AI search for law firms: why the engines name the other firm, and the 12-point fix list.
Someone in your city typed this into ChatGPT last night: "My mother passed without a will and my brother is contesting the house. Which estate lawyer in [county] should I call?" The engine answered with three firms, a sentence on each, and the sources it used. One of those firms will get the call. It probably wasn't chosen because it has the best lawyers. It was chosen because the engine could find, confirm and quote enough about it to name it with confidence.
I audit law-firm websites for a living, and the gap between the firms that get named and the firms that don't is rarely about reputation or budget. It's about twelve specific, fixable things. Here they are, with the prompts to run first so you can see where your own firm stands.
How the engines decide which firm to name
ChatGPT (with search on), Perplexity, Gemini and Google's AI Overviews and AI Mode all do a version of the same thing: retrieve pages relevant to the question, extract the facts they can support, and write a short answer that cites or links to the sources. For "which lawyer should I call," the answer is a shortlist. To make the shortlist, a firm needs four things the engine can find:
- A stable identity. One name, one set of facts, everywhere. Firms whose review count, founding year, office list or even name vary from page to page get hedged or skipped.
- Corroboration. Third parties that say the same thing the firm says: Google reviews, Avvo, Justia, FindLaw, Martindale, Super Lawyers and Best Lawyers listings where they apply, the state bar directory, local press, Reddit threads. The engines cite these constantly, and a firm that exists only on its own website has nothing to be corroborated by.
- Quotable pages. Sentences that state who, what, where, since when and for how much, in a form the engine can lift without editing. "We're passionate about protecting families" can't be quoted. "Flat-fee estate plans from $1,800, two meetings, documents signed within three weeks, Onondaga and Madison counties since 2009" can.
- Verifiable people. Lawyers with bar numbers, law schools, admissions and bylines on the firm's guides. The engines weight authorship, and a licence number they can check against the state's directory is the strongest kind.
Google's own guidance on its AI features, published in May 2026, says these features run on the same core Search ranking and quality systems as the rest of Google, and that content should bring "real experience" rather than summarize what already exists. For a law firm that's good news: the work is mostly the SEO work you should have been doing, plus a layer of fact hygiene most firms have never done.
Run these ten prompts first
Before fixing anything, find out what the engines say today. Run each prompt on ChatGPT with search enabled, Perplexity, Gemini, and a normal Google search (to see whether an AI Overview appears and what it says). Log whether the firm is named, who is named instead, and the sources cited.
1. Who are the best [practice area] lawyers in [city]?
2. I need a [practice area] attorney in [county/state]. Who should I call and why?
3. How much does a [practice area] lawyer cost in [state]?
4. Is [your firm] a good law firm? What do clients say?
5. Compare [your firm] and [competitor].
6. Which [practice area] firms in [city] offer flat fees?
7. My [matter] was denied / rejected / dismissed. Which lawyer in [state] handles that?
8. What should I look for in a [practice area] lawyer?
9. Who is [lead attorney] and what is their background?
10. [A question your intake team answers every week, word for word]
Forty answers, one spreadsheet. The sources column is the diagnosis: it tells you which directories, review platforms and pages the engines trust for your practice area in your market. You don't have to guess where to build presence. They just told you.
The short version of this check is here; the ten prompts above are the law-firm version.
Want the twelve points checked on your firm?
Send the firm's website and your closest competitor. You'll get the ten prompts run across four engines, who gets named, and the twelve-point scorecard with the fixes in order. Free, written, within 24 hours.
Prefer to talk? 315.430.0782. I answer my own phone.
The 12-point fix list
In the order I'd do them, which is roughly cheapest-and-most-blocking first.
1. One page per intent, and no near-duplicates in the index
The most common structural problem I find on law-firm sites: several versions of the same page live and indexed. A paid landing page that's a copy of the homepage. A "Toronto lawyers" page and a "lawyers in Toronto" page. Two versions of the practice-area page from two agencies. A firm I audited this month had six near-copies of its homepage in Google's index, each with slightly different numbers. The engine has to guess which one is the firm, and it guesses differently week to week. Fix: one canonical page per intent, redirects from the rest, and canonical and noindex tags that agree with each other.
