When a prospective client asks an AI model which personal injury attorney to call in their city, the answer is decided long before the conversation starts. Three or four firms get named. The other 400 in the market do not exist in that conversation. This is no longer a marketing question. It is a question of whether your firm is structured to be cited.
Where the case actually starts now
Ten years ago, an injured driver searched Google and clicked the first three results. Five years ago, they asked a friend on Facebook. Today, increasingly, they open ChatGPT and type something close to the way they actually talk: "I was rear-ended on I-95 last week, my neck is still bad, what kind of attorney handles this and who's good in Newark." The AI returns two to four names, a sentence on each, and a recommendation on who to call first.
That short list is the new top of funnel. The user does not scroll through ten ranked links. They do not compare Avvo profiles side by side. They get a confident shortlist with reasoning attached, and most of them call name one or name two. Whether you got a chance to earn that case was decided weeks or months earlier, when the model built its picture of who the credible personal injury attorneys in your market are.
For a firm operating on contingency, this is consequential math. A single shortlist placement on a serious injury matter is worth more than a year of pay-per-click on case types you do not actually want. The question is not whether AI is in the funnel. The question is whether your firm is in the answer.
The signals AI uses to pick attorneys
Five categories of signal drive whether a model names your firm. None of them are mysterious, but they are different from the SEO signals firms have been buying for fifteen years.
1. Entity clarity
AI models build a knowledge graph of named entities. Your firm needs to be a clean, unambiguous entity in that graph. That means a single canonical firm name across every property, a complete LegalService and Attorney schema on the site, full NAP (name, address, phone) consistency across Google Business Profile, Avvo, Martindale, Justia, FindLaw, and the state bar directory, and bio pages for every attorney with structured data covering bar admissions, years in practice, education, and practice areas.
Firms routinely fail at this in small, expensive ways: the firm is "Smith and Associates" in one place and "Smith Law Group" in another, two attorneys share a bio page, the address on the Avvo profile is two offices ago. Each inconsistency weakens the entity. A clean entity gets cited. A blurry one does not.
2. Citability
AI models cite specific sentences they can extract cleanly. The page formats they extract from best are question-and-answer FAQ pages, defined-term explainers, and structured case-result summaries with numbers.
A FAQ page that answers "what is the statute of limitations for a car accident claim in New Jersey" or "what does a contingency fee actually cover" gets cited. A homepage block that says "we fight for the compensation you deserve" never gets cited, because there is nothing in it to extract. The unit of GEO value at the page level is the citable sentence, not the keyword.
3. Topical authority
Models trust firms that cover their practice areas with depth, not breadth. A firm that wants to be the cited rear-end collision firm in its market needs a pillar page on auto accident cases, sub-pages on the major fact patterns (rear-end, intersection, multi-vehicle, commercial truck, uninsured motorist), explainers on the procedural arc from intake through demand letter through deposition through settlement or trial, and content on the specific medical and economic issues common to those cases.
Generic personal-injury copy does not produce topical authority. It produces noise that the model discards. The firms getting cited write the way an associate would brief a client: specific, concrete, and grounded in the actual mechanics of the case.
4. Off-domain authority
AI models do not learn about you from your website alone. They learn from everywhere you appear: Avvo, Martindale-Hubbell, Justia, FindLaw, Super Lawyers, the local bar association, news mentions, podcast appearances, guest articles on legal blogs, verdicts and settlements posted to industry databases, and credible directories specific to your practice area.
For legal, this off-domain layer matters more than it does for almost any other vertical. Models lean hard on legal-specific authority sources because the stakes of recommending a bad attorney are high. A firm with complete, accurate, claimed profiles across the legal directory layer gets triangulated as legitimate. A firm with a beautiful website and nothing else looks thin.
5. Technical surfacing
Crawler access. JSON-LD schema. Sitemap discipline. The largest GEO win we have ever delivered for a law firm was a single line removed from robots.txt that had been blocking AI crawlers for fourteen months. The site looked fine in Google. It was invisible to ChatGPT. Within ninety days of unblocking and adding LegalService and Attorney schema, the firm started appearing in ChatGPT shortlists for its core practice area.
What to actually build
The list of GEO work that moves the needle for a personal injury firm is short and concrete:
- Practice-area pillar pages. One per major case type. Auto accidents. Slip and fall. Medical malpractice. Workers comp. Each one structured with H2 sub-sections that answer the specific questions an injured person actually asks.
- Attorney bio pages with full schema. Name, headshot, bar admissions, years in practice, education, practice areas, notable results, languages spoken. Attorney schema, not generic Person schema.
- Case result library. A page per significant settlement or verdict, with the case type, the injury, the venue, the recovery, and any bar-compliant context on the result. Numbered, dated, and structured for extraction.
- FAQ pages on the questions clients actually ask. What does contingency cost. How long does a case take. What is the statute of limitations. What happens at a deposition. Each FAQ marked up with FAQPage schema.
- Complete directory presence. Avvo, Martindale, Justia, FindLaw, Super Lawyers, state bar, local bar. Claimed, populated, consistent.
That stack handles the work. Most firms do not have any of it built to the standard the models reward.
Bar compliance is not the blocker
Attorneys hear "rebuild your content for AI" and reasonably ask whether this is going to land them in front of disciplinary counsel. State bar advertising rules govern how a firm describes itself, what claims it can make about outcomes, what disclaimers are required, and how testimonials and case results have to be presented. None of those rules prevent the GEO work.
Clean entity data is not advertising. Schema markup is not advertising. A FAQ page that explains contingency or statute of limitations or what happens at a deposition is education, not solicitation. A case result library, presented with the required jurisdiction-specific disclaimers and presented factually rather than promotionally, is compliant in every state we have worked in. GEO and bar compliance run on parallel tracks. They do not collide.
The 60 to 90 day window
AI models build their citation graphs incrementally and reinforce them each retraining cycle. The first AI citations for a firm that does this work properly typically appear 60 to 90 days after implementation. Citation density compounds across the following 6 to 12 months as the directory profiles age in, the practice-area depth accumulates, and the model retrains on the new signals.
Firms that establish citation authority through 2026 will be cited by default for years. Firms that wait will be optimizing against an entrenched set the model already trusts. In a top-twenty metro for personal injury, three to four firms now claim the bulk of the AI shortlist real estate. The other firms in that market are working against a model that has already decided who it trusts.
The honest version. Most firm marketing budgets are pointed at the funnel that used to exist. Billboards, pay-per-click, TV. Those still work, but they are converting the small share of clients who have not yet asked an AI for a recommendation. Reallocating even a fraction of that spend to the structural work that gets a firm cited in the AI shortlist is the highest-leverage marketing move available to a personal injury practice in 2026.
Where to start this week
Start with a diagnostic, not a campaign. Run the AI Visibility Scan against your firm's domain. It evaluates entity clarity, schema coverage, citability, off-domain presence, and technical surfacing against the actual signals AI models use. It takes under a minute, surfaces the specific gaps, and tells you which ones move the needle in the next 60 days versus which are longer compounding plays.
From there, the work is sequenced: fix the technical and entity issues first, build the practice-area pillar and FAQ pages second, populate the case result library and directory profiles third. None of it requires reinventing the firm. All of it requires treating AI visibility as infrastructure, not as a campaign.