Blog

How to Get Your Law Firm Found in ChatGPT

Legal is one of the sectors where people will search through AI the most, in English and in Spanish. You don’t hire a lawyer very often, so word of mouth and personal referrals count for less than in other services. People look for references that are impartial, or that they believe are impartial, and that is the role ChatGPT, Claude and Gemini now play.

Right now it is still easy to get positioned through press articles, lawyer directories and public or media appearances. Firms need to understand that link. When someone asks an AI “who’s the go-to divorce lawyer in San Diego?”, it will recommend the firms that invested now, while it is still early. Almost anyone can still get in, but it is a game you have to start now, with a mid-term view.

What did we ask, and how?

We ran the test twice. On October 1, 2026 we asked four questions the way a client would type them, across divorce, personal injury, employment and estate planning in four US cities. Each went once to ChatGPT (with search), Perplexity and Gemini. On October 19 we asked the same four again on the same three engines, and added an immigration question for Los Angeles, asked once in English and once in Spanish. That is 30 answers. For each one we logged the firms named and, where the engine showed them, the sources behind the answer.

QuestionFirms named Oct 1Firms named Oct 19Named on both datesNamed by all three engines
Who is the best divorce lawyer in San Diego?9124None, then none
Best personal injury lawyer in Houston14104Zehl & Associates, then none
I need an employment lawyer in Denver. Who would you recommend?8107None, then none
Top estate planning attorney in Chicago9155Peck Ritchey, then none
Who is the best immigration lawyer in Los Angeles, California?Not asked16Not askedNone
¿Quién es el mejor abogado de inmigración en Los Ángeles, California?Not asked16Not askedNone

The numbers are different firms named across the three engines. “Named on both dates” counts a firm if any engine named it on October 1 and any engine named it on October 19.

Do AI engines recommend the same lawyers twice?

Not for long. On October 1 the three engines gave 54 recommendations between them, counting each engine’s list separately. Eighteen days later, the same engine repeated 21 of them. Perplexity changed the most: 3 of its 12 names survived. In San Diego it went from one firm to five different ones, and in Chicago it replaced all four of its names with six new ones, each cited to the same Expertise.com list.

Bar chart: of the firms each engine named on October 1, 2026, Perplexity named 3 of 12 again on October 19, ChatGPT 9 of 23 and Gemini 9 of 19. Firms named by all three engines went from 2 to 0.

Some answers held. ChatGPT named the same four Denver employment firms it had named on October 1 (Swain Law, Rathod Mohamedbhai, Lowrey Parady Lebsack DeFazio and Murray Law) and dropped a fifth. Gemini repeated all four of its Chicago estate planners and added two more.

The two firms all three engines agreed on in the first round each lost one engine. Zehl & Associates is still named by Perplexity and ChatGPT for Houston personal injury, but Gemini’s list was now Arnold & Itkin (also named on October 1), The Lanier Law Firm and Perdue & Kidd. Peck Ritchey is still named by ChatGPT and Gemini in Chicago, but not by Perplexity. On October 19, not one of the six questions had a firm named by all three.

It moves within the hour, too. We asked ChatGPT the English Los Angeles question a second time about ten minutes later: of the seven firms across the two runs, two appeared in both.

That is what “still early” looks like in the data. A single check tells a firm very little. What counts is how often it appears across many runs, which is why we measure AI share of voice with repeated runs instead of one screenshot.

In Perplexity, completely. We asked “Who is the best immigration lawyer in Los Angeles, California?” and the same question in Spanish, minutes apart. Perplexity named five firms each time, and not one appeared in both answers. In English, 1 of its 10 sources was a law firm’s own website; the rest were Avvo, Super Lawyers, Lawyers.com, Best Lawyers, Justia, Expertise.com, Yelp and Forbes Advisor. In Spanish, 7 of its 10 sources were firm websites, and six of those were Spanish-language pages. It described its first pick as a firm with a strong Los Angeles presence and Spanish-language content.

Perplexity's answers to the same Los Angeles immigration question in English and in Spanish: two different shortlists with no firm in common

Perplexity, October 19, 2026: the English question (top) and the Spanish one (bottom). Different firms, and different kinds of sources behind them.

ChatGPT and Gemini moved less. ChatGPT kept three firms across both languages and, in Spanish, added one whose public profile, in its words, confirms service in Spanish. Its second English run also named that firm, so for ChatGPT the language gap was within its normal run-to-run variation. Gemini shared two names out of twelve, and its Spanish answer opened with Fragomen, a corporate immigration firm, and never mentioned Spanish-language service. Across the three engines, the English and Spanish answers named 16 firms each and only 6 in common.

