What AI assistants tell people looking for a lawyer
GapWatch measures what AI assistants say when someone asks for a lawyer, an accountant or a professional adviser — and which specific websites those answers are built from. We have measured two United States legal markets and one Canadian one, and the same pattern holds in each.
The finding that surprises firms most
Advertising does not carry into these answers.
In a large United States metropolitan market, two firms with heavy local advertising presence were named in none of eighteen discovery answers. Not once, by any assistant, on any run. Both are real, active firms in that market — the kind of names locals can recite from memory.
An assistant does not see advertising. It assembles an answer from sources it retrieves, and money spent making a name memorable to people does not enter that process at any point.
What decides these answers instead
Directories and editorial rankings, not review platforms
In both United States markets we measured, legal directories and editorial rankings out-cited consumer review platforms — by roughly 1.7× in one market and 4.8× in the other. Forbes, Super Lawyers, Best Lawyers, Best Law Firms, Justia, FindLaw and Expertise.com outweighed Yelp, the BBB, Avvo and Birdeye in each.
The finding holds in both; the size of the gap varies by market, which is itself worth knowing before deciding where to spend. It is close to the opposite of the advice most firms are given, and it explains the result above. The firms being named were not the ones with the largest budgets. They were the ones present in the sources these systems actually read.
Which means the practice area you are filed under matters more than your reputation
In one measurement, a firm with genuine recognition — Chambers, Benchmark, Best Lawyers, Lexpert — was named in 24 of 24 answers when asked about by name, and in 3 of 48 when asked who should handle that kind of matter.
The cause was in the assistants' own words: every directory description led with the firm's other practice area and mentioned the target work as a trailing clause. The firm was not unknown. It was filed under the wrong heading — and the directories are most of what these answers are built from.
And the audiences are different lists
Asked who should handle a matter, and asked who a lawyer should refer one to, the assistants return different firms. In one market a firm placed seventh with prospective clients and second with referring counsel. Most measurement blends the two into a single number that describes neither.
What clients ask us
Is this different from SEO?
Yes. Search optimisation targets ranked links. An assistant produces one answer naming two or three firms, assembled from a much narrower set of sources. A firm can rank first on Google and be named by no assistant at all.
Does this apply outside personal injury?
The measurements above cover personal injury, medical malpractice and class action work. The mechanism — directories and editorial rankings as the substrate — is not specific to a practice area.
What about French or Spanish?
Answers in different languages are built from substantially different sources. In one bilingual market, the assistants named a firm in twelve of twelve English answers about it and none of forty-six French discovery answers. Each language is measured separately, never translated.
Can you guarantee we get named?
No. Nobody controls what an assistant says, which is exactly why measuring it matters.
What an engagement looks like
A frozen baseline before anything changes. Monthly measurement against it across four assistants. The ranked list of sources deciding your market. Then the work on those sources — directory entries, the practice-area descriptions, third-party evidence, and content structured to be quoted rather than only read.
Reported whichever way it goes, including the months nothing moved.