Blog

Why is my law firm not showing up in ChatGPT?

John Rice9 min read

You've probably already run the test. You opened ChatGPT, typed "best [your practice area] lawyer in [your city]," and got back a confident shortlist with someone else's name on it, or no local firms at all. (If the same competitor keeps appearing where you don't, the sibling diagnostic, ChatGPT recommends your competitor, audits their footprint specifically.)

That silence has causes. Six of them cover almost every case, and you can check each one yourself today. No tools, no "free audit" sales call. In rough order of how often they're the culprit:

1. AI answers lean on directories — and you're thin on them

When someone asks an AI assistant to recommend a lawyer, the assistant rarely reasons from your website. It leans on legal directories.

The number is stark. A legal directory was the first source cited in 77.8% of AI answers to lawyer-hiring questions (1,254 of 1,612 valid answers, 540 queries asked three times) — InterCore Research, July 2026. Directories also took at least 51.8% of all 18,900 citations in that dataset. InterCore sells AI-visibility services; this is vendor-published research, cited here with that caveat. A separate April 2026 report from 5W Public Relations and Haute Lawyer found seven directories — Chambers, Legal 500, Super Lawyers, Best Lawyers, Martindale, Avvo, and Justia — dominating AI citations for legal queries, while individual firms "appear inside these directories, not as independent voices."

Not all directories pull equal weight. InterCore's measured citation shares, read as tiers:

TierDirectoryShare of all AI citationsTypical cost
1Justia9.7%Free profile available
1Super Lawyers8.1%Selection-based; profile upgrades paid
2Avvo5.7%Free profile available
2Lawyers.com5.7%Paid listings
2Expertise.com4.9%Selection-based
3Best Lawyers / Best Law Firms4.3% / 3.9%Peer-selection; paid enhancements
3FindLaw2.2%Paid listings

Citation shares are InterCore's data. The "Typical cost" column is Briefly's annotation, compiled from the directories' own pricing pages; it is not part of InterCore's dataset.

Note where the money and the citations diverge. Some of the priciest legacy listings sit at the bottom of the citation table, and some of the most-cited profiles are free to claim.

How to check it yourself: search your own name and firm name on Justia, Avvo, Super Lawyers, and Lawyers.com. Is the profile claimed? Does it list your current address, practice areas, and a real bio, or is it an auto-generated stub?

What fixes it: claim and fully complete the free tiers first (Justia, Avvo), then decide on paid ones with citation data rather than a sales call. A complete profile with practice areas, locations, and reviews gives the assistant something to cite.

2. There's almost nothing about you that you didn't write

AI assistants trust what other people say about you more than what you say about yourself. Muck Rack's May 2026 "What Is AI Reading?" study, built on more than 25 million links cited by ChatGPT, Claude, and Gemini, found that earned, third-party media accounts for 84% of AI citations — a figure that has held between 82% and 89% across every edition of the study. Your own website contributes a sliver.

If the only pages that mention your firm are your website and your Facebook page, the assistant has nothing independent to stand on. It will name the firm that shows up in the local business journal, the bar newsletter, and three "best car accident lawyers in Columbus" listicles.

How to check it yourself: search Google for "Your Firm Name" -site:yourfirm.example. Count the real third-party pages: news mentions, directory profiles, roundups, community coverage. Under a dozen is thin. Do the same for the firm that is showing up.

What fixes it: earned mentions, built patiently. Local press comment on cases in the news, bar association publications, community sponsorships that get covered, credible local roundups. Not guest-post farms; one real local news mention outweighs twenty spam placements.

3. The internet can't agree on who you are

AI models assemble a picture of your firm from every mention across the web. If your website says "Marsh & Cole Trial Lawyers, PLLC," your Google Business Profile says "Marsh and Cole," Avvo lists your old office on Fifth Street, and the state bar shows a partner who left in 2022, that picture is blurry. Blurry entities don't get confidently recommended; a model unsure whether two half-profiles are the same firm may credit neither.

How to check it yourself: open five tabs: your website's contact page, your Google Business Profile, your Justia and Avvo profiles, and your state bar listing. Compare name, address, and phone number character by character. Note every variant: "&" vs "and," PLLC vs LLP, suite numbers, old numbers, old partners.

What fixes it: pick one canonical rendering of the firm name and address, then sweep every listing to match it. Tedious, one afternoon, permanent payoff.

4. Your own website is telling AI crawlers to go away

This is the five-minute check that occasionally explains everything. Some firms, or their web vendors, or an overzealous security plugin, block the crawlers AI assistants use to read the web.