2. The fact sheet, applied everywhere
One page listing every fact about the firm that should be identical across the web: exact name and punctuation, legal name, founding year, offices, phone per office, practice areas with prices or ranges, review count by platform with the date, the principals and their bar numbers. Then every page, bio, listing and piece of markup is checked against it. The template is in the fix-it guide. This is a morning's work and it's the single highest-leverage item on the list.
3. LegalService and Person schema, with bar numbers
A LegalService block for the firm (not generic LocalBusiness), a Person block for every lawyer with job title, bar or licence number, law school, admissions and memberships, and Service blocks on each practice-area page with the fee where there is one. Google says structured data isn't required for its AI features; what it does is state your facts in a form that can't be misread, and for a firm with fourteen lawyers it's the only practical way to tell the engines who each one is. Copy-paste blocks are here.
4. Bios that say verifiable things
Year admitted, courts, bar associations, the practice areas they actually handle, a count of matters where you can state one honestly, and a byline link to the guides they've written. The adjectives can stay; they just don't do anything.
5. Practice-area pages that answer the buyer's question in the first screen
Who it's for, what the firm does, where, since when, what it costs or how fees work, what happens next. Two or three plain sentences near the top that a model could lift verbatim. The rest of the page can be as long and as persuasive as you like; the quotable part is what gets you named.
6. Prices or ranges on the page, where the rules allow
"How much does a [practice area] lawyer cost in [state]" is one of the most-asked questions in every market, and the engines answer it from pages that say. Flat-fee practices should publish the fee and what it includes. Hourly practices can publish a range and a typical total for common matters. Firms that won't say get represented in the answer by a competitor's number.
7. Reviews where the engines look
Google first, because every engine reads it. Then whichever platforms the sources column showed for your practice area: Avvo and the state bar directory come up constantly for consumer-facing areas; Martindale and peer-review listings more for business and litigation. Build the review request into the close of every matter so it's a system, not a favour. And make sure the count on your site matches the platform, dated.
8. Listings that agree with the fact sheet
Same name, address and phone on Google Business Profile, Bing Places, Avvo, Justia, FindLaw, Lawyers.com, the state bar directory, the county bar, the chamber, and whatever else the sources column named. Bing matters more than its search share suggests because ChatGPT's search draws on Bing's index among its sources. Register in Bing Webmaster Tools while you're at it.
9. A real FAQ, in the buyer's words
Pull the twenty questions intake answers every week and answer them on the site, plainly, with a specific in each answer (a timeframe, a price, a court, a county). Mark it up as FAQPage. Google rarely shows FAQ dropdowns anymore; the AI engines lift question-and-answer pairs constantly, which is the reason to do it.
10. Guides written by the lawyers, with their names on them
The cost guides, process guides ("how probate works in [county], step by step"), and the problem guides ("what to do when your [application] is denied") are what the engines cite when a buyer asks a question that isn't "who should I hire" yet. Google's guidance is explicit that content "easily produced by a generative AI model" adds little; a guide written from a hundred real matters, under a named lawyer's byline with a Person block behind it, is the opposite of that. Draft with AI if you like; the lawyer edits, fact-checks and signs it.
11. The problem cluster nobody owns
In every practice area there's a set of searches from people whose matter has gone wrong: denied, rejected, dismissed, audited, revoked, missed a deadline. They're urgent, they convert, and almost no firm writes for them because the pages don't feel like marketing. Write them. They're often the first thing a firm gets named for, because there's nobody to compete with.
12. Measurement that isn't a feeling
The ten prompts, four engines, once a month, in the same spreadsheet. Search Console's Generative AI performance report (rolling out since June 2026) for how often your pages appear in AI Overviews and AI Mode. Call tracking and booking events so organic and AI-referred leads are counted by page. Without this, "are we showing up in AI" stays an opinion, and opinions don't get budget.
