AI Visibility Tracking for Law Firms: What to Measure and How Often

Most law firms already know how to read an SEO report: rankings, organic traffic, calls, cost per case. AI visibility needs its own set of numbers, because an AI answer does not have a position one through ten and often does not produce a click at all. This post covers what we track for law firms, how to collect it, and how to tell a real change from noise.

The core number: answer share

The single most useful number is how often your firm is named when someone asks an AI assistant the questions your clients ask. We call it answer share: the percentage of AI answers for your market’s discovery questions that name your firm.

To calculate it you need three things:

1. A fixed set of prompts. Questions a potential client would ask without already knowing your name, such as “best car accident lawyer in [city]” or “who should I call after a truck accident in [city].” In The Answer Share we use ten.
2. A fixed set of platforms. At minimum ChatGPT, Gemini, and Perplexity. Add Copilot and Google’s AI Overviews if you can collect them consistently.
3. A count. Ten prompts on three platforms gives you 30 answers. If your firm is named in 6 of them, your answer share is 20%.

For context, across the 15 markets in our September study, 81% of the firms we scored had an answer share of 10% or less. The highest figure we recorded was 63.3%.

Track answer share for your top competitors with the same prompts. The number means most when you can see it next to the firms you compete with for cases.

Supporting numbers

Answer share tells you whether you are named. These tell you why, and what to fix.

Per-platform share. Break answer share out by platform. A firm can be named often on Gemini and never on Perplexity, because the platforms weigh different signals. The platform where you are weakest usually points to the kind of work you need.

Citations. Which pages each answer used as sources, and whether any of them are on your site. If your pages are cited but your firm is not named, you have an authority problem rather than a content problem. If you are named but never cited, the assistant knows your firm but is not reading your site, which is worth checking for crawler blocks.

Accuracy. Whether the assistant describes your firm correctly: current name, right domain, right city, practice areas you actually handle. Errors here usually trace back to inconsistent listings or an old identity still circulating on the web.

Competitor list. Every firm named in place of yours. Confirm each one actually practices in your market before you treat it as a competitor, since assistants occasionally name a firm from another state.

Branded search and calls. People who see your name in an AI answer often search for you or call from your Google profile instead of clicking through. Rising branded search volume and call volume are the downstream signs that AI visibility is turning into cases.

Where the data comes from

Running prompts yourself. The simplest method: a logged-out browser, the same prompts every month, results recorded in a spreadsheet. It is free and accurate for what it shows. It is also slow, and a single person in a single location sees a narrow slice. Our step-by-step setup is in How to Tell If Your Law Firm Is Invisible in ChatGPT, Gemini, and Perplexity.

Tracking platforms. Several tools now submit prompts to the AI platforms on a schedule and record the answers. They save a great deal of time. Two habits keep the data honest. First, check how many platforms actually returned answers on each run, because a platform that fails to respond can quietly shrink the denominator. Second, spot-check the matches. Firms with common surnames get credited with mentions of a different firm, and firms with unusual names sometimes get missed.

Bing Webmaster Tools. Bing now reports how often your pages are cited in Microsoft Copilot answers and which queries led there. This is measured data rather than a simulation, which makes it valuable. Keep in mind it covers Copilot only. It does not tell you about ChatGPT.

Google Search Console. Search Console does not report AI Overview citations separately in the standard query report, but it can still show the effect. A page holding a strong position with plenty of impressions and almost no clicks is often a page whose query is being answered by an AI Overview above it. Before rewriting the title tag to fix click-through, look at the live results page for that query.

How often to measure

Monthly is the right rhythm for answer share and citations. AI answers vary from day to day, so weekly numbers swing more than they should and tempt you to react to noise. Three or four monthly readings show a real trend.

Quarterly, re-check crawler access and directory accuracy. A website update or new security plugin can block an AI crawler overnight without anyone noticing.

After any big change, such as a site migration, a firm name change, or a new domain, run a full check within a few weeks.

Reading the changes

A swing of one or two mentions out of 30 is usually noise. A sustained move across several months, or a move that shows up on more than one platform, is more likely real.

Also allow for the platforms changing underneath you. We have seen the answers for the same prompts shift between two runs a few weeks apart, moving many firms up or down at once, with nothing announced by the platform and nothing changed by the firms. When your numbers jump, look at whether your competitors’ numbers jumped too. If everyone moved together, the platform changed. If only you moved, it was probably your work.

Putting it on one page

A monthly AI visibility report for a law firm can fit on a single page:

• Answer share overall and by platform, with last month and three months ago for comparison
• The same figures for your three to five closest competitors
• Your pages that were cited, and the outside sources cited most often in your market
• Any errors in how the assistants describe your firm
• Branded search and call volume
• What changed this month and the one or two actions that follow from it

If the report cannot point to a next action, it is only describing the month, so keep the last item even when the answer is “keep doing what we did.”

How we do it

At First Mind AI, this report is the starting point for every client: we measure each platform with a fixed prompt set, confirm the crawlers can read the site before trusting any other number, and tie each month’s movement to the work that caused it. If you would like to see what that looks like, you can test drive the dashboard our clients use, or see how we work with law firms and book a discovery call.