marks it as unavailable. The publication date on an entry tells you when it was added to the directory; the source publication date is separate and appears only when known. Counts, medians, and ranges are calculated from published entries. If there are not enough comparable figures to support a number, we leave it out or mark it as missing.
Ranked and listed software vendors
A software hub ranks vendors by one signal: the number of published case studies in that category that are linked to the vendor’s published profile. Payment does not change position. If a vendor has records in the category but its current site does not present the hub’s product, the hub leaves it out of the ranking and names it with the reason. A hub is open to search engines only when at least 3 vendors are ranked in the category and the hub shows at least 8 vendors in all. A hub says so when one vendor holds half or more of its records.
Some hubs also list vendors that are not ranked there because no published case study in the category is linked to a published profile for them. We list a vendor only if its product is sold today for that category. Each fact in its row, such as who it serves, pricing, and integrations, comes from a page we link on the row, with the date we checked it. When a fact is not published, the row says “Not published.” Listed vendors appear in a separate table in alphabetical order. They have no rank, no profile page, and no outcome claim, and they are not affiliated with this directory. A listing is not an endorsement. A vendor moves into the ranked table once a published case study in the category is linked to its published profile.
Keeping sources current
After publication, we periodically check whether each source link still works. The most recent successful check is shown as source link checked. That date tells you the link was reachable then; it does not mean the facts were rechecked or that the source published new information. If a link stops working, we flag the entry for review instead of presenting it as current evidence.
Accountability and corrections
The directory is maintained by a named editor with a public professional profile (see the byline above). “Maintained by” identifies who is accountable for the directory; it does not imply that the editor personally reviewed every entry. If you find an inaccuracy, such as a wrong metric, a misattributed company, or a stale source, please tell us and we will investigate, correct, or remove the entry.