Local & Industries · 9 min read · July 15, 2026
What drivers actually ask the AI, and what it means for your shop
More drivers skip the search box entirely and hand the whole problem to ChatGPT, Gemini or Google's AI Overview: "Where can I get my brakes done this week?" What comes back is two or three shop names, not ten blue links. If yours isn't one of them, you were never in the running. Generative Engine Optimization is the work of getting named, and you can test where you stand in an afternoon.
Why drivers ask the AI before Google
Your customers changed how they search without telling you. The driver who used to type "brake replacement price" into Google now writes a full sentence into ChatGPT or reads Google's AI Overview sitting above the results. Either way the ten blue links collapse into one answer that names two or three shops. AI Overviews now reach over two billion people a month across 200-plus countries, so this is not a fringe habit you can wait out.
The reason is convenience, not novelty. Drivers are not mechanics. They have a symptom and a worry: a grinding noise on the way to work, a light on the dash, an inspection due Friday. One answer reassures, explains and recommends in a single breath. Whoever gets named in that breath is already halfway to the counter. Whoever doesn't is out before the phone rings.
This layer sits in front of Google rather than replacing it. Pew Research, watching real browsing behavior across nearly 69,000 Google searches, found people clicked through to a website in 8% of visits when an AI summary appeared, against 15% when it didn't, and clicked a link inside the summary itself in just 1%. The summary is increasingly the entire visit. Serve both layers and you win twice. Serve only the old one and you are working on the half of the journey that is shrinking.
The questions drivers actually ask
Here is what those questions really look like. Panicked: "My engine light just came on yellow, is it safe to drive and who can check it today?" Price-led: "What does a timing belt replacement on a VW Golf cost, and is an independent shop in [your city] cheaper than the dealer?" Long, specific, local, and phrased as a whole problem rather than two keywords. That shape is exactly what these systems are built to answer.
Then the scheduling questions: "Which garage near me does the safety and emissions inspection together on short notice?" Or "who can swap my tires on a Saturday without a two-hour wait?" And the trust questions: "Which shop in [your city] has good reviews for gearbox work?" The driver is not shopping for a business. They are shopping for the certainty of not being taken for a ride.
Notice how much gets packed into one sentence: make, model, symptom, town, urgency, sometimes a budget. The more precisely your pages and profiles cover those combinations, the likelier you are to be the name that comes back. "Your competent partner for all things automotive" covers nothing. The academic paper that gave GEO its name found keyword stuffing bought no measurable gain in its simulated tests, while adding quotations, statistics and cited sources moved the needle most.
What Generative Engine Optimization changes for you
Generative Engine Optimization, GEO for short, is what happens to classic search engine optimization when the answer, not the link, is the product. The job is no longer to rank on page one; it is to be named inside the generated answer and cited as one of its sources. Ahrefs found only about 6-8% of the URLs ChatGPT cites also sit in Google's top ten for the same query, and roughly 80% don't rank in the top hundred at all. Being findable and being quotable are two different competitions.
The practical consequence is that clicks fall while the phone keeps ringing. A driver who hears your name from ChatGPT dials it and never touches your website, so your analytics record nothing at all. SparkToro and Similarweb measured 68% of US Google searches ending with no click to any site in early 2026, up from 60% two years earlier. Your traffic graph can drop while your bays are fuller than ever. Read that graph as failure and you will cut the thing that is working.
GEO runs on facts a machine can lift without guessing. Google is blunt about this: its AI features are rooted in the same Search ranking and quality systems as everything else, and no AI-specific file or markup is required. There is no private channel to buy. What helps is plain language and structured data: unambiguous hours, one line per service, the makes you actually work on, real reviews, prices as numbers. Treat markup as an extractability aid rather than a ranking lever, since Google has said since 2018 that schema is not a direct ranking factor. The cheapest advantage available to a small shop is being the clearest source in town.
How to check whether the AI names you
The first check is free and slightly humbling: be your own customer. Open ChatGPT, Gemini and Perplexity and ask what your customers would ask, word for word. "Best shop for brakes in [your city]". "Cheapest tire change in [your city]". "Independent garage in [your city] with good reviews". Write down which names come back, and whether one of them is yours.
Do it systematically, not once. Take ten to fifteen typical questions and run them over several days, because these answers move. ChatGPT's share of citations to Reddit reportedly fell from around 60% to around 10% inside a month in 2025 after a platform-side change, and Profound found citation frequency inside a conversation dropping from 12.6% of first exchanges to 3.0% by the twentieth. One good result proves nothing. Log question, engine, shops named and where you landed, and you have a baseline instead of a hunch.
Read the sources, not just the answer. Perplexity and Google's AI show where the material came from. If a local directory, your Google reviews or a car forum keeps appearing, that is the machine telling you where your effort belongs. Ahrefs, looking at 1.4 million real ChatGPT prompts, found citation rates vary wildly by channel: around 88% for general search results against under 2% for Reddit. Chase the sources that actually get cited for your questions rather than every platform at once.
