How People Find a Lawyer Now: Referrals, Maps, and AI

Ask ten partners how clients find them and nine will say referrals. They are right, and it is the least useful true thing in legal marketing. Because a referral is not a hire. It is a nomination — a name handed over at a barbecue, in a doctor's office, or in a group chat.

What happens in the next four minutes decides whether that nomination survives. The prospect types your name into a phone. And in 2026 that phone increasingly answers back — not with ten blue links, but with a paragraph a machine wrote about you, assembled from sources you have probably never audited.

This post traces the actual path from legal problem to signed retainer, names the machine standing at each step, and gives you a free 20-minute routine to see what the machines currently say about your firm.

How does someone actually go from problem to hiring a lawyer?

The two channels overlap rather than compete. Clio's 2019 Legal Trends Report found that 59% of clients sought a referral from someone they know, 57% searched on their own through some other means, and 16% did both — which means referral and search are not two separate audiences. They are two steps in one journey for a large share of people.

Add up 59 and 57 and you get more than a hundred. That is the whole point. A referred prospect is also a searching prospect. Here is the sequence as it actually runs:

• The trigger. A crash, a termination letter, a served petition, a spouse moving out. The problem shows up before the search does.
• The private question. Before asking a human, many people now ask a machine. Clio's 2025 Legal Trends Report found more than half of consumers have used or would consider using AI to answer their legal questions — and more than half of those who did felt the AI sufficiently answered it.
• The nomination. They ask a friend, a colleague, a prior lawyer, or a Facebook group. One or two names come back.
• The verification. They Google the name. This is where your map pack listing, your reviews, and now the AI answer either confirm the nomination or quietly kill it.
• The shortlist. If the name does not survive verification — or if nobody gave them a name at all — they search for the category instead: 'personal injury lawyer near me,' 'best divorce attorney Tampa.' Now you are competing with everyone.
• The call. Whoever answers the phone, and how fast, closes the loop.

Firms spend their whole budget on step five and their whole reputation on step three. Step four — verification — is where the money leaks, because nobody owns it and nobody measures it.

Does the referral still win if the search results contradict it?

Frequently, no. BrightLocal's 2026 Local Consumer Review Survey of 1,002 US consumers found that 47% of consumers will not use a business with fewer than 20 reviews, and 31% will only use a business rated 4.5 stars or higher — up from 17% the previous year. A warm referral into a 3.6-star profile with eleven reviews is a referral you lose.

Freshness matters as much as volume. In the same survey, 74% of consumers said they only care about reviews written in the last three months. That makes review velocity — a steady trickle of new reviews — the thing to manage, not your lifetime count. Forty reviews from 2022 read like a firm that stopped practicing.

Silence reads badly too. BrightLocal found 89% of consumers expect business owners to respond to reviews, and 42% say they are unlikely to use a business that never replies. For a law firm this is delicate — you cannot discuss a matter — but 'Thank you for the feedback. We take every concern seriously and would welcome a call to discuss it' is a response, and it is bar-safe in most jurisdictions. Check your state's rules before you write a policy.

The verification step is mostly a local-search problem, and it is winnable. Our guide to winning the map pack for law firms covers the profile-level work — category, service pages, proximity, review cadence — that decides whether a referred name shows up looking like the obvious choice.

What does ChatGPT use when someone asks it to recommend a lawyer?

Third-party sources it already trusts — almost never your website's about page. Pew Research Center's tracked-browsing study of 900 US adults found the most frequently cited sources in Google's AI summaries were Wikipedia, YouTube, and Reddit, which together accounted for 15% of all sources listed in those summaries (Pew Research Center, March 2025 data).

That is the uncomfortable structure of the thing. A model asked to recommend a lawyer is not reading your practice-area copy and grading it. It is running what Google calls a query fan-out — issuing multiple related searches across subtopics and data sources, then assembling a paragraph from what comes back (Google Search Central, 2025). In legal, what comes back is Avvo, Justia, Martindale-Hubbell, your state bar's member directory, a local news story, and a Reddit thread from someone who hired you three years ago.

This is not a hypothetical channel. Clio's 2025 Legal Trends Report found that, of the consumers who used AI for a legal question, 28% were directed by it to contact a lawyer. Read the denominator carefully — that is 28% of AI users, not 28% of everyone. The machine is already making referrals. The only question is whether it knows your name.

The same report found the reverse also happens: in 12% of cases, consumers were convinced by an AI that their legal problems were not worth pursuing. Some of those were bad cases you would have declined anyway. Some were not.

Why do directories and Reddit outrank your site inside AI answers?

Because ranking and citation have come apart. Ahrefs analyzed 863,000 SERPs and 4 million AI Overview URLs and found that just 38% of pages cited in AI Overviews also rank in Google's top 10 for the same query — down from about 76% a year earlier (Ahrefs, March 2026). Being number one is no longer a ticket into the answer.

There is also a shape problem. The query is 'best personal injury attorney in Sacramento.' That question wants a list. A directory page is already a list. A Reddit thread is a list with human deliberation attached. Your homepage is a sales page for exactly one firm — structurally the wrong answer to a question that expects several. The model does not dislike you. You just are not the right shape.

Now the anti-hype counterweight, because the agencies pitching you 'AI search packages' will not offer it. AI Overviews barely show up on local searches at all: only 7.9% of local searches trigger an AI Overview, versus 22.8% of non-local queries (Ahrefs, 146M SERPs, September 2025). Someone typing 'divorce lawyer near me' overwhelmingly still gets a map pack. Do not torch your local budget chasing AI.

