Skip to content
Original research · August 2026

The State of AI Search in Legal

We ran our 20-point AI visibility scan on 84 North Carolina law firm websites, all of them already ranking on the first two pages of Google for high-intent searches. The question was simple: when someone asks ChatGPT, Google AI or Perplexity to recommend a lawyer, can the machines read these firms well enough to name them? For most of them, no.

Scan your own firm free →

The headline numbers
Average score
55.0/100

The median was 52.5. These are not neglected websites. Every firm in this sample is already winning at traditional SEO, which is exactly what makes the number interesting.

Scored under 60
53 of 84

Roughly two firms in three are close to unreadable for an answer engine. Only 4 scored 80 or above.

The leader gap
78.2 vs 36.7

The top ten firms averaged 78.2; the bottom ten averaged 36.7. The best firm in the sample scored 92, so this is plainly achievable. Most simply have not done it.

The distribution

Ranking on Google is not the same as being readable by AI.

Every firm here earned its Google position the hard way. That work does not carry over automatically. Search engines rank pages, and these pages rank. Answer engines recommend businesses, and to do that they need structured, machine-readable facts about who a firm is and what it does. That layer is mostly missing.

AI visibility scores of 84 North Carolina law firm websites (out of 100)9Under 404440–592760–79480+

Where they fail

The gaps, ranked.

Every item below is fixable, most of them in weeks rather than quarters. That is what makes this an opportunity rather than an obituary, for whichever firm in each market moves first.

AI crawler access policy100%
robots.txt says nothing about GPTBot, PerplexityBot or ClaudeBot, so the engines fall back to guessing.
Review/rating markup82%
Reviews exist on Google and the directories. Nothing on the firm’s own site lets an engine connect them.
Service schema81%
Practice areas are not published as structured data, so an engine cannot tell what kind of law the firm actually practises.
Phone/NAP consistency81%
Several different phone numbers across a site reads to a machine like several different firms, or a stale one.
LocalBusiness or Attorney schema80%
No machine-readable record of the firm itself: its offices, its hours, or who its attorneys are.
FAQPage schema79%
The question-and-answer format assistants quote back to users most often, and four firms in five do not have it.
llms.txt guide for AI crawlers68%
The emerging file that tells AI crawlers what a business is and which pages matter.
Answer-shaped content48%
Pages are written as brochures rather than as answers to the questions clients actually ask.
What it means

Answer engines name two or three firms. Not ten.

A search results page has room for everybody. An AI answer does not. When someone asks an assistant who to call after a charge or an injury, it returns a short list, usually two or three names, assembled from whatever it can reliably read. In this sample, most firms have given it very little to read. The firms that fix that first are not merely improving their marketing, they are becoming the answer, because there is not much competition for the machine’s attention yet.

If you run a firm

The bar is a 55.0.

You do not have to be perfect to win your market’s AI answers. You have to be readable, consistent and answer-shaped before the firm down the road is.

Get your free scan →

Full disclosure

We scored a 58 ourselves.

Before rebuilding our own site we ran this same scan on us and scored 58 out of 100. We took our own medicine first, and we will show you the before and after.

See our ledger →

Method

How we measured.

94 firm websites were drawn from Google organic results for high-intent North Carolina legal searches across criminal defence and DWI, family law and workers’ compensation. Directories, bar associations and public defender offices were excluded, so every domain is a working firm. Scanned 84 successfully in August 2026 with our automated 20-point engine; 10 could not be fetched and were excluded rather than scored. No manual scoring and no cherry-picking.

Want to know your firm’s number?

The free AI Visibility Scan is this study run on your firm: your score out of 100, the specific gaps, and a fifteen-minute walkthrough. The report is yours either way. Press and researchers: we are happy to share methodology and aggregate data, contact support@dbadvertising.co or (844) 809-9223.

Get your free AI Visibility Scan →