Your Google Business Profile does the heavy lifting
Nothing gets pulled into these answers as often as your Google Business Profile. Hours, address, phone, services and reviews feed straight in. When the underlying data is thin or stale, the model does not go quiet, it guesses. Columbia's Tow Center ran 1,600 source-identification queries across eight AI search tools and found more than 60% of answers wrong, with the tools rarely admitting any doubt. Confident and wrong about your hours hurts you more than being absent. So audit the details: are the holiday hours right, is air-conditioning service listed, is the safety inspection listed?
The service list is the most underused field in the whole profile. Enter every job on its own line, named the way a customer would say it: brake service, timing belt replacement, wheel alignment, tire change, manufacturer-scheduled service, safety and emissions inspection. Add the makes you work on. Each line is another question you can be the answer to. A profile with three vague categories gives a machine almost nothing to quote.
Reviews are the other lever, and the text matters more than the star count. "Brakes done same day, price matched the quote" hands the machine exactly the link between a job and an outcome that drivers ask about. Third-party mentions are also the best-evidenced signal anyone has found: across roughly 75,000 brands, Ahrefs measured brand web mentions correlating with AI citation at about 0.66, roughly three times as strongly as backlinks. Ask happy customers to name the specific problem you solved, and answer reviews yourself.
Writing a site the machine can quote
Most shop websites are photographs and adjectives, and a machine can do nothing with either. Give every service that matters its own page, headed the way a human would actually search: "Brake replacement in [your city]". Say what the job involves, roughly what it costs, how long the car is with you and which makes you do it on. That is the shape of an answer, which is why it gets used as one.
An FAQ block earns its keep faster than anything else on the site. Use the real question as the heading and answer it in two or three sentences. "What does a service on a BMW cost?" "How long does a tire change take?" "Does my car need a new timing belt at 100,000 kilometers?" You are pre-assembling the paragraph the model would otherwise have to build itself. Keep it honest, though: Google explicitly warns against spinning up separate content "for AI", which it treats as scaled content abuse.
Technically, schema markup for a local business is worth having. It lets a machine read your address, hours and services without inferring them, and it cuts down on invented details like the wrong price or the wrong closing time. It is not a ranking lever, and neither is llms.txt: John Mueller confirmed in 2025 that no Google Search system reads it, and an Ahrefs look at roughly 137,000 sites publishing one found about 97% got no measurable referral traffic from it. Have your web provider add the schema, skip the folklore, and write for humans who happen to be read by machines.
Three mistakes that keep you out of the answer
The most expensive mistake is data that disagrees with itself. Different hours on the site than on Google, an old number in a directory, a third address on Facebook. A retrieval system facing three versions of you has no way to choose, and the shop with one consistent story wins by default. More than half of Gemini's and Grok's citations in the Tow Center study pointed at fabricated or broken URLs, which is what happens when machines fill gaps on their own. Name, address and phone should match everywhere, down to the hyphen.
The second is refusing to talk about money. Plenty of shops publish no prices at all, on principle. For a cost-anxious driver, and for a system trying to produce a useful sentence, that is a dead end. You don't need a full price list. A range or a starting figure is enough: "front brakes from 180 euros including parts" is something an answer can carry. "Price on request" is something an answer quietly drops.
The third is neglect, though not for the usual reason. Freshness itself is not the magic: in that same Ahrefs study of ChatGPT prompts, the median cited page was around 500 days old. What hurts you is stale data being wrong data, a number nobody answers, hours that changed last spring, a service you stopped offering. Ten minutes a week keeping photos, posts, hours and services true beats one big campaign that ages badly.
A plan for the next six weeks
Start small and keep it honest. Week one: run your fifteen most important questions through the AI systems and write down exactly what comes back. Week two: get the Google Business Profile right, every service on its own line, correct hours, current photos. Week three: ask five to ten recent customers for a review and ask them to name the specific job you did. None of that needs a budget.
Then the website. Build one concrete page for each of your three highest-revenue services, with a price range and a realistic turnaround. Add the FAQ block. Get the schema added if it isn't there. Re-run the same fifteen questions six to eight weeks later and compare against your first table. That comparison, not a vendor's dashboard, tells you whether anything actually moved.
Then keep going, because this is a habit rather than a project. The engines change how they retrieve without announcing it, competitors catch up, your own service list drifts. Whoever tests, corrects and sharpens on a schedule stays in the answer. The work is smaller than most people expect, and the payoff is specific: your name arrives at the moment a driver is standing next to a car that won't start.
Common questions
How do I tell whether ChatGPT or Google knows my shop exists?
Ask them. Put your customers' questions into ChatGPT, Gemini and Perplexity, things like "good shop for brakes in [your city]", repeat over several days because the answers move, and note whether your name appears. On Perplexity and Google's AI, read the cited sources too: they tell you which directories, forums and profiles are actually feeding the answer in your market.
Do I have to publish prices for the AI to recommend me?
You don't need a full price list, but a range or a starting figure helps a lot. Cost is one of the things drivers ask about most, and "from 180 euros including parts" is a sentence an answer can pass along. "Price on request" gives it nothing to work with, so it reaches for the shop that was willing to say a number.
Which matters more, my website or my Google profile?
Both, but fix the Google Business Profile first, because it is what these systems pull from most often for a local trade. Get it complete and consistent: every service listed individually, correct hours, current reviews. Then build out the site with real service pages and an FAQ block. The deciding detail is that both tell exactly the same story about your name, address, hours and phone number.
Read on