A prospective client with a herniated disc and a police report used to open Google, type “personal injury lawyer near me”, and start clicking. Many still do, but a growing share now open a chat window instead, describe what happened in a paragraph of plain language, and ask what they should do next. Most firms have no visibility into that second conversation. It leaves no trace in Search Console, and if your firm gets named, or skipped, you are unlikely to find out why.
The scale is no longer speculative. The Pew Research Center’s June 2026 report on Americans and AI found that 44% of U.S. adults say they have ever used ChatGPT, up from 34% the year before, though Pew notes the question wording changed between those two years. The same survey found 60% say they have read the AI summaries that now sit at the top of search results.
For legal specifically, a 2026 consumer survey by iLawyer Marketing asked 1,110 people aged 18 to 65 how they would research an attorney. Nearly 42% said they would use ChatGPT, and 50% said they would use at least one AI answer engine. Two caveats worth stating plainly: that survey measures stated intent rather than observed behavior, and iLawyer Marketing is a legal marketing agency, so read it as directional rather than definitive.
The Part That Changed This Year
For a while, the reassuring answer to all of this was that AI systems mostly cite whatever already ranks well, so a firm with solid SEO would be carried along. That is getting less true, quickly.
Ahrefs re-ran its analysis of Google AI Overview citations in March 2026, across 863,000 keyword result pages and 4 million cited URLs, and found that 38% of cited pages also appeared in the top 10 organic results for the same query. Eight months earlier, that figure was 76%. A separate BrightEdge analysis in February 2026 put the overlap closer to 17%. The two studies disagree on the number and agree on the direction: the link between ranking on page one and being named in an AI answer is weakening.
That is the practical problem. Page one used to be the finish line, but now it is one input among several, and the other inputs are ones most firms have never audited. Anyone looking at how law firms show up in AI search results is looking at a different question than a rankings report answers, which is why a firm can hold three top five positions and still go unnamed when a prospect asks a chatbot the same question.
Check Where You Stand
Open ChatGPT, Google’s AI Mode, Perplexity, and Gemini. Log out of your accounts or use a temporary chat, because incognito mode does not disable account level memory and your own history will skew what you see. Run the same prompts through each, substituting your city and practice area:
- “I was injured in a car accident in Greenville, South Carolina. Which law firms should I call?”
- “Who are the best family law attorneys in Asheville, NC?”
- “What should I look for when hiring a personal injury lawyer in South Carolina?”
- “Is [your firm name] a good firm for this?”
- “Compare [your firm name] and [a competitor].”
Record more than whether you appear. Record who does appear, and which sources the model links to underneath the answer. Those citations are the useful part.
A word of caution before you act on prompt five: running comparison queries privately is research. Repeating what a chatbot says about you versus a competitor in your own advertising is a different thing, and ABA Model Rule 7.1 and its state equivalents govern comparative and unverifiable claims in attorney advertising. Check your state bar’s rules before any of this reaches a brochure.
Reading What Comes Back
No one local is named: The model answered generically, describing how to choose a lawyer rather than naming one. This is common in regulated practice areas. Your opening here is prompt three, the advice question, because that is the one the model will answer with substance and citations.
Competitors are named and you are not: The actionable case. Look at what the model cited. It is often not the competitor’s own website. It is a directory profile, a bar association listing, a review platform, or a local news mention. That tells you which sources the model trusts for your market, and those are the surfaces to go fix.
You are named, but the details are wrong: Wrong practice areas, an old office address, an attorney who left three years ago. Fix this first. It usually traces back to stale directory listings and an inconsistent Google Business Profile rather than anything on your website. It is also not purely a marketing problem, since inaccurate public statements about your services touch Rule 7.1.
You are named accurately: Do not ask the model why it recommended you. Language models generate plausible explanations for their own output after the fact, so that answer tells you nothing reliable. Instead, look at which sources it actually cited and treat those as the list of places worth defending.
What Actually Moves This
Consistency about who you are: Before a model can name your firm confidently, your basic facts need to agree across your site, your Google Business Profile, the state bar directory, and the major legal directories. Name, address, phone, practice areas, attorneys, jurisdictions. When those conflict, retrieval surfaces whichever version is most common, which may not be the current one. Reconciling listings is dull work and it is usually the highest return hours available.
Answers a machine can lift: Structure key service pages so each one opens with a direct answer to the question a prospect would actually ask, then supports it. Use headings phrased as questions. Keep the answer near the top and let it stand on its own, without requiring three preceding paragraphs for context. You are still writing for readers, but you are also writing for a system that pulls a passage out of context, and a passage that survives being pulled out is more likely to get used. Worth noting: Google retired FAQ rich results for most sites back in 2023, and there is no published evidence that FAQ schema increases AI citation, so treat markup as hygiene rather than as a lever.
Access, checked properly: Confirm you are not blocking the crawlers that matter: GPTBot and OAI-SearchBot for OpenAI, PerplexityBot, ClaudeBot, and Google-Extended. Note that these do different jobs, so blocking GPTBot, which is for training, does not remove you from ChatGPT’s search results, which use OAI-SearchBot. And robots.txt is only half the check. Cloudflare and similar services block AI crawlers at the network layer by default, which returns a 403 that a clean robots.txt will never reveal, so fetch your own homepage using each crawler’s user agent or check your CDN’s bot rules directly. While you are there, confirm your site is indexed in Bing Webmaster Tools, since ChatGPT’s search leans on Bing’s index.
Corroboration from elsewhere: Bar association profiles, legitimate directories, local press, published commentary. The concrete version of this: pick the three sources that showed up as citations in your own prompt testing and make sure your firm has a complete, current, accurate presence on each. That beats a generic campaign to “get mentioned more.”
The Attribution Gap
The hard part is not that AI traffic hides its origin. Clicks from ChatGPT and Perplexity do pass a referrer, and ChatGPT tags outbound links, so the visits you get are traceable. The gap is the visit that never happens. A prospect reads an answer naming your firm, never clicks through, and turns up four days later as a branded search or a phone call that looks like it came from nowhere. Your intake software logs it as “found us online” and your marketing report quietly credits it to something else.
The fix is procedural, not technical. Add one question to your intake script: “Before you called, did you use ChatGPT or any AI assistant to look into this?” Log the answer. In a quarter you will have real numbers for your firm rather than an argument about the industry.
One last thing worth saying to anyone reading this from the other side: an AI naming a law firm is not a vetting. These systems get details wrong and cannot assess whether a firm suits a particular matter. Being named is a marketing outcome, not a credential, and prospective clients should still check the bar directory.
Start with the prompts and the intake question. That is one afternoon of work, and it will tell you more about your firm’s actual position than most reports you could buy.
Vyron Johnson is the founder of VerdictIQ, where he runs AI visibility and search audits for law firms.
Our dedicated team gathers information from all the reliable sources to make the law accessible and understandable for everyone. We provide the latest legal news stories from across the country, delivered straight to you.