For firms with several practice areas

Why AI recommends you for one practice area and a rival for the rest

Evangelos Pantidos · WhoAINames · 13 October 2026 · 9 min read

Why are we recommended for one practice area but not the others?

Because a buyer never asks AI about your firm in general. They ask about their problem: “Who’s the best family solicitor in Cork?”, “Which accountants do R&D tax claims for tech startups in London?” The assistant answers that question, using whatever it finds that matches it.

If the web says a lot about your employment team (directory rankings, client reviews that mention the case, a partner quoted in the press) and very little about your commercial disputes team, the assistant has plenty to work with for the first question and almost nothing for the second. It fills the gap with the firm that does have material for that service.

You can see this in the data. In a July 2026 study of 3,200 US legal prompts across 32 practice areas, the assistants’ overlap on firms varied widely by practice area: highest for corporate law (a 45.2% three-way agreement score), lowest for estate planning, where they never once named a common firm. Whether the assistants agree on a firm really does change from one practice area to the next.

Doesn’t our reputation carry across every service?

Only partly. A strong firm name helps an assistant recognise you, but it still has to connect you to the specific service someone asked about. If your website lists twelve practice areas in a menu and only two have pages that actually answer buyers’ questions, the other ten are names without evidence.

Wording matters too. A UK study by QualitySolicitors (over 700 manual tests across ChatGPT, Google AI Overviews and Claude) reported that different phrasings of the same need returned different firms. Buyers don’t search for your internal practice-area names; they describe their situation. Each way of asking is its own question.

Why does a smaller firm get named for our other services?

Because AI doesn’t rank by size. As far as anyone can tell from the outside, it names the firms the wider web agrees on for that question: the ones that show up consistently in the directories, review sites, lists and articles it reads. A ten-person boutique that does one thing, and is described the same way everywhere, often beats a larger firm whose specialism is buried in a long list.

The same QualitySolicitors study reported that a small cluster of firms dominated recommendations across platforms and locations, while many well-established local firms barely appeared. Being established isn’t the signal. Being clearly and repeatedly associated with the service is.

What it looks like: one firm, several practice areas

Here’s a real check on a national UK law firm best known for personal injury, anonymised. We asked 20 buyer questions across its practice areas, each once on four AI engines, which gives 80 answers.

Even inside its best-known practice area, the firm was named for one kind of question and not the rest:

Three things stand out. First, the firm is a regular answer when a buyer asks for the best personal injury solicitors, and absent when the same buyer asks for the best personal injury lawyer. In this check, one change of wording went with a different answer (each question was asked once per engine). Second, it was named once in 32 answers to the questions buyers ask before they shortlist anyone: how to choose, and what it costs. Third, outside personal injury it barely appears.

The sources the engines cited most were a legal directory (Chambers), a lawyer-ranking site (Best Lawyers) and a competing firm’s own website. The firm’s own site wasn’t among them. That’s the per-service gap in practice: the firm’s reputation is real, but for most questions, someone else is supplying the answer.

How many questions should we check?

One per practice area per city is the minimum: “best {service} in {city}”, phrased the way a client would say it. For your two or three most valuable services, add the follow-up questions buyers ask once they have a shortlist: “{service} lawyer for {type of client}”, “{firm A} vs {firm B}”, “is a specialist {service} firm worth it?”

Test the wording as well as the service. In the check above, “best personal injury solicitors” and “best personal injury lawyer” produced different answers for the same firm. Use the words your clients use, and try the common variants.

Ask each one several times on more than one assistant. Answers vary from run to run, and the assistants often disagree: the same US study found only 2.9% of the firms it recorded were named by all three assistants tested. One screenshot from one assistant tells you very little. A count across runs and assistants tells you where you stand.

For a firm with four practice areas and two office cities, that’s at least eight core questions before follow-ups. That’s normal, and it’s the reason the work has to cover the whole firm, not one page. If your offices are in different cities, check each city separately: AI can recommend you in one and a rival in the next.

Which practice area do we fix first?

Pick the service where three things meet:

1. It’s worth the most to you. High client value, capacity to take on more, a partner who wants to grow it.
2. A rival is clearly named for it, which tells you buyers are asking and the assistants have an answer they like.
3. You’re close. You’re named some of the time, or you already have the raw material: rankings, reviews, case outcomes or talks that simply aren’t connected to that service online.

Then, for that one service, work through the same short list:

• Find the sources behind the answer. Which directories, lists and review sites do the assistants draw on when they name the rival? Are you listed there for this service?
• Describe the service the same way everywhere. Same name, same specialism, same city, on your site, your directory profiles, your partners’ LinkedIn profiles and your Google Business Profile.
• Answer the buyer’s questions on your own site. One page per service that answers the question first, uses the questions clients actually ask as headings, and backs its claims with something only you have: outcomes, fee ranges, timelines, credentials.
• Get others to say it. Reviews that mention the service, a partner quoted on it, a place on the lists the assistants already read.

Then repeat for the next service. Firms with several practice areas or offices usually need this done across the whole site rather than one page at a time. That’s the difference between a single fix and a programme.

Questions partners ask

Sources

1. CLM Sequoia (Custom Legal Marketing), “ChatGPT, Claude and Google AI Mode rarely recommend the same law firms”, data collected 28–29 July 2026, 3,200 prompts across 100 US metros and 32 practice areas. custom.legal (US data; the study’s authors sell legal marketing).
2. QualitySolicitors AI recommendation study, reported in Today’s Family Lawyer, 28 April 2026: over 700 manual tests, 14 prompts, ChatGPT, Google AI Overviews and Claude, seven UK locations. todaysfamilylawyer.co.uk (full methodology not published).