How do you measure whether AI assistants mention your firm?

To measure whether AI assistants mention your firm, keep three separate records: a fixed set of test prompts you run on a schedule and save in full, the AI-related data that Google Search Console, Google Analytics, and your referral tracking actually report, and what new clients tell intake about how they found you. Keep mentions, citations, visits, consultations, and signed matters apart, and never put real client facts into a test prompt.

The question often starts with a screenshot. A partner asks ChatGPT for a divorce lawyer in their city, sees the firm named, and forwards the image. Another partner tries the next day and gets a different list. Both observations are real. Neither tells you how often the firm is named, for which questions, or whether any of it produces a signed matter.

This is a measurement question for firm owners and marketing leads, not legal advice. Our article on measuring AI search mentions, citations, visits, and leads covers the general method. This page focuses on what changes when the business is a law firm.

Build a prompt set from how clients describe their problems

Start with a manageable list of prompts drawn from real intake conversations, organized by practice area and location. Include situation prompts ("my employer cut my hours after I filed a complaint, who can help in Phoenix"), comparison prompts, and prompts that name the firm. Keep those groups separate in the report, since a prompt that names the firm is likely to return the firm and says little about unprompted mentions.

The phrasing should come from callers, not from the firm's practice area labels. Ask the intake team: "What did the caller call the problem before we gave it a legal name?" Those words make better prompts than the headings on your site.

One law-firm rule applies from the first prompt. Test prompts should be invented scenarios, never facts from a real matter. The ABA has addressed lawyers' use of these tools: its announcement of Formal Opinion 512 says the opinion "provides guidance on ethical issues involving the use of generative artificial intelligence tools," and that a lawyer "must be cognizant of the duty to keep confidential all information relating to the representation." Your state bar may have its own guidance too, so check with ethics counsel before building any process that touches client information.

Record every answer the same way

Each saved answer is one observation from one moment. Answers can change with the date, the account, the location, and the mode, so record those alongside the text.

Read each answer for what it actually does. A mention is not a recommendation. An answer can name a firm in a list of options, cite one of its articles as a source on a legal topic, or suggest it for the reader's situation, and those are three different findings. Report how many answers matched out of how many you checked, so the sample size is visible beside the result.

Schedule matters as much as format. Run the same prompts at the same point each month, by the same person if possible, and change the prompt set on purpose rather than casually. When you add prompts, note the date, because the comparison across that change is no longer like for like.

Read Google's reports within their scope

Google Search Console now includes a generative AI performance report. Google's help page says it "shows data about how your site performs in generative AI features on Google Search," covering AI Overviews and AI Mode, and that it reports impressions. The report may not appear if "your site hasn't received enough impressions in generative AI features on Google Search," and "the usual data limitations (1,000 row limitation, time period, etc)" of the Search performance report apply.

Google Analytics 4 has a separate classification. Its default channel definitions say "AI Assistant is the channel by which users arrive at your site from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok," and that it excludes Google's AI Overviews and AI Mode, whose non-ad traffic falls under Organic Search. OpenAI's publisher FAQ says "ChatGPT automatically includes the UTM parameter utm_source=chatgpt.com in referral URLs," and advises sites that want to appear in ChatGPT summaries and snippets to "make sure you aren't blocking OAI-SearchBot." Check what your own analytics receives before building a report on these labels. Search Console counts impressions of links to your site; it does not describe what an answer said about the firm. That gap is one reason the saved prompt record exists.

Connect visits to consultations and signed matters

Visits are not clients. Add a "how did you hear about us?" question to intake, record the answer in the caller's words, and keep it separate from the source analytics recorded. A person may read an AI answer, remember the firm's name, and call days later from a search.

Then define a qualified inquiry for each practice area: the right matter type, the right jurisdiction, no conflict. Count one person who fills a form, calls, and books as one inquiry. The person building the report should not have to guess which inquiries became real matters, because intake already knows, and its records should feed the report directly.

Declined matters deserve their own line. If callers who mention an AI assistant are often outside the firm's practice areas or jurisdiction, that says something about how the firm's pages describe its scope. It is a content fix, not a measurement one.

A fictional monthly review

Consider a fictional four-attorney family law firm in Colorado. Each month it runs 40 invented prompts across divorce, custody, and support, in three cities. The review shows the firm named in a handful of situation prompts in one city and none in the other two, steady AI Assistant sessions in GA4, and two callers who mentioned ChatGPT at intake. The finding is modest and specific: the firm's custody content seems to be reaching one market. The next decision is whether the other two cities lack pages that answer the same questions, which is a content question the firm can act on.

In Hudson South's process, Market Intelligence researches the buyer questions and the market, the same kind of question research a prompt set draws on. The Authority Asset gathers approved business knowledge and evidence, Hudson Twenty-One records how the firm communicates across 21 dimensions, and AEO Creation turns those inputs into answer-first content. For a first look at where the firm stands, see the Hudson South AI visibility check.

For the content side of this work, see how law firms get recommended by ChatGPT and Google's AI, and for the drafting side, whether a firm should use AI to write its blog. More pages are in our law firm content guide.

Start with a clear baseline

On a free 30-minute call, Hudson South walks through how your site and competing firms' sites answer the questions clients ask before choosing a lawyer.

Sources

1. ABA issues first guidance on a lawyer's use of AI tools. ABA News. Read 2026-10-15.
2. Generative AI performance report (Search). Google Search Console Help. Read 2026-10-15.
3. Default channel group. Google Analytics Help. Read 2026-10-15.
4. Publishers and Developers - FAQ. OpenAI Help Center. Read 2026-10-15.