Measurement & Reporting · 9 min read · July 15, 2026
How Often Does the AI Recommend Your Law Firm? A Framework for Measuring It
When a prospective client types "best employment lawyer in Cologne" into ChatGPT, a language model decides which firms get named — and whether yours is one of them isn't something you have to guess at. You can measure it. This guide walks through how to track your firm's AI visibility with a fixed set of test prompts, a handful of metrics, and a repeatable monthly process.
Why AI visibility now matters for law firms
A client's journey increasingly starts not with Google, but with a chat window. Someone with a legal problem is anxious, worried about cost, and just wants to understand what's happening. These are exactly the questions people now ask ChatGPT, Gemini , or the AI overview at the top of a Google search: 'What does a divorce cost?', 'Do I have to sign this warning letter?', 'Do I need a lawyer for a wrongful termination claim?'. Somewhere in that conversation, a line like 'For that, you should talk to a specialist lawyer' tends to appear — and sometimes a specific name follows.
For your firm, that shifts the moment that matters. The old question was: do I rank on page one of Google? The question now is: does the AI even say my name when someone asks about exactly my specialty in exactly my city? This new discipline is called Generative Engine Optimization, GEO for short. It isn't about a click on the third result — it's about whether your name gets spoken as a recommendation, or the AI quietly skips over you.
The uncomfortable truth: most firms have no idea whether they're being recommended. They've never tested a single question. If you don't measure it, you're relying on a gut feeling — and that's especially risky with AI answers, since the same question can get a very different response in Berlin than in a town of 40,000 people.
What 'visibility' actually looks like with an AI
Unlike Google, there's no ranking list you can scroll through. An AI answer is just running text. So your visibility comes down to three questions: are you named at all? Are you recommended, or just mentioned in passing? And is the context right — your actual specialty and your actual location?
Take an everyday example. You're a family law specialist in Freiburg. A test prompt reads: 'I'm getting divorced in Freiburg and I'm looking for an empathetic family lawyer. Who would you recommend?' The AI might name you directly, it might write only 'a specialist firm in the Freiburg area,' or it might name three other firms and skip you entirely. All three outcomes are measurably different — and only the first one actually brings you clients.
It's also worth separating a neutral mention from a genuine recommendation. 'There are several firms in Freiburg, including X, Y, and Z' is not the same as 'For empathetic family law advice, firm X is often recommended.' When you measure, keep those two apart, or you'll talk yourself into a rosier picture than reality.
Step one: an honest baseline measurement
Before you optimize anything, you need a starting point. Sit down with your team and put together 15 to 25 questions your typical clients actually ask — not phrased in legal terms, but the way an anxious person would type them: 'My boss wants to let me go for operational reasons, what do I do?', 'Landlord won't return my deposit, which lawyer can help?', 'Renouncing an inheritance — do I need a notary or a lawyer?'. This is the same language AI users type.
Now run each question through ChatGPT, Gemini, and Google's AI overview. For local questions, always include your city or region — without a location, the AI rarely names specific firms. For each answer, note three things: were you named, yes or no; how were you described — recommendation, mention, or inaccurate; and who was named instead. That last one is your real competitive set in AI search, and it's often not who you'd expect.
Log everything in a plain table: one row per question, a column for each AI system, plus the date. That table is your baseline. It's often a sobering read at first, but it's worth its weight in gold — in three months you'll be able to see in black and white whether anything has actually moved.
The metrics that actually matter for your firm
From your table, a few simple metrics fall out that you can track over time. The most important is your mention rate: what share of your test questions surface your firm's name? If you watch that number climb over several months, that's real, provable progress you can bring to your partners.
The second metric is your recommendation rate: of those mentions, how many are genuine recommendations rather than just a name buried in a list? The third is your correctness rate: is the practice area and location actually right? It doesn't help you if the AI describes you as a traffic-law attorney in Munich when you actually practice medical malpractice law in Hamburg. Errors like that cost you mandates and damage your reputation.
Track all three rates separately by practice area. Many firms find they're already well visible in employment law but essentially invisible in inheritance law, even though both are part of what they offer. Without that breakdown, the averages hide the real gap.
