Why One AI Search Tells You Almost Nothing About Your Firm’s Visibility
A firm owner asks ChatGPT for the best personal injury lawyer in their city. Their firm comes up second. They screenshot it and send it to their partners. A week later, a colleague asks the same question and the firm does not appear at all.
Neither screenshot is wrong. Both are incomplete. AI answers are not fixed rankings, and anyone measuring AI visibility by running a single search is measuring noise.
AI Answers Vary by Design
Traditional Google results are relatively stable. Search the same phrase from the same place twice in a row and you will usually see the same list. AI tools work differently. Language models generate answers rather than retrieve a fixed list, and many of them introduce some variation on purpose so that answers do not read identically every time.
On top of that, the answer can shift based on how the question is worded, which model version is answering, whether the tool searched the web live, the user’s location, and what the conversation looked like before the question was asked. “Best car accident lawyer in Tampa” and “Who should I call after a car accident in Tampa?” can produce different firms.
That variation is not a flaw to work around. It is the nature of the channel, and it changes how visibility has to be measured.
What a Single Check Misses
One search can tell you that your firm is capable of appearing. It cannot tell you how often it appears, how it compares to competitors, or whether things are getting better or worse. It also cannot tell you how the firm is described when it does appear, which matters as much as whether it appears at all.
A firm that shows up in a third of answers for its main practice area is in a very different position from one that shows up in nearly every answer, even though both could produce the same screenshot.
How We Measure It
That problem is why MileMark built its own AI visibility measurement tool. The approach is straightforward even if the execution takes work.
We start by building a set of questions that real prospective clients would ask for each practice area and market a firm cares about. Some are direct requests for recommendations. Some are questions about a legal problem where a firm might be mentioned. Some name the firm directly to see how it is described.
We then run those questions across multiple AI models, repeatedly, and record which firms appear, how often, in what position, and with what description. Repeating the questions is what turns individual answers into a measurement. The result is a share of appearances rather than a yes or no.
Finally, we run the same set again over time. A single measurement is a baseline. The trend is what tells a firm whether its website and visibility work is changing anything.
Watch the Description, Not Just the Mention
Appearing in an AI answer is only half the picture. The other half is what the answer says. An AI tool might mention a firm but describe it as handling the wrong practice areas, list an attorney who left, or attach a vague description that does not set the firm apart.
Tracking the descriptions over time shows whether the firm’s own messaging is getting through. When a firm rewrites its homepage and practice pages with clear, specific statements about what it does, those statements often start to show up in AI descriptions. That is one of the clearest signs the work is landing.
What Moves the Numbers
The work that tends to move these numbers comes down to a handful of things. A website with real depth in the practice areas the firm cares about. Consistent name and details across the web. Detailed attorney bios. Reviews and independent mentions that confirm what the website says. And patience, because AI visibility builds over time rather than flipping on.
What does not seem to move the numbers is shortcuts: invented claims, inflated credentials, or content that says nothing specific about the firm. AI systems are built to find useful, trustworthy information, and that is a hard thing to fake.
How to Do a Rough Version Yourself
A firm without access to a measurement tool can still get a better picture than a single screenshot. Write down ten questions a prospective client might ask. Run each one in ChatGPT, Perplexity, Gemini, and Google, and record whether the firm appears. Do it again on a different day. Then repeat the whole exercise a few months later.
It is manual and imperfect, but it replaces one lucky or unlucky result with a pattern, and patterns are what decisions should be based on.
Common Questions
Does it matter which AI tool we focus on?
Prospective clients use several, and usage keeps shifting. It is better to measure across the major tools than to optimize for one.
Can an agency guarantee placement in AI answers?
No. Nobody controls what AI tools say. What an agency can do is improve the sources those tools draw from and measure honestly whether that work is having an effect.
MileMark is a digital marketing agency for law firms that builds websites designed to get found in Google and AI search. Our free website audit includes a look at how your firm currently shows up in AI answers for the practice areas you care about.