Here is how the surfaces actually divide today:

The honest verdict: for a local law firm in 2026, the map pack is still the highest-return surface, and AI visibility is the cheapest one to be early on — because almost nobody in your market is deliberately working it yet. Fix the map pack first. Build AI citations while your competitors argue about whether AI search is real. That is the sequencing we recommend, and it is what our GEO service is built around.

How does a firm become an entity the models recognize?

By being mentioned, consistently, on the sources the models already read. Whitespark's 2026 Local Search Ranking Factors survey — 47 local-search experts scoring 187 factors — ranked 'presence on expert-curated best-of and similar lists' as the single highest-scoring factor for AI search visibility, with three of the top five being citation factors. Whitespark's own conclusion: 'In AI SEO, mentions (citations) are the new link.'

Read that again, because it inverts twenty years of legal marketing advice. You do not become an entity a model recognizes by publishing more. You become one by being named elsewhere, in enough places, the same way, that the model's picture of your firm resolves into a single confident thing rather than a smear of half-matches.

The practical list, in the order we would work it for a firm starting from zero:

• Fix the name collision first. One firm name, one spelling, one address, one phone — identical on your site, your GBP, your bar listing, and every directory. 'Smith & Associates' in one place and 'Smith and Associates, PLLC' in another gives the model two firms and confidence in neither.
• Claim the legal directories. Avvo, Justia, Martindale-Hubbell, FindLaw, your state and county bar member directories. These are the pages models fan out to on legal queries. A complete profile with practice areas, bio, and reviews is table stakes — and we cover whether the paid tiers are worth it in are legal directories worth it.
• Name your attorneys. Individual lawyer pages, consistent bios, bar admission dates, real headshots. Models resolve people, not just businesses, and a partner with a coherent cross-web identity pulls the firm along with them.
• Get on lists you did not write. Local 'best lawyers in [city]' roundups, chamber directories, industry association pages, bar section pages, local news quotes. This is the highest-scoring AI visibility factor in the Whitespark survey and the one most firms ignore because it cannot be bought in a dashboard.
• Show up on Reddit honestly. It is a top-cited source in AI summaries. Answer questions in your practice area as a lawyer, disclose who you are, and do not astroturf — our Reddit AEO guide covers the line between participating and getting banned.
• Check your bar's advertising rules before any of this. Several states restrict comparative claims, 'best' language, and testimonials. Being cited by a model is not worth a grievance.

Notice what is not on that list: buying links, spinning out fifty thin practice-area pages, or paying anyone who guarantees you a ranking. Nobody can guarantee a ranking — not in Google, and certainly not inside a model's answer that regenerates on every prompt.

What kind of page passage gets lifted into an AI answer?

Short, self-contained, and attributable ones. Pew found the typical AI summary in its study was 67 words long, and 88% of those summaries cited three or more sources (Pew Research Center, 2025). A model building a 67-word answer from three-plus sources is not reading your 2,000-word essay. It is looking for one clean sentence it can lift without breaking.

Google's own guidance says there are no additional technical requirements to appear in AI Overviews or AI Mode — a page must simply be indexed and eligible to show with a snippet. Its advice is ordinary: crawlable pages, internal links, 'important content available in textual form,' and structured data that matches the visible text (Google Search Central, 2025). No secret schema unlocks this.

What separates a liftable passage from an unliftable one is almost entirely writing discipline:

The reward for getting cited is real and measurable. Seer Interactive analyzed 5.47 million queries across 53 brands and found that being cited inside an AI Overview delivers 120% more organic clicks per impression than appearing on the same SERP uncited — though still 38% fewer clicks than a SERP with no AI Overview at all (Seer Interactive, 2026). Citation does not replace the click. It rescues it.

The tactical version of this — heading structure, answer-first paragraphs, what to do with FAQ blocks — is in how to get cited by Google AI Overviews and how to get recommended by ChatGPT.

How do you check your own firm's AI visibility for free?

In about 20 minutes a month, with a spreadsheet and ten prompts. No tool, no subscription, no $500/mo AI-visibility platform. A partner can run this personally, and honestly should — the results are usually more persuasive than any agency deck.

Open ChatGPT, Perplexity, and Google AI Mode. Use a logged-out window or a temporary chat so your own history does not flatter you. Then run the same prompt set in each, and log what comes back.

• 'Who are the best personal injury lawyers in [your city]?' — swap in your actual practice area.
• 'I was rear-ended in [your city] and the insurer is lowballing me. Who should I call?'
• 'What should I look for when hiring a [practice area] attorney in [your state]?'
• 'Is [Your Firm Name] a good law firm?' — the brand-recognition test.
• '[Your Firm Name] reviews' — what the model thinks your reputation is.
• 'Top-rated [practice area] firms near [your neighborhood or county].'
• 'How much does a [practice area] lawyer cost in [your city]?' — the money question you want to be the answer to.

For each prompt, record four columns: were you mentioned (yes/no), were you linked (yes/no), who was mentioned instead, and — the one everybody skips — which sources did the model cite. Then repeat it on the first business day of every month and watch the trend, not the snapshot.

That fourth column is the entire strategy. The sources the model cited are, by definition, the sources it trusts for your practice area in your city. That list is your target list. If it cites a Reddit thread, a county bar page, and an Avvo category page, then those three pages — not your homepage — are where your next month of effort goes. You are not guessing at what the model reads. It just told you.

If you want the fuller version of this measurement loop, including how to track it without buying software, see how to measure AI search visibility on a budget.

Run the ten prompts. If your firm is absent from all three engines while a directory and a Reddit thread do the recommending, that is a fixable visibility problem, not a verdict on your lawyering. Our GEO program works exactly that gap — entity consistency, third-party mentions, and liftable passages — alongside the local search work that still wins most legal clients. Get my free audit and we will show you what the models currently say about your firm before you pay us anything.