Which Legal Directories Actually Get Cited by AI? We Measured 1,770 Answers Across 20 Cities

Everyone has a theory about what gets a law firm recommended by AI. We got tired of the theories and measured it directly.

We ran 10 buyer-intent prompts per city — the kind of thing real prospects type when they need a lawyer, like "best personal injury lawyer in Austin" or "who do I call after a car accident in Phoenix." Each prompt ran three times against ChatGPT, Gemini and Perplexity, across 20 US metros. Coverage wasn’t perfect and a small number of calls failed and are excluded, so the total lands at 1,770 answers instead of a clean 1,800. For every answer, we logged which firms got named and which sources the engine cited. Here’s how often each directory turned up in a city’s top five cited sources:

Two things stood out.

Avvo comes up constantly in legal marketing conversations, but it made the top five in only 4 of 20 cities and never ranked above #4. Super Lawyers and Justia were in the top five in all 20. If you only have time to keep a couple of directory profiles current, that’s where the data says to put the hours.

Best Lawyers never appeared once. Its firm-level product, Best Law Firms, turned up twice (Nashville #4, Orlando #5). Firms tend to treat Super Lawyers and Best Lawyers as the same kind of listing. The engines clearly don’t.

Across all 20 cities, citations broke down as 68% directories, 23% law firm websites, 8% Reddit, and about 1% news and trade press. Reddit was a top-five source in 8 of 20 cities, and the single most-cited source in Phoenix — that one surprised us.

One caveat worth stating plainly: this measures two different things, and they don’t move together — which firms get named in the answer, and which sources get cited as the answer’s evidence. A firm can be named without any of its own pages ever being cited. Google Business Profile and Yelp both scored zero in our source rankings here, and that doesn’t mean they don’t matter. Engines can lean on that data without citing a google.com or yelp.com link, so this method would miss it entirely. Treat the zeros as "not cited," not "not useful."

Legal also sits in a category some call YMYL — your money or your life, alongside finance and health — where a wrong recommendation carries real consequences. That’s likely part of why citations skew so heavily toward established directories rather than firm websites: before naming a name, these engines seem to be reaching for third-party corroboration rather than taking a firm’s word for itself.

This data covers personal injury only. Tax, family and criminal law will have different citation patterns; we haven’t run those yet.

It’s also one snapshot. Re-running the same prompts moves the answers around a lot, which is a finding on its own — there’s no fixed ranking sitting there waiting to be claimed. The same prompt can hand you different firms on different runs, so the goal isn’t winning a single query. It’s showing up across more of the runs.

This is the same audit we run for individual firms, scoped to one firm instead of a whole city.

See the city-by-city AI visibility reports →