Your Baseline: The Ten-Question Test
Enough theory. Before you fix anything, you need to know where you stand today. Not a guess, and not a feeling. A number you can write down, check again in a few months, and compare.
That number comes from a simple test anyone can run. You ask AI the questions your clients ask, and you write down whether your firm is named.
Why start with a test and not a fix
It is tempting to jump straight to the work: new pages, schema, reviews. Resist that for one afternoon.
Without a baseline, you cannot tell whether anything you do is working. Worse, you may fix the wrong thing. The test often shows a problem you did not expect, like an old address, a partner who left years ago, or a competitor you never think about who keeps getting named instead of you.
A baseline turns AI visibility from an opinion into something you can measure. That is the whole point of this course.
Write ten real questions
Use the questions real clients ask, in the words they would use. Not the keywords you wish you ranked for.
A good set has a mix:
- "Best" questions. "Who is the best estate planning attorney in Coral Gables?"
- Situation questions. The way a worried person describes their problem. "My landlord kept my deposit. Who should I call?"
- Comparison and cost questions. "Is a flat fee or hourly better for a divorce?" or "How much does this usually cost here?"
- Questions with your firm's name in them. "What is [your firm] and who do they serve?" These show you what AI believes about you.
Write the ten questions once and keep them. You will reuse the same set every time, because changing the questions changes the result.
Run them across the engines your clients use
Ask each question in ChatGPT, Google's AI answers, and Perplexity, in a private or logged-out window so your own history does not color the result. If many of your clients speak Spanish, run your top questions in Spanish too.
For each answer, note three things:
- Named: your firm appears in the answer.
- Named with wrong facts: you appear, but with an old address, a wrong specialty, or someone who no longer works there. Treat this as urgent. A wrong fact can cost you more than being left out.
- Not named: someone else got the spot. Write down who.
Save screenshots as you go. They are your proof of where you started.
Your share of answer
Count how many times you were named and divide by the number of answers you checked. Ten questions in three engines is thirty answers. If you were named in three of them, your share of answer is 10 percent.
That one number is your baseline. It will not be perfect. AI answers change by platform, by day, and by how a question is worded, so a single run is a snapshot, not a verdict. That is fine. What matters is the trend over time, measured the same way each time.
What the test tells you
The test does not tell you why you were left out. It tells you where to look.
If AI gets your facts wrong, the problem is usually agreement between your sources. If AI never mentions you, the problem is often that it cannot tell clearly who you are, or that nothing outside your own site vouches for you. If AI names you but describes you vaguely, your pages may not answer questions in a way it can quote.
Those causes have names. The next lesson walks through the five signals AI checks before it names a firm, so you can connect what the test shows you to what to fix.
Make it a habit
Run the same ten questions again every month. Same questions, same engines, same private window. Keep the results in one place, and read the trend each quarter, so you can see the line move.
The firms that make real progress are not the ones that run one big audit. They are the ones that measure, fix the weakest thing, and measure again.
Go deeper:
The rest of this academy is free
See every lesson in AI Visibility: Found, Trusted, and Cited by Name