Why the AI names other firms instead of you
When the models skip over you, it's rarely random. Language models rely on what they can find and classify about you across the open web. Firms that get named consistently tend to share three things: clearly structured expert articles on their own site, unambiguous location and specialization details, and mentions on independent sources like legal directories, industry portals, and local press.
A common reason for invisibility is the classic law firm website built entirely out of platitudes: 'competent, committed, personal.' To a language model that's content-free. It has nothing to work with to conclude that you're the lawyer who handles operational layoffs in the chemical industry around Ludwigshafen. The more concrete and thematically coherent your content is, the more confidently the AI matches you to the right case.
Timeliness matters too. An article on a recent ruling, clearly explained with your own take, signals to the AI that you're active in that area. Generic marketing copy without real substance does not.
From measurement to improvement: a worked example
Take a mid-sized firm focused on traffic law in Dortmund. Its baseline shows it named in only two of twenty test questions, both times as a passing mention. The competitors who keep showing up have detailed advisory pages on 'contesting a fine notice,' 'avoiding a driving ban,' and 'challenging an MPU' — each with concrete deadlines and local context.
The firm responds directly: it writes a well-researched, well-structured article for each of its five most common case triggers, clearly naming the city and jurisdiction in each one, and builds out a complete profile on two reputable attorney directories. Three months later, it reruns the exact same 20 questions. The result: eight mentions, four of them genuine recommendations. The mention rate roughly quadruples.
What matters here is repeating the identical questions. That's the only way the comparison stays clean. If you ask different or easier questions the second time around, you're not measuring anything — you're just fooling yourself. Sticking to the same question set every time is the core discipline behind a measurement system you can trust.
A monthly measurement routine that fits real firm life
You don't need an expensive tool for this. A fixed slot once a month is enough, where someone on the team runs through the question set and logs the answers. Budget one to two hours for 20 questions across three AI systems. Do it in a freshly opened, non-personalized browser window, so the AI isn't just naming you because you searched for yourself beforehand.
Add a short note to each measurement round: what changed since last time, what new content went live, which competitors newly showed up. That connects the raw number to a cause. If your mention rate climbs after you publish three expert articles, you know the work paid off.
Keep in mind that AI answers fluctuate. The same question can get a different answer today than tomorrow. That's why the trend across several months matters, not any single day's snapshot. One missed mention isn't a crisis, but a steady decline across three measurement rounds in a row is a real warning sign.
Limits, honesty, and the rules of professional conduct
Be realistic about all of this. You can measure and influence AI visibility, but you can't force it. No credible provider can guarantee that ChatGPT will name you for a given question. Anyone who promises that is selling you a fantasy. What you can actually influence is the probability — through good, concrete, findable content and a clean presence across the web.
As a lawyer, you're also still bound by professional conduct rules. Your AI-optimized content has to stay just as factual and truthful as any other firm communication. Sensational claims or invented success rates, just to get the AI to respond, aren't just unprofessional — they carry real regulatory risk. Good GEO for law firms is always grounded in fact, never loud or boastful.
And finally: measurement is a means, not an end. The number in your table is only worth something if it leads to better, more honest content and, ultimately, to the right clients. Keep that in view, and AI visibility stops being a vague gut feeling and becomes a metric you can actually manage, like any other in your practice.
Frequently asked questions
Is it permissible under professional conduct rules to specifically optimize my firm for AI systems?
Yes, as long as you stay factual and truthful. Attorney advertising rules allow you to describe your practice but prohibit misleading or boastful claims. For law firms, GEO mainly means presenting your real specializations clearly, concretely, and in a way that's easy to find. Invented success rates or sensational claims just to show up in AI answers are both unprofessional and legally risky.
How many test questions do I need for the measurement to be meaningful?
To start, 15 to 25 questions covering your typical case triggers are enough. What matters more than sheer volume is breaking them down by practice area and using the exact same wording every time you repeat the test. That's the only way the comparison stays clean across months. Phrase the questions the way real clients actually type them, not in formal legal language.
Why does the AI name other firms, even though I've been practicing for 20 years?
Experience alone is invisible to a language model. It relies on what it can find and thematically classify about you across the open web. Firms that get named usually have concrete expert articles, clear location and specialization details, and mentions on independent portals. A website built out of pure platitudes like 'competent' and 'committed' gives the AI nothing to work with.
Read on