How to check it yourself: go to yourfirm.example/robots.txt in a browser. Look for lines like these:

User-agent: GPTBot
Disallow: /

User-agent: PerplexityBot
Disallow: /

A Disallow: / under any of these user agents tells that crawler to skip your entire site. The names to scan for: GPTBot, OAI-SearchBot, and ChatGPT-User (OpenAI's three crawlers, per its own documentation — the second feeds ChatGPT search specifically), PerplexityBot (Perplexity's index crawler), Google-Extended, and ClaudeBot. A firewall or CDN "bot protection" setting can also block them silently even with a clean robots.txt. Ask your web vendor directly: "are we serving pages to GPTBot and PerplexityBot?"

What fixes it: remove the disallow lines for AI crawlers, or scope them narrowly (block a client-portal path, not the whole site). If you want to appear in AI answers, blocking the readers is self-defeating.

5. You describe your practice in words clients never use

Your site says "catastrophic injury litigation" and "dissolution of marriage." Your future client types "lawyer for a bad car accident" and "divorce lawyer near me who does payment plans." AI assistants match questions to sources; if no page of yours speaks the client's language about the client's city, you're a weaker match. City phrasing matters too: a firm marketed "Twin Cities-wide" can miss questions that name Bloomington.

How to check it yourself: read your practice-area pages and ask one question: does this page contain the actual words a stressed non-lawyer would type at 11pm? Then check whether each city you serve has a page that names it, or just a footer mention.

What fixes it: plain-language pages that answer real hiring questions ("How much does a DUI lawyer cost in Austin?" "Do I need a lawyer for a custody case?") for the specific places you practice. Genuine answers only; thin duplicated city pages now attract search-spam penalties.

6. You checked once — and AI answers are unstable

Here's the uncomfortable one: your absence might be partly noise. The same question does not reliably produce the same answer.

This has been documented since the beginning. In March 2023, AttorneySync asked ChatGPT for personal injury firm recommendations and reported that when they ran "these same prompts over and over again, I got different lists of law firms." It runs deeper than repetition: Carnegie Mellon-led researchers found that ordinary paraphrases of a prompt shifted the likelihood of an LLM mentioning a given brand from never to always (Lin et al., LLM Whisperer, CHI 2025). The same study's adversarial experiments, in which synonyms were deliberately optimized to steer the model, moved mention likelihood by up to 78.3 percentage points — an attack result, not everyday variance, but it marks how far wording alone can push an answer.

So one screenshot proves very little, in either direction. You may be absent in the phrasing you tried and present in three phrasings you didn't.

How to check it yourself: ask your core question five times in fresh chats, then three reworded variants, across at least two assistants. Log every answer. Now you have a sample, not an anecdote.

What fixes it: nothing "fixes" variance. You manage it with measurement discipline: many questions, many phrasings, repeated on a schedule, so you read a trend line instead of a coin flip.

How you'd know for sure: this cause, more than any other, is why systematic checking exists. Frequency across repeated runs is the real number — the metric we define as Case Recommendation Share — and a single answer never is. The tracking guide walks through running that measurement yourself, spreadsheet and all.

FAQ

Does ChatGPT even know my firm exists?

Probably, in fragments. The real question isn't whether your name sits somewhere in the training data. It's whether the sources assistants cite at answer time (directories, reviews, third-party pages) present your firm strongly enough to make a shortlist. That's causes 1–3 above.

How long until fixes show up in AI answers?

Weeks to months, and it varies by assistant. Directory and crawler fixes can surface quickly in browsing-based answers (Perplexity, ChatGPT search, Google AI Overviews); shifts in a model's baked-in knowledge move much slower. Checking weekly beats checking once: you see the trend instead of guessing.

Should I pay for premium directory listings to fix this?

Not reflexively. The citation data shows free-to-claim profiles (Justia, Avvo) among the most-cited sources, while some expensive legacy listings are cited least. Complete the free tiers first; justify paid spend with evidence that a directory actually gets cited in your market.

Can't I just ask ChatGPT why it didn't recommend me?

You can, but don't trust the answer. Models generate plausible-sounding explanations for their own outputs without reliable access to the real reasons. Diagnose from the observable inputs (the six causes above), not from the model's self-report.

Is being invisible in ChatGPT actually costing me cases?

The demand is no longer speculative. In iLawyerMarketing's 2026 survey of 1,110 US consumers, 41.9% said they would use ChatGPT to research a lawyer — up from 9% in 2023 — and 9.5% would skip Google entirely. (The stats hub collects every number like this, with sources.) Whether it costs you cases depends on who's in the answers you're missing from.

John Rice builds and operates the scan engine behind Briefly, which runs client-style lawyer-hiring questions across ChatGPT, Gemini, Perplexity, Google AI Overviews, and Google AI Mode on a recurring schedule and stores every answer it collects. He is not a lawyer; he measures what AI assistants say, with receipts. The observations here come from operating that system; the statistics come from the third-party studies linked above. Methodology · About John

Briefly measures what AI assistants say. It does not rank, rate, or endorse attorneys.

Sources