Key Takeaways
- AI assistants recommend firms based on corroborated mentions across independent sources, not on who ranks #1.
- Consistent name, address, and phone data across the web is a prerequisite for being named at all.
- Review volume and recency influence which firms models surface for 'best' and 'top' prompts.
- Test your visibility with a fixed prompt set monthly — it's the only reliable measurement available.
- Structured data and clear, factual page copy make your firm easier for a model to quote correctly.
A meaningful share of people who would once have typed "car accident lawyer near me" into Google now ask an assistant instead — and the assistant answers with two or three firm names, not ten blue links. If your firm isn't among them, you were never in the consideration set.
This isn't a ranking problem in the traditional sense. Models don't have a results page; they have a synthesis. Understanding how that synthesis gets built is the difference between guessing and working.
For the broader strategy this sits inside, see AI search and legal marketing and our AI search optimization approach.
How an AI Assistant Actually Picks a Firm to Name
Three mechanisms are in play, often at once.
Training data. The model absorbed the web up to a cutoff date. Firms mentioned frequently across independent sources — news coverage, bar association pages, legal directories, local publications — are simply more available to it. A firm whose only web presence is its own website is nearly invisible here.
Live retrieval. Most assistants now search the web mid-answer. This is where current rankings, Google Business Profile data, and directory listings re-enter the picture, and it's why traditional SEO still underpins AI visibility rather than being replaced by it.
Corroboration. Models weight claims that appear in multiple independent places. Your website saying you're the region's leading trial firm counts for nothing. A verdict reported by a local paper, a bar association leadership page, and a directory profile that all point the same direction counts for a great deal.
The practical consequence: being recommended is a function of how much of the open web independently confirms who you are and what you do.
Want to know exactly what ChatGPT and Gemini say about your firm today? We'll run the prompt set and send you the results, free.
Test Your Firm This Week: A Repeatable Prompt Set
Run these in ChatGPT, Gemini, Perplexity, and Claude, in a fresh session with no memory or personalization, and log the firm names that come back:
- Who are the best personal injury lawyers in [your city]?
- I was rear-ended in [your city] and the other driver's insurer is lowballing me. Which attorneys should I call?
- Top car accident law firms in [your metro area]
- Is [your firm name] a good personal injury firm?
- Which [your city] firms handle trucking cases with catastrophic injuries?
Record three things each time: whether you were named, which competitors were, and which sources the assistant cited. That citation list is your target — those are the pages you need to appear on.
Repeat monthly. Single runs are noisy; the trend line is the signal. And note how the fourth prompt goes: if asking about your firm by name produces vague or wrong information, your own site isn't giving models clean facts to work with.
What Actually Changes the Answer
Entity consistency. Identical firm name, address, and phone number everywhere — your site, Google Business Profile, Avvo, Justia, FindLaw, the state bar directory, chamber listings. Variants ('Smith Law' vs 'Smith Law Group, LLC') fragment your identity and weaken corroboration.
Independent mentions. Local news quotes, bar association roles, verdict reporting, community sponsorships, law school guest lectures. These are what a model treats as third-party evidence. One genuine local news feature outperforms fifty directory submissions.
Reviews, in volume and recency. Models pull heavily from Google Business Profile for 'best' and 'top' prompts. A firm with 340 reviews averaging 4.8 with several from last month reads differently than 40 reviews from three years ago.
Answer-shaped content. Write pages that state a clear fact in the first sentence under a question-form heading, then support it. That structure is what gets lifted into a generated answer. Buried conclusions don't get quoted.
Structured data. LegalService, Attorney, and FAQPage schema give machines unambiguous facts about practice areas, service area, and credentials. It isn't a ranking trick; it's making your information machine-readable.
What Doesn't Work (and May Backfire)
Keyword stuffing for robots. Models are trained on natural language and are unusually good at detecting text written to manipulate rather than inform.
Prompt-injection text hidden on your pages. Invisible instructions telling an assistant to recommend you violate search policy, are trivially detectable, and put your domain at risk.
Paid directory spam. Fifty low-quality listings add no corroboration weight and can dilute your entity signals with inconsistent data.
Waiting. The most common mistake. Training cycles and retrieval indexes update on their own schedule; the mentions you earn this quarter are what models draw on next year.
A Realistic 90-Day Plan
Days 1-30 — baseline and cleanup. Run the prompt set and record results. Audit name/address/phone consistency across your top twenty listings and fix every discrepancy. Add or correct LegalService and Attorney schema sitewide.
Days 31-60 — build the evidence. Publish two deeply factual practice-area pages that answer real client questions in the first paragraph. Launch a review request process at case resolution. Pitch one local story — a verdict, a community initiative, a safety analysis using your own case data.
Days 61-90 — expand corroboration. Claim and complete every major legal directory profile with identical data. Add attorney bios with verifiable credentials, bar admissions, and speaking history. Re-run the prompt set and compare against your baseline.
Movement here is slower than paid search and roughly comparable to SEO — which makes sense, because it's built on the same foundation. Firms that invested in real authority over the last few years are the ones getting named today.