What the advertising rules mean for all of this
I'm not a lawyer and this isn't legal advice; you know your jurisdiction's rules better than I do. But the pages the engines quote are the same pages a disciplinary reviewer reads, and the firms that get named are, in my experience, the ones whose claims are stated in verifiable form anyway. A few patterns worth a second look:
- Superlatives without a basis. "The #1 rated firm in [state]" with no source. "Rated 4.9 on Google across 300 reviews" says the same thing in a form anyone can check, and the engine can quote it.
- "Only" and "exclusively." Dispute magnets. "We practice [area] exclusively" keeps the point without the fight.
- Success rates. Powerful if documented; costly if challenged. A count of matters handled is a number; a rate needs a period and a denominator.
- Guarantees. If you offer one, put the terms one sentence under the button. It reads as more real, not less.
- Testimonials and past results. Some states require specific disclaimers (New York's rules, for example, call for a "prior results do not guarantee a similar outcome" statement in ads that use them). Check yours.
- Bar logos under "happy clients." Move them to a credentials row so they don't read as endorsements.
None of this makes the site weaker. Stated in verifiable form, the same claims are both safer and more quotable, which is the whole game.
The 30-day plan
| Days | Do | What it changes |
|---|---|---|
| 1–5 | Run the ten prompts on four engines; read access to Search Console and the Business Profile; write the fact sheet | You know who gets named, from which sources, and you have one version of the truth |
| 6–15 | Consolidate duplicate pages; fix canonical and noindex; apply the fact sheet to every page; add LegalService, Person and Service schema; set up call tracking and booking events | The engines see one firm with one set of facts and verifiable people |
| 16–25 | Rewrite the first screen of each practice-area page to be quotable; publish prices or ranges; rebuild the FAQ from intake questions; correct every listing the sources column named | There is now something worth quoting, and third parties agree with it |
| 26–30 | Outline the problem-cluster guides and assign a lawyer to each; put the review request into the matter-close checklist; schedule the monthly prompt check | The compounding part starts, with a way to measure it |
After that: two to four guides a month under lawyers' bylines, the prompt log re-run monthly, the listings and reviews maintained as a system, and a quarterly look at whether paid is still buying what organic now wins. That last check tends to fund the rest.
Questions I get about this
Do AI engines actually recommend specific law firms?
Yes, when the question asks for one. "Who should I hire for X in [city]" returns a short list with reasons and citations on ChatGPT with search, Perplexity and Gemini, and Google's AI Overviews and AI Mode increasingly do the same for local service queries. The list is short, three to five names, and it's drawn from firms whose facts the engine could corroborate.
Is this different from SEO?
Mostly no. Google's May 2026 guidance says its AI features run on its core search and quality systems and that optimizing for them "is still SEO." The additions are the ones in this guide: consistent facts, quotable pages, corroboration where the engines look, and a monthly check of what they say. Firms that rank well organically are usually most of the way there.
Can we put prices on the site under the advertising rules?
In most jurisdictions, yes, as long as the price is accurate and not misleading about what it includes. Flat-fee practices do it routinely. Check your state's rules and your malpractice carrier's guidance; the usual requirement is clarity about what the fee covers and what it doesn't.
We have great Google reviews. Why doesn't the engine mention them?
Two usual reasons: the count on your site doesn't match the count on Google (so the engine can't be sure which is right), and the reviews aren't where the engine looked for your practice area. Run the prompts and read the sources column: if it cites Avvo and the state bar directory and you've never touched either, that's the gap.
How long does it take to get named?
The identity and schema fixes show up in search-grounded answers within weeks. Corroboration and content take longer, typically a quarter before the prompt log shows the firm consistently. It compounds from there, which is the argument for starting with the cheap fixes this month.
Does our Google Ads budget affect AI answers?
Not directly; ads don't appear inside the AI answer as recommendations. Indirectly, yes: if the paid pages are near-duplicates of the organic ones and indexed, they muddy the firm's identity for the engines. The paying-for-clicks guide covers consolidating them.