Method note: the consumer figures come from LLM Listed's Q2 2026 survey of 871 US adults, and the citation tables come from BrightLocal's AI directory study. Both companies sell services to firms doing exactly this work, so we label both vendor research and treat the numbers as directional. Compliance note: this article discusses how AI engines describe law firms. It contains no legal advice.

When someone in your city asks ChatGPT for a lawyer, the answer either contains your firm or it does not, and for most firms today it does not. The behavior behind that question is already mainstream: LLM Listed surveyed 871 US adults in Q2 2026 and found 62% had used AI when looking for a lawyer, 96% say they trust AI recommendations, and 84% would not trust a firm that does not appear in AI results at all. Whether those exact percentages travel to your market is worth doubting; the direction is not. A large share of your future clients now get legal recommendations from an engine that has probably never heard of you.

This article is about the fix: where ChatGPT's legal answers actually come from, what moves a firm into the answer, and what to do when the answer describes your firm wrongly. The companion piece, AI search for law firms: what the data shows, covers the studies in depth; here we stay on the corrective side.

What ChatGPT reads before it names a firm

An AI answer is assembled. The engine pulls from sources it already trusts and stitches them into a paragraph. For legal questions those sources are consistent enough to map: BrightLocal's AI directory study recorded 13,354 citations across lawyer-sector AI answers, and Google Business Profile led with 42%, more than five times the next source. Yelp followed at 7.9%, then Justia at 3.2%, Avvo at 2.5%, Lawyers.com at 1.8%, and FindLaw at 0.9%.

FindLaw's white paper, which analyzed legal queries run through ChatGPT, grouped what the answers draw on into four buckets of roughly a quarter each: ratings, reviews, recognition (awards, rankings, published cases), and roots (where you studied, where you practice). In other words, the engine is not looking for your blog. It is verifying a handful of facts about you across platforms you do not control directly.

One more finding reorders most firms' priorities. Martindale-Avvo compared AI answers against Google results and found ChatGPT matched Google's page one less than 25% of the time. Ranking first in local search does not get you named in the answer, and firms that have never outranked anyone still get named. The full citation tables live on our AI visibility for law firms page.

Five fixes that change what engines read

  1. Get your Google Business Profile verified and current. It is 42% of lawyer-sector citations, and 5W PR's study found 78% of legal queries trigger a Google AI Overview, the highest rate of any industry measured. Practice areas, hours, address, and intake information belong there in plain text, kept accurate.
  2. Claim the legal directories ChatGPT actually reads. The most-cited directories in legal answers are Super Lawyers, Avvo, Martindale-Hubbell, and FindLaw, with Justia close behind in the citation table. An unclaimed profile is a fact the engine fills in from somewhere else.
  3. Keep reviews current on the platforms the answer reads. Ratings and reviews together are about half of legal answer content. Volume matters less than recency and responses: an engine summarizing your reputation reads what is newest first.
  4. Make recognition verifiable. Bar listings, awards, published results, and press mentions are the third bucket. If a claim about your firm appears nowhere but your own site, it contributes nothing to the answer.
  5. State your facts plainly on your own site. Practice areas, city, languages, fee structure, who handles what: in sentences, on pages, not in image sliders. The engine needs quotable, checkable text that agrees with your listings.

None of this is speculative engineering. Every item above is a listing you can claim or a fact you can correct this week, on surfaces the citation data says engines already read.

When AI gets your firm wrong

Wrong is worse than absent. In the same LLM Listed survey, 95% of legal consumers said they would not consider a firm if AI presented inaccurate or negative information about it. A wrong practice area, a wrong city, a stale “not accepting clients” note in a directory. Each one actively disqualifies you with the audience that matters.

The correction sequence that works is source-first:

  1. Inventory what the engines say. Ask ChatGPT, Perplexity, and Gemini the questions a real client would ask, and record every description of your firm, right or wrong. One run per engine; save the outputs with dates.
  2. Fix your Google Business Profile first. Edit your profile directly on Search or Maps (Google documents the exact steps in its Business Profile help), and use the “suggest an edit” path on details you cannot control directly.
  3. Correct the directories. Claim the profiles on Avvo, Justia, Martindale-Hubbell, Super Lawyers, and FindLaw; fix name, address, practice areas, and status on each. Where a listing is not yours to edit, submit a correction request and keep the confirmation.
  4. Re-run the same questions monthly. Engines refresh their sources on their own schedule. The measure of progress is simple: the wrong details stop appearing, and the right ones start.

One compliance line, because it matters here: ABA Model Rule 7.1 prohibits false or misleading communications about a lawyer's services. Everything you correct in an AI's source pool is a communication about your services. Fix facts; never plant flattering ones.

What not to do

Do not keyword-stuff your profiles, buy reviews, or publish claims you cannot substantiate on any surface an engine reads: engines cross-check sources, and contradictions across them are exactly what produces vague or wrong descriptions. The durable move is boring: identical, verifiable facts everywhere your clients already look.

To see whether AI engines can read your site at all (the precondition for any of the above) run CiteGap's free scan. It checks seven technical aspects of your homepage in seconds, no signup.

Get the written diagnosis ($490)

Frequently asked questions

Do AI engines actually recommend specific law firms?

Yes. LLM Listed's Q2 2026 survey of 871 US adults found 62% have used AI when looking for a lawyer and 96% say they trust AI recommendations. 5W PR's study of legal AI visibility found 78% of legal queries trigger a Google AI Overview, the highest rate of any industry measured. Both sources are vendor research, so treat the exact figures as directional.

Which sources do ChatGPT legal answers read?

In BrightLocal's lawyer-sector table (13,354 observed citations), Google Business Profile led with 42%, followed by Yelp at 7.9%, Justia at 3.2%, Avvo at 2.5%, Lawyers.com at 1.8%, and FindLaw at 0.9%. FindLaw's white paper grouped ChatGPT's legal answer content into four roughly equal buckets: ratings, reviews, recognition, and roots. Being ranked first on Google does not settle it: Martindale-Avvo found ChatGPT matched Google's page one less than 25% of the time.

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

No honest answer exists, and we do not promise timelines. Engines update their sources on their own schedule, and answers shift between runs. What you can measure is direction: run the same buyer-style questions each month, keep a dated record, and check whether your firm's descriptions become accurate and your citation footprint grows. The free scan is a first pass; the written diagnosis runs the full three-engine version.

AI got my firm's details wrong. What do I fix first?

Fix the source, not the engine. 95% of legal consumers in LLM Listed's survey said they would not consider a firm if AI presented inaccurate or negative information, so accuracy is the bottleneck. Correct your Google Business Profile first (42% of lawyer-sector citations), then the legal directories ChatGPT reads (Avvo, Justia, Martindale-Hubbell, Super Lawyers, FindLaw), then request corrections on any wrong directory listing. Never publish claims you cannot substantiate: ABA Model Rule 7.1 applies to your website and your listings alike.

Sources