How to Track Your Law Firm's AI Citation Rate (Method and a Copyable Template)

Short answer: Citation rate is the percentage of AI answers, across a fixed set of questions and engines, in which your firm is named or your page is cited. Measure it by building a query list of 40 to 60 questions a prospective client would ask, running each on ChatGPT (search on), Perplexity, Gemini, Google AI Overviews and Google AI Mode three times, and recording whether the firm appears and who else does. Monthly. The sheet takes about four hours a month by hand. The result is a number that goes up or down for reasons you can act on, which is more than a ranking report offers now.

Why measure this at all?

Because the AI answers are where a growing share of the "which lawyer" decisions are being shaped, and because they do not appear in Search Console or in any rank tracker. A firm that does not measure them is guessing. We have logged a fixed Los Angeles query set since 2024; the number moves, and it moves in response to specific changes (a courthouse page published, a directory profile corrected, a review pace sustained) in ways that are visible only because the baseline exists.

What is the query list?

Forty to sixty questions in three groups:

The list is fixed for a year. Adding a question mid-year breaks the trend; add it to next year's list.

Which engines and how many runs?

Five surfaces, three runs each per question per month, from a logged-out or fresh session, with the location set to the target city where the product allows it:

1. ChatGPT with search enabled
2. Perplexity
3. Gemini
4. Google AI Overview (record "none" when no Overview appears)
5. Google AI Mode

Three runs because the answers vary. In our logs, the same question returns a different set of named firms on roughly a third of repeat runs. One run per month is noise; three lets you say "named in 2 of 3".

What gets recorded?

One row per question per engine per run:

• Date, engine, question, run number
• Firm named? (yes / no)
• Firm's page cited as a source? (yes / no, with URL)
• Other firms named (up to five)
• Other sources cited (domains)
• Notes (Overview absent, answer refused, etc.)

From that, the monthly numbers:

• Citation rate: rows with "firm named" or "page cited" divided by total rows.
• By engine: the same, per engine, because they differ a lot.
• By group: find-a-lawyer vs process vs local.
• Share of voice: how often each competitor is named, ranked.
• Source share: which domains are cited most (directories, courts, publishers, firms).

What does a baseline look like?

For most Los Angeles firms with one to five attorneys, the first month's citation rate is in the low single digits, and it is often zero on ChatGPT for find-a-lawyer questions. That is not a failure; it is the starting line. Local and process questions move first, because a specific page can be cited within weeks of publication. Find-a-lawyer questions move last, because they depend on the directory and review signals that take months.

What are the rules that keep it honest?

• Fixed list, fixed engines, fixed runs. Change nothing mid-year.
• Log the misses. A month with no citations is data.
• Do not count the firm's own name in the question. "Is Smith & Lee a good firm" is not a citation test.
• Do not count a directory page that lists the firm among fifty as a citation. The firm is named in the answer or its page is cited, or it is a no.
• Separate the engines. Perplexity cites far more freely than ChatGPT; averaging them hides the difference.
• Publish the method with the number. A citation rate without the query list and run count is not a number.

The template

A sheet with these tabs: Queries (the list, with group), Log (one row per run, the columns above), Monthly (formulas: citation rate overall, by engine, by group; share of voice; source share). We keep ours as a spreadsheet, not a tool, on purpose; it is auditable by anyone. Ours is described on the methodology page.

What does not work?

• Tools that report a single "AI visibility score" with no query list. Ask what the questions were.
• One run, one engine, one month.
• Asking the model "do you recommend Smith & Lee". Leading question; not a citation.

What do you do with the number?

Look at the misses. For each process or local question where a competitor or a directory is cited and the firm is not, check whether the firm has a page that answers it; usually it does not, or the answer is buried. Write or fix the page, wait a month, re-run. For find-a-lawyer misses, check the directory consistency and the review pace. The AI citation audit is this method run for a firm as a one-time report; the free visibility check is a shorter version of it. The wider context is in the AI search visibility guide.