Industries · How law firms get recommended by ChatGPT and AI search
Clients ask AI which lawyer to call. Make sure it can read your firm.
"Who's the best DUI lawyer in Tulsa?" is now a question people put to ChatGPT and Perplexity. The answer usually names a few firms and cites the pages it read. Here's how to be one of them.
Free, no account. We only read public files.
What people ask AI assistants
- “Best personal injury lawyer in [city] with no upfront fees”
- “Divorce attorney in [city] who handles custody”
- “Do I need a lawyer for a minor car accident in [state]?”
- “Estate planning attorney near me, flat fee”
Why does AI search matter for law firms?
Legal clients do more research before calling than most local customers, and they ask long, specific questions. That suits AI assistants, which answer exactly those questions and then recommend a firm.
Assistants lean on practice-area pages, attorney bios, bar admissions, and third-party sources like Avvo, Justia and local news. A firm whose site says clearly which matters it takes, where, and who handles them is easier to recommend than one with a generic "we fight for you" homepage.
Bar rules on advertising still apply. Factual, specific pages are the safest kind of content anyway.
What we found on 33 law firm sites
Full report and method →0 / 33
0%
block any AI crawler in robots.txt
robots.txt is rarely the problem
23 / 33
70%
have no llms.txt at all
11 / 32
34%
homepages with no Organization or LocalBusiness schema
of homepages we could read
1 / 33
3%
returned a firewall 403 to our automated request
From GeoLumen's own scan on 2026-10-19 of public files only (robots.txt, llms.txt, homepage). 13 of 32 readable homepages used an industry-specific schema type, and 3 of the 10 sites serving something at /llms.txt served a broken, empty or HTML file. Small samples: treat these as a snapshot, not a census.
The law firm checklist
Things to do this week, roughly in order.
- Write one page per practice area and location you want cases for, naming the court or county where it matters.
- On every attorney page, list bar admissions, years in practice and the matter types they handle, as text.
- Say how consultations work: free or paid, by phone or in person, and how fast you call back.
- Explain fee structures in plain words (contingency, flat fee, hourly) for each practice area.
- Add LegalService schema for the firm and Attorney or Person markup on bio pages.
- Match your Avvo, Justia and state bar profiles to your site: same firm name, address, phone and practice areas.
What are the common mistakes, and the fixes?
- 01
Practice areas without their own pages
A single "Practice Areas" list gives an assistant nothing to cite for "car accident lawyer in [city]". One clear page per practice area and location does.
- 02
No business schema, or the wrong type
About a third of the firm homepages we could read had no Organization or LocalBusiness-type JSON-LD at all. LegalService with your address, phone and areaServed is the minimum.
Free tool: Schema / JSON-LD Generator → - 03
llms.txt left to a plugin
Several firm sites that serve an llms.txt got it from an SEO plugin. That's fine if it's valid, but check it lists your practice areas, not just your blog archive.
Free tool: llms.txt Generator & Validator → - 04
Attorney bios as images or PDFs
Bar admissions, years in practice and case types should be plain text on each attorney's page. Assistants can't quote a headshot with a caption baked in.
Which schema should law firms use?
LegalService describes the firm; Attorney describes an individual lawyer. Many firms use both: one for the firm, one per attorney page.
Generate it for your law firm →Recommended type
"@type": "LegalService" // (or Attorney)
Start with the free check. If you'd rather hand it over, there's the Snapshot.
The free AI Crawler Checker shows in seconds whether AI crawlers can read your site. The $99 Snapshot audit goes further for one law firm: 10 customer questions × 4 AI engines, crawler, robots.txt & CDN review, llms.txt and schema written for you, prioritised fix list (PDF). Delivered in 5 working days.
FAQ