Somebody asks ChatGPT which firm handles wrongful termination cases in Tampa. They get three names, a summary of what each does, and a short explanation of how contingency fees work. They never open a search results page. Whether your firm was one of those three names had nothing to do with your Google Ads budget, and increasingly it has less to do with your blue-link rankings than firms assume.
Key Takeaways
- AI assistants and Google’s AI overviews answer legal questions directly, so a growing share of research ends before anyone visits a firm website.
- Getting cited depends on clear, quotable answers, strong structured data, and third-party mentions the model already trusts.
- Attorney credentials, bar admissions, and named authorship carry unusual weight for legal topics because the stakes are high.
- Directory profiles, bar listings, and local news mentions feed the sources these systems draw from, so off-site presence matters as much as your own pages.
- Track branded search volume and direct traffic alongside rankings, since AI-driven discovery often shows up there first.
What changed, in practical terms
Legal research by consumers used to follow a predictable path. Type a question, scan the first page, open three or four results, form an opinion, then look for a firm. Every step of that path gave you a chance to be seen.
Now a large share of those questions get answered in place. Google shows an AI overview at the top. ChatGPT, Perplexity, and Claude answer conversationally and cite a handful of sources. The person still hires a lawyer eventually, but the education phase happens somewhere you can’t measure with a rank tracker, and the shortlist forms before they ever land on your site.
Firms notice this as a strange pattern in their analytics: organic traffic flat or down, while consultations hold steady or the callers seem better informed than they used to be. That’s the new funnel showing itself. People are arriving already knowing what a contingency fee is and roughly what their case might be worth.
What gets a firm cited
These systems pull from a mix of their training data, live search results, and whatever sources they consider reliable for the topic at hand. For legal questions in particular, the reliability bar is higher because the systems are tuned to be careful with advice that could hurt someone.
Answer the question in the first two sentences
Pages built as slow reveals, where the actual answer sits under a heading two-thirds of the way down, don’t get quoted. Put a direct answer at the top, then expand. “In Florida, you generally have two years from the date of injury to file a personal injury claim, though several exceptions extend or shorten that window.” That sentence is extractable. A paragraph about how every case is unique is not.
Be specific about jurisdiction
Generic national content is exactly what these systems already have plenty of. What they lack, and what they’ll reach for, is content that says which state, which statute, which court. A page covering the actual rule in your jurisdiction with the citation attached is far more likely to be pulled than a broad overview.
Put a real attorney’s name on it
Author attribution matters more for legal content than almost any other category. Every substantive page should carry an attorney byline linked to a bio that lists bar admissions, years in practice, notable cases, and any publications or speaking. That bio should have Person schema on it. Unsigned content on a topic that affects people’s legal rights reads as low-trust to both search algorithms and the systems built on top of them.
Use structured data properly
LegalService or Attorney schema on practice pages, FAQPage schema on question content, Person schema on bios, and Organization schema tying it together with your address, phone, and licensing details. None of this guarantees a citation. All of it makes your content easier to parse correctly, and easy to parse is a real advantage.
The off-site half nobody wants to hear about
Here’s the uncomfortable part. A meaningful share of what these systems know about your firm comes from places you don’t control.
Avvo, Justia, FindLaw, Martindale, your state bar directory, Chambers if you’re in that world, local news coverage of cases you handled, podcast appearances, bar association committee listings. When a model builds a picture of “employment lawyers in Tampa,” it’s synthesizing across all of that. A firm with a thin presence outside its own website is a firm the model has little reason to mention.
Where to put effort
- Complete and update every major legal directory profile, with matching firm details across all of them
- Get quoted in local and trade press, which is easier than firms think if you respond quickly to reporter queries
- Publish on bar association sites and industry publications, not just your own blog
- Keep your Google Business Profile active, since local data feeds a lot of these answers
- Appear on podcasts in your practice area, which produces transcripts that get indexed
Consistency across all of it matters. If three directories list an old office address and two list a former partner as managing attorney, you’re feeding contradictory information into systems that resolve conflicts by trusting the source that seems most reliable, which won’t necessarily be you.
Content that AI search rewards
Some formats hold up better than others in this environment.
- Definitive answers to narrow questions. “Can my employer fire me for filing a workers comp claim in Florida” beats “understanding employment law.”
- Process walkthroughs with timelines. What happens at each stage, how long it takes, what the client does. This is the kind of structure that gets summarized well.
- Comparison content. Chapter 7 versus Chapter 13. Mediation versus litigation. These map directly onto how people phrase questions to a chatbot.
- Cost and fee explanations. Money questions come up constantly and most firm websites dodge them, leaving the field to directories and forums.
What holds up badly is the thin service page written purely for a keyword. Those pages were already borderline. They’re now close to invisible.
Measuring something you can’t see directly
There’s no dashboard for AI citations yet, at least not a reliable one. What you can do is triangulate.
Ask the major assistants your core questions once a month and record whether your firm appears. It’s manual and imperfect and still more useful than guessing. Watch branded search volume in Search Console, since people who hear about you from an AI answer often search your firm name next. Watch direct traffic. And add a question to your intake script: how did you first hear about us? A rising number of “I asked ChatGPT” answers tells you more than any tool currently sold for this.
None of this replaces conventional SEO. The pages that get cited are largely the pages that already rank well, so the foundation is the same. What’s changed is that being findable is no longer the same thing as being visited, and firms that only measure sessions will misread their own results for the next couple of years.