What AI visibility means for a driving school, and how to measure it

What does being named mean for a driving school?

Being named means the school's name appears in the answer to a question it could have been the answer to. Three things can happen there, and they are not worth the same. The school can be mentioned, one name in a list of five. It can be recommended, singled out as the one to call for automatic lessons or for a nervous beginner. Or it can come first, the name a parent reads before any other.

Our methodology does not weight them, and it does not separate a recommendation from a mention. Its visibility score is the share of the panel's twelve questions in which the assistant named the school, from 0 to 100, with first place and the other schools named reported beside it. With one answer per question, that is the mention rate for the panel, the share of answers that name the school at all. For a school the useful part is what sits under both numbers: which questions it is named for, which it is missing from, and who is named instead.

What the answer says counts as much as whether the name is in it. An answer that names a school but gets its pickup area wrong, or quotes a package price from two seasons ago, sends a parent to the phone expecting something the school does not offer. So the measurement keeps the description as well as the verdict, word for word.

Why a driving school's panel needs two voices

Measurement starts with who asks. For a driving school that is two people who rarely use the same words: a learner who wants to know whether lessons come in an automatic and how soon the test can be, and a parent who wants the total cost and to know whether the instructor collects from school. A panel written in one voice measures half of the people who decide.

Where a school teaches adults who have moved from another country, they are a third voice. Their questions turn on facts only the school can state for its own area: lessons in a language other than English, a refresher before the road test, a car unlike the one they drove at home. Write those questions the way they are typed, doubt included ("do i need lessons if i already drive in my country").

The three shapes do different work. A problem question shows whether the school turns up at all when someone asks for a school. A comparison question shows how it is described beside the one a parent is weighing it against. A named question shows whether what an assistant says about the school is true. Our panels for paying clients hold twelve questions, mixing all three shapes and every voice.

Leave lesson questions out of the panel. "How many lessons does a beginner need" is a question the school's own site should answer, and collecting those is a separate job, covered in what learners and parents ask an assistant. The panel holds only the questions whose answer is a school.

How to run the panel so one month compares with the next

1. Fix the questions and the place. Write each question once, with the town or neighborhood in the words your callers use, and do not edit it afterward. Months only compare when the questions behind them are identical.
2. Ask each question of ChatGPT, Claude, Gemini and Perplexity, and ask it three times in each. Answers differ between runs, and one run of one question is close to a coin flip.
3. Read every answer and record four things: whether the school is mentioned, whether it is recommended, whether it comes first, and which other schools are named.
4. Keep each answer word for word, with the date and the assistant, so any verdict can be checked later.
5. Score the month. Run the same panel the next month, and the month after that, without changing a word.

Note which answers cite sources, and which pages they cite. A source citation shows where a description came from: a school whose pickup area is wrong in the answer can see whether the error came from its own page, a directory or a competitor's. An answer written from training data alone carries no such trail, and that data stops at a knowledge cutoff, so it can describe a school as it was.

The arithmetic of telling a real change from run-to-run noise is in how to measure AI visibility, and the file format for the panel and its records is in setting up prompt monitoring. Neither needs repeating here. What is particular to a driving school is who the questions come from, which the table above settles.

What one question per business showed across 40 trades

Our audit ran a narrow version of this across its whole sample. For 9,240 small businesses in 40 trades, driving schools among them, we put one buyer question from their own market and trade to Claude, with web retrieval on, and checked whether the business was named. 14.9% were. Among the 5,928 questions where Claude answered with a clearly differentiated list rather than a hedge, the figure was 17.7%.

That is not a driving-school rate, and we publish no cut for the trade. With one question per business, put once to one assistant, it is a direction rather than a measurement of any school. What it shows is the shape of the gap: most of the businesses were absent from the answer to a plain question about their own trade in their own metro.

When we inspected the businesses that were named, their sites had answer-shaped content and a consistent identity. For a driving school, our reading of that is packages, prices and pickup area stated as text, and one name, one phone number and one address used the same way on every page. The overview for driving schools lists those facts in full.

What to read in the first run

The first run is the baseline, and it is worth most before anything on the site changes. Read it against your own pages, question by question. If a parent asked for pickup from a high school and your site never says you offer it, the answer had nothing of yours to go on; that gap is a fix, not a mystery.

Then read what the answers say about the schools named instead. Where an answer cites them, open the cited page and see what it states that yours does not: automatic cars, a pickup area, a language, a price. Across a month of answers, your mentions as a share of every school named is your share of voice, and the other names tell you who you are measured against in your area.

Last, check the description of your own school line by line. A wrong price or an old address in an answer can often be traced through its sources, and then the fix is on the page it came from. Publishing the facts is the subject of the other articles for driving schools: the prices page, pass rates and reviews, and what the crawlers can reach.