On This Page
TL;DR Google ranks pages. AI tools recommend firms by name, and each one picks differently. We ran 38,000 “best lawyer” searches across ChatGPT, Perplexity, Gemini, and Claude in 15 cities, and found every platform crowns a different short list, built mostly from directories like Super Lawyers and Justia.
Open ChatGPT tonight. Type the question a hurt or scared person types: best [your practice] in [your city]. Read the names it gives back.
Then do the same in Perplexity. Then Gemini. Then Claude.
Four tools. Four different answers. And for a lot of good firms, their name isn’t in any of them.
That’s the thing nobody told you was coming. The front page moved, and it stopped looking like a list of links.
The Difference Between A Ranking Engine And A Recommendation Engine
For 20 years the game was Google, and Google is a ranking engine. It hands a searcher ten blue links and steps back. The searcher does the choosing. You could sit at number four, number six, even down the page, and still get the call, because the person scanned, compared, and made up their own mind.
A recommendation engine works a different way. You ask it for the best lawyer in town and it hands back three names and a reason for each. It already did the choosing. There’s no page two to scroll. There’s no list of ten to compare. There’s a short answer, and you’re either in it or you’re not.
That’s what ChatGPT, Perplexity, Gemini, and Claude are. Recommendation engines. And the move from ranking to recommendation changes the whole job. Ranking rewards the page. Recommendation rewards the name the machine already trusts.
So we stopped guessing about it and ran the numbers.
What We Did
Over about six weeks, from late April into early June 2026, we asked four AI tools the questions real clients ask. Best personal injury firm in this city. Best divorce lawyer in that one. Best business attorney, best immigration lawyer, on down a list of 25 practice areas across 15 cities: Atlanta, Austin, Boston, Charleston, Charlotte, Chicago, Columbia, Dallas, Denver, Houston, Los Angeles, Miami, New York, Raleigh, and San Francisco.
We captured 44,411 answers. Some came back as platform errors, quota limits and server hiccups, and we threw those out: 6,238 of them, about 14%. That left 38,173 clean answers to read. Inside them sat 96,988 times a real firm got named, spread across more than 15,000 different firms.
Here’s what the data said.
Finding 1: Every Platform Crowns A Different Short List
This was the one that stopped us cold.
We expected the four tools to mostly agree. They don’t. Ask ChatGPT and Perplexity the same question, in the same city, for the same practice area, and their lists of named firms overlapped about 6%.
Six percent. Almost no agreement at all.
Think about what that means for a firm. You could own ChatGPT in your market and be a ghost on Perplexity. You could be the first name Claude gives and a name Gemini has never heard. There’s no single “AI result” to win. There are four front pages, and they’re reading from four different scripts.
So the firm that says “we tried the AI thing, we show up in ChatGPT” is reading one quarter of the board and calling the game.
Finding 2: In Any One City, A Few Names Take The Top
Inside a single city and practice area, the answers crowd around a few firms.
The top 3 firms took about 34% of every name the tools handed out. The single most-named firm in a given market averaged 13% on its own. Below that sat a long tail, around 64 different firms per city and practice area, most of them mentioned once or twice and then gone.
So it’s top-heavy at the front and crowded in the back. A handful of names soak up a third of the attention, and everyone else fights over scraps.
Real examples, straight from the study. These are the firms the tools named most often, not our ranking and not a verdict on who’s the better lawyer.
- In Atlanta personal injury, the top three were Morgan & Morgan, Kaine Law, and Butler Law Firm.
- In Chicago divorce, three names took nearly half the answers: Schiller DuCanto & Fleck, Beermann, and Berger Schatz.
- In Dallas business law, Vinson & Elkins, Haynes and Boone, and Jackson Walker.
- In Houston personal injury, Arnold & Itkin led, then Zehl & Associates and The Krist Law Firm.
- In Los Angeles criminal defense, Eisner Gorin, Geragos & Geragos, and Werksman Jackson & Quinn. In Miami immigration, Revilla Law Firm, Kurzban Kurzban Tetzeli & Pratt, and Pozo Goldstein.
- In Boston personal injury, Breakstone, White & Gluck, then Sweeney Merrigan and Lubin & Meyer. Down in Charleston, the Steinberg Law Firm.
- Up in Raleigh family law, Charles R. Ullman & Associates and the Rosen Law Firm.
