A homeowner in your service area sat down last night, opened an assistant, and typed some version of a question people used to ask their neighbors over a fence: who should I call about my air conditioner? The machine answered. It named companies, described them, and in some cases explained why it liked them. It did this without visiting a showroom, taking a sales call, or knowing that you exist — because everything it believes about the contractors in your town came from what it could read.
What the machine can actually read
Assistants do not browse the way a customer does. They assemble an answer from sources that are structured, consistent, and quotable — a narrower set than most contractors assume. A handsome page whose text arrives by script, a phone number that exists only inside an image, a service list buried in a scanned PDF: all effectively invisible. What survives the filter is unglamorous and specific.
- Structured data — the schema markup stating in machine-readable terms what you do, where, and when you are open. A page without it is a rumor.
- A tended Google Business Profile whose name, address, and hours match the website exactly; contradictions make a company look unreliable to something sorting for confidence.
- Real reviews in volume and recency — how a model distinguishes a going concern from a listing, per our word-of-mouth filing.
- Genuine city-level pages answering local questions specifically, because "near me" is precisely what these machines are being asked to resolve.
The same work, cashed twice
The convenient part is that none of this is a new discipline. Everything a language model needs in order to recommend you with confidence is what search engines have rewarded for a decade: clean structure, fast pages, consistent listings, honest content with a town in it. Shops that built the boring fundamentals are already being read. Shops that bought the handsome template are being skipped by two audiences at once — and the load-time argument in our autopsy of slow contractor sites applies here as well, since crawlers read the same site your customers do.
“Being the shop a machine has already learned about is cheaper this year than it will be in three.”
What is genuinely new is the timing. This channel is young enough that most markets have no settled answer yet, and models tend to keep trusting sources they have already learned to trust. That is an ordinary compounding advantage of the kind local search has always paid — except the field is, for the moment, nearly empty in most towns.
A necessary caution
None of which requires believing anything in particular about artificial intelligence, and this shop will make no promises about ranking inside a system nobody outside it can audit. Anyone selling guaranteed placement in an assistant’s answer is selling weather. The defensible position is narrower and sturdier: a growing share of the questions once asked over a fence are now asked of a machine, the machine reads, and a company that is legible to it costs nothing extra to build that way.
The preparation is described plainly on the AI search page, the groundwork is ordinary local SEO, and it is the same groundwork that has been paying since long before any of this arrived. The estimate for either is a phone call to the man who does the work.
Filed by Casey Kinnison
Building websites for the trades since 1999. San Antonio, Texas.