For a firm with Spanish-speaking clients, the lesson is concrete. In Perplexity, the Spanish question is a different search with different winners: four of its five picks were cited through Spanish-language pages on their own sites.

What do the engines read before naming a firm?

Mostly directories, rankings and review sites. Across both rounds the engines leaned on Expertise.com, Super Lawyers, Avvo, Justia, FindLaw, Best Lawyers, Lawyers.com, Yelp, Forbes Advisor and Trust Analytica, plus local “best of” pages such as a 2026 list of San Diego divorce practices and a “10 best personal injury lawyers in Houston” page.

ChatGPT built its October 1 San Diego answer on Doyle’s Guide, a peer-reviewed legal ranking, and its Chicago answer on Best Law Firms and Super Lawyers.

ChatGPT's answer to "Who is the best divorce lawyer in San Diego?", built on Doyle's Guide's 2026 peer-reviewed rankings

ChatGPT, October 1, 2026: “The strongest current signal is Doyle’s Guide’s 2026 peer-reviewed San Diego family/divorce rankings.”

The mix is not fixed either. Perplexity’s sources for San Diego went from zero firm websites on October 1 to seven of ten on October 19. In Houston they went the other way, from seven of ten to two.

Press, bar associations and credentials fill in the rest. On October 1, one of Perplexity’s Houston sources was a press release on Business Wire, and in Denver it cited the Colorado and Denver bar associations. On October 19, in six of the 18 answers, a certification as a specialist in family law or immigration law was given as a reason to consider a lawyer. Credentials came up again in our November test of Denver dentists, where the engines repeated oral surgeons’ and orthodontists’ board certifications.

What has ChatGPT changed since October 1?

On October 1 it opened every answer with a map of firms and their Google ratings. On October 19 the map appeared in three of six answers; the others used profile cards, some with star ratings, review counts, address and phone. Gemini showed Google Maps listings in all four questions on October 1 and in two of six on October 19.

The bigger change is how ChatGPT ends. All six of its answers closed with a short form to narrow the list: the type of case every time, then questions such as budget or estate size, urgency, whether you are the employee or the employer, in-person or virtual meetings, and in three of the six, the language you prefer.

ChatGPT's intake form at the end of its answer: type of immigration case and preferred language

ChatGPT, October 19, 2026, end of the second run of the Los Angeles question.

That form is the brief your website has to answer. If ChatGPT is matching firms to “Spanish preferred”, “employee side” or “urgent deportation”, the firm that states those things plainly gives it something to match. Denver shows the difference: ChatGPT said it prioritized employee-side practices rather than relying on star ratings alone, while Perplexity’s answer to the same question included two firms it described as primarily or often on the employer side.

What didn’t fit?

Two things, and a firm should know both.

The firm’s own website gets cited more than “directories decide everything” suggests. On October 19, 8 of Perplexity’s 10 sources for Denver were firm websites, and 7 of 10 for San Diego. On October 1, ChatGPT quoted one Denver firm’s own line that it represents employees exclusively. A page that says plainly who you represent is quotable.

A low rating does not keep a firm off the shortlist. On October 19, ChatGPT shortlisted two Chicago estate planners that showed 3.4 and 3.8 stars on its own map, and on October 1 Gemini listed a San Diego firm with 3.0 stars. The rating is one signal among several, not a filter.

Limits of this test: one run per question, engine and date, on logged-in accounts, so answers can change between runs and between users, as the second ChatGPT run shows, and they can be personalized to the test account and connection. On October 19 we used a ChatGPT temporary chat, so earlier conversations could not shape the answer; on October 1 we used regular chats. Gemini ran in Flash on October 1 and in automatic mode on October 19. Personalization still leaks in: one Perplexity answer closed by asking about assets in two countries that came from the test account, not from the question. On October 1, ChatGPT’s first reply to the Houston question named nobody and only answered after a follow-up. Read it as two snapshots, not a ranking.

What should a law firm do now?