The pattern held in every market we studied.
- In New York, the corporate names ran the board: Sullivan & Cromwell, Paul Weiss, and Skadden.
- In San Francisco, Morrison & Foerster, Orrick, and Cooley. In Denver, Holland & Hart and Brownstein Hyatt.
- In Austin, Jackson Walker and Graves Dougherty Hearon & Moody.
- In Charlotte, Moore & Van Allen and Parker Poe.
- In Columbia, Nelson Mullins and Haynsworth Sinkler Boyd.
Different cities, same shape every time. A short stack at the top, and a long line of firms the tools mention once and forget.
Finding 3: Gemini Mostly Refuses To Name Anyone
One tool barely played.
Asked the same buyer questions, Gemini named an actual firm in only about 12% of its clean answers. The other 88% of the time it gave a little lecture on how to pick a lawyer, things like “check reviews” and “schedule consultations,” and named no one at all.
ChatGPT and Claude named firms about 99% of the time. Perplexity, 96%. Gemini, 12%.
That’s not a small footnote. It tells you these systems don’t behave alike, and a plan built for how one of them acts can fall flat on another.
Finding 4: Directories Write The Answer
Here’s the part that points straight at what to do.
Perplexity shows its sources, so we read every one. Across more than 30,000 cited sources, the same short list of directories carried most of the weight. Super Lawyers showed up in 15.6% of all citations. Justia in 13.2%. Those two alone were close to a third of every source the tool leaned on. Add Best Law Firms, Best Lawyers, BCG Attorney Search, and Chambers and the top six were nearly half.
Look at it from the firm’s side and it’s sharper still. 96% of Perplexity’s answers cited at least one directory. Only 68% cited a law firm’s own website.
So when the AI says your competitor is the best injury firm in town, it usually isn’t reading their About page and deciding they’re great. It’s reading what Super Lawyers and Justia say, and repeating it. The directories you maybe ignored for years are the AI’s reading list.
Being Named Is Not The Same As Being Better
Say this plainly, because most agencies won’t.
The study counts who gets mentioned and who gets cited. It says nothing about who wins cases, who treats clients right, or who you’d actually want in your corner. A firm with a thin record and a strong directory presence will out-appear a stronger firm that’s invisible to the machines.
So if you’ve watched a weaker competitor get named while you get skipped, you read it right. They didn’t out-lawyer you. They got recognized, and you didn’t.
That should sting. Good. Because recognition is the part you can fix.
What To Do If You’Re Not Coming Up
Here’s the work. None of it is a trick, and there’s no secret prompt. It’s plumbing, and the whole house runs on it.
- See where you actually stand. Open all four tools and ask each one the question a client would: best [your practice] in [your city]. Write down who gets named. On Perplexity, write down which sites it cites. That’s your starting scoreboard, and it’s free.
- Get onto the sources the AI quotes. The names that keep coming up live on Super Lawyers, Justia, Best Lawyers, Best Law Firms, Chambers, Avvo, and Martindale. Claim those profiles. Fill in every field. These aren’t vanity badges. They’re what the machine reads before it answers.
- Make your firm one consistent entity. Your name, address, phone, and practice areas should read the exact same way on your website, your Google Business Profile, and every directory. When those details disagree, the AI gets unsure about who you are, and unsure means unnamed.
- Write plain pages for what you do and where you do it. One page per practice area, one per city you serve, in the words a client would use. “We handle [practice] for people in [city].” Not clever. Clear. The machine has to be able to tell, in seconds, exactly what you do and exactly where.
- Earn outside mentions. Local press, bar association pages, legal publications, podcasts, guest articles. Every credible site that names you is one more vote the AI can count when it builds its answer.
- Win each engine on its own. Remember the 6%. Showing up in ChatGPT tells you almost nothing about Perplexity. Check all four, find the ones where you’re missing, and work those gaps one at a time.
- Keep showing up. These tools learn by repetition. Every month your name is out there in the sources they trust, the pattern hardens in your favor. Every month it isn’t, it hardens for someone else.
The New Front Page
The old front page was ten links and a searcher who made up their own mind. The new one is three names and a machine that already decided.
Tonight, someone in your city is going to open one of these tools and ask it to name a lawyer. It will give them a short answer, fast, with no page two.
The only question worth asking is whether your name is in it.
