The light comes on, and the driver asks an assistant who can look at it today.

A driver asks about the job, not about a trade

Most drivers who put a car question to an AI assistant are describing something: a noise, a warning light, an inspection due this week, a part that has given out, or a routine job they need fitted in. The question names a job, a place and often a time: today, this weekend, before a trip.

With web search on, an assistant writes its answer from pages it fetches for the question; without it, from what its model learned in training. On either route it can only name a shop for brakes if something it read says that shop does brakes.

The follow-ups are what a driver would otherwise call around to ask: what the diagnosis costs, whether they can wait, whether the work is guaranteed, how they get home. Answered in text on your site, they are there for an answer to repeat.

Ten facts a driver compares, and where each one goes

AutoRepair, the schema.org type for this trade

AutoRepair is defined on schema.org as a car repair business. It sits under AutomotiveBusiness (car repair, sales or parts), which sits under LocalBusiness; we checked both entries. Declare it and fill in what Google's local business documentation reads (name, address, telephone, hours), plus areaServed for a towing radius and makesOffer for the services, each an Offer.

AutoRepair has no property for the makes a shop works on (we read the full list), so the makes go in the text and in each Service name. Leave ratings out: Google's local business documentation recommends aggregateRating only for sites that collect reviews of other businesses. Every value in the JSON-LD must also be on the page, as Google's structured data introduction asks of all structured data.

{
  "@context": "https://schema.org",
  "@type": "AutoRepair",
  "@id": "https://www.example.com/#shop",
  "name": "[Shop name]",
  "url": "https://www.example.com/",
  "telephone": "+1-555-0100",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "[Street and number]",
    "addressLocality": "[Town]",
    "addressRegion": "[State or province]",
    "postalCode": "[Postal code]",
    "addressCountry": "[Country code]"
  },
  "openingHoursSpecification": [{
    "@type": "OpeningHoursSpecification",
    "dayOfWeek": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday"],
    "opens": "[opening time]",
    "closes": "[closing time]"
  }],
  "areaServed": ["[Town you tow from]", "[Town]"],
  "makesOffer": [{
    "@type": "Offer",
    "name": "Check engine light diagnosis",
    "price": "[price]",
    "priceCurrency": "[currency code]",
    "itemOffered": { "@type": "Service", "name": "Check engine light diagnosis" }
  }]
}

What our crawl of 9,483 sites found, read for a repair shop

In late August and early September 2026 our audit sent a crawler, running no scripts and obeying robots.txt, to the homepage, contact page and about page of 11,060 small business sites in 40 trades and 90 metros across the US, Canada, Europe and the UAE; 9,483 answered. It publishes no repair-shop figure, so every number here is sample-wide.

On 61.5% of the sites there was no question-and-answer content at all. 43.1% gave inconsistent answers about who and where they are across their own three main pages, with more than one phone number on 36.9% and no postal address a machine could parse on 34.1%. 17.8% had no schema.org markup at all. 7.2% served our crawler a challenge page instead of the site, and 2.1% kept real content, such as a service menu, in a PDF.

64.7% showed no review signal a crawler can read. Close that with text: real reviews that name the job, quoted with the reviewer's first name, the date and the platform. Keep platforms' ratings out of the markup, since Google's review snippet documentation treats reviews a business controls about itself as self-serving, and such pages get no review stars.

For 9,240 of the businesses we also put one buyer question from their trade and metro to Claude, with web retrieval on: 14.9% were named, 17.7% among the 5,928 answers that gave a differentiated list rather than a hedge. One question, one assistant, one day: a direction, and Claude's only.

Where a shop's work starts, with the article for each step

1. Check what a crawler receives: run the crawler access check and view a service page with scripts off, as what OpenAI's crawlers fetch from a repair shop explains.
2. Settle one name, one address and one phone number for every page and listing.
3. Write the ten facts as text: a page per main service, with a starting price or range and the warranty.
4. Bring the Google listing into line with the site. The repair shop's Google Business Profile goes field by field.
5. Give each specialty you can staff a page of its own; specialty pages sets out what goes on one.
6. Measure before and after with a fixed set of drivers' questions, as checking whether your shop is named describes. A dealership's service department has its own article.

Each step leans on the ones before it. The measurement comes last, by the method our methodology sets out, because it shows whether any of it changed an answer; nobody outside the companies that build the assistants can promise that it will.

Sources: our audit of 9,483 small-business sites, with its method · schema.org, LocalBusiness and its subtypes · schema.org, Service · schema.org, Offer · Google, local business structured data · Google, introduction to structured data · Google, review snippet