Be present and consistent in the places AI engines read before they name a lawyer, and make your own site answer the questions they now ask:

  1. Directories and rankings. Claim and complete Super Lawyers, Avvo, Justia, FindLaw, Best Lawyers, Lawyers.com and Expertise.com, and find the local “best of” pages that rank for your city and practice area. Same firm name, address, practice areas and jurisdictions everywhere.
  2. Peer rankings and press. Doyle’s Guide, Best Law Firms and Super Lawyers were cited directly, and Gemini named Chambers and U.S. News Best Law Firms as the recognition behind its picks. One wire press release was enough to become a source in Houston.
  3. Credentials the engines can quote. If you are a certified specialist, say so on your site and every profile, in the certifying body’s exact wording. Bar sections, talks and media comments count too.
  4. Pages that answer ChatGPT’s intake form. Who you represent, which case types, languages spoken, fee model (contingency, flat fee, free consultation), how fast you respond, in-person or virtual. One page per practice area and city.
  5. A Spanish-language page if you serve Spanish speakers. Written for that client, not run through a translation widget. In Perplexity, that is what the Spanish question cited.
  6. Google Business Profile. Correct category, consistent details and real reviews. ChatGPT and Gemini still pull ratings and maps from it in many answers. ChatGPT has had business profiles of its own since November 2026, but the Business Hub beta doesn’t have an option built for service firms yet.
  7. Measure every month, per engine. The same questions on each engine, more than once. In 18 days, 33 of the 54 recommendations changed.

Why now: Clio’s 2025 Legal Trends Report found that more than half of consumers have used or would consider using AI for a legal question, and 28% of those who used it were directed to contact a lawyer. Meanwhile a paid click for “personal injury lawyer near me” costs $131.33 in the US, and “geo for law firms” has a keyword difficulty of 16 (Semrush, US database, pulled September 30, 2026). Being named in the answer skips one of the most expensive auctions in advertising, and very few firms are competing for it yet.

The playbook is not unique to legal, but the sources are. Compare it with what moves SaaS (review platforms and comparison pages) or ecommerce (product schema and marketplaces), and with how Perplexity picks its citations in general. Accounting firms looked similar on October 13, 2026: local answers leaned on Google Business Profiles and directories, and at most one firm per city was named by all three engines (GEO for accountants). For hotels and vacation rentals on October 19, the properties’ own websites carried most of the weight, not the booking platforms (direct bookings from AI search).

Frequently asked questions

Can ChatGPT actually refer clients to a law firm?

Yes. Clio’s 2025 Legal Trends Report found that 28% of consumers who used AI for a legal question were directed to contact a lawyer. In our October 19, 2026 test, every ChatGPT answer ended with a short form to narrow the list to the user’s case, asking about the type of case and details such as budget, urgency or preferred language.

Which sources do AI assistants use to recommend lawyers?

In our October 2026 tests: legal directories (Super Lawyers, Avvo, Justia, FindLaw, Lawyers.com, Expertise.com), peer rankings (Doyle’s Guide, Best Law Firms, Best Lawyers), local “best of” lists, review sites (Yelp, Forbes Advisor), press, bar associations, Google ratings and the firms’ own websites. The mix changed between runs, even for the same question.

Do AI recommendations for lawyers change over time?

Yes, quickly. We asked ChatGPT, Perplexity and Gemini the same four questions on October 1 and October 19, 2026. Only 21 of the 54 recommendations from the first round, counting each engine’s list separately, were repeated by the same engine. Two firms were named by all three engines on October 1; on October 19, none.

It can change who gets named. For the same Los Angeles immigration question, Perplexity named five firms in English and five different firms in Spanish, and 7 of its 10 Spanish sources were firm websites, six of them Spanish-language pages. Across ChatGPT, Perplexity and Gemini, only 6 of 26 firms appeared in both languages.

Do Google reviews matter for AI recommendations?

They matter as a way in, not as a ranking. ChatGPT showed star ratings and review counts in most answers, and both ChatGPT and Gemini used Google maps and listings. But ChatGPT also shortlisted firms with 3.4 and 3.8 stars, so a low score did not keep a firm off the list.

How long does it take for a law firm to show up in AI answers?

Directory and profile fixes are the fastest lever, because engines re-check those sources often. First citations usually take 30 to 60 days, and competitive practice areas take longer. Measure monthly with the same questions per practice area and city, the way we do in how we work.

See where your firm stands today

Before changing anything, find out what the engines say when someone asks for a lawyer in your practice areas and cities. Our free AI visibility diagnosis runs questions like these across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, and shows which firms they name, which sources they cite, and where yours does or does not appear.

If AI doesn’t name you, it names a competitor.

The free diagnosis runs real buying prompts on five engines, shows who gets named instead of you and where you can win first.