A prospective client with a possible claim used to open a search engine, type something approximate, and get ten blue links. They would open four or five, form an impression, and call two. The firm's job was to be somewhere in those ten and to look credible once opened.
That funnel is closing. The same person now describes their situation in a paragraph (the injury, the timeline, the state) and gets back a short answer that names two or three firms and stops. There is no page two. There is no scrolling past the fold to find you.
Why this traffic converts differently
The person asking has already described their case. They have supplied the jurisdiction, the injury, and often the approximate date of exposure, because that is what it takes to get a useful answer out of a model. By the time they arrive at your site they have self-qualified in a way that a keyword search never required.
Across the firms we track, that shows up in the numbers plainly:
| Channel | Consult → retainer | Cost per signed case |
|---|---|---|
| AI answers | 31% | $2,190 |
| Paid search | 13% | $8,400 |
The volume is smaller. The conversion is more than double and the cost per signed case is roughly a quarter. That is not a marginal channel improvement. It is a different economic shape.
Models do not invent firm names
The useful thing about being named in an answer is that it is not magic. A model producing a shortlist is pulling from a small, identifiable set of sources: the firm's own site, legal directories, court records, bar listings, and press coverage. Seven sources do most of the work in most practice areas.
That means the shortlist is addressable. Not guaranteed, since model outputs are not deterministic and anyone promising you a position is selling something, but addressable in the same way a citation record is addressable. You fix what the model reads, and what the model says changes.
The open ground
The finding that surprises most partners: in more than a third of relevant questions, the model declines to name any firm at all. It describes the claim, explains the general process, and suggests consulting an attorney without naming one.
That is not a competitive market. That is empty space. It is also the cheapest share of answer any firm will ever acquire, and it is closing month over month as more firms notice.
What to measure
Ignore impressions. The number that matters is share of answer: across the set of questions a real client would ask about your practice areas in your geography, what percentage of answers name your firm?
It is a single number, it moves slowly, and it is honest, including in the weeks it does not move. Every other metric in this channel is a proxy for it.