Google's AI Is Calling Your Business. Now What?

A homeowner searches for a roofer, taps a button, and Google's AI starts dialing. It calls the roofers near them, asks about price and availability, then texts the answers back. That is agentic calling. It is live across most of the United

A homeowner searches for a roofer, taps a button, and Google's AI starts dialing. It calls the roofers near them, asks about price and availability, then texts the answers back.

That is agentic calling. It is live across most of the United States right now, and if you run a home repair, beauty or pet care business, your number is already on the list.

Nobody asked your permission. Google's I/O 2026 Search announcement named those three categories and said users could "ask Google to call businesses on your behalf," rolling out to everyone in the U.S. over the summer. In late July Google widened it again to cover product availability at local retailers, so a shop selling toys, electronics or beauty products is now in scope too.

The caller is a machine and the customer behind it is real

The instinct is to treat this as spam. It isn't. Every one of these calls has a person at the other end of it who wanted an answer badly enough to ask Google to go get it.

Think about what that person just did. They didn't browse. They didn't compare five websites. They outsourced the shortlisting entirely and asked for numbers. That is the most qualified inbound you will get all week, arriving through the least glamorous channel you own.

Google's own Business Profile documentation is blunt about the mechanics: the calls are placed on behalf of customers using Google Search or Maps, the service is free, and it runs in the United States excluding Indiana, Louisiana, Minnesota, Montana and Nebraska. What it does not publish is the script. Google tells you how to make the calls stop. It does not tell you what the agent says when it starts.

Home repair, beauty and pet care are the categories with the worst phone hygiene

This is not a coincidence. Google picked three categories where the owner is physically unable to answer the phone.

The roofer is on a roof. The groomer has both hands on a dog. The technician is under a sink with the water running. These businesses run on callbacks and voicemail, and they have done for thirty years. Google just pointed an automated caller at the exact segment least equipped to pick up.

Ranking got you the call. Ranking has never once been able to answer it.

— Bunny Honey Club

The uncomfortable part is that ranking is still the entry ticket. Google calls the businesses it considers relevant to the query, and relevance comes out of the same local Search and Maps results you have been fighting over for years. So the four things that actually move local SEO still matter. They just stopped being the finish line.

Voicemail is now a disqualification, not a delay

Here is the mechanical difference from a normal missed call.

A human caller who hits your voicemail might leave a message. They might call back tomorrow. They might remember your sign on the van. An AI caller does none of that. It logs "no answer," moves to the next number on the list, and the customer receives a text with three price quotes and yours is not among them.

There is no second impression. There is no callback window. The comparison closed while you were under the sink.

That changes the math on call handling that we have written about before. Missed-call text-back was always the cheap patch: you miss the call, an automation fires a text, the human reads it later and re-engages. Against an AI caller, that patch does nothing. There is no phone to text back. The agent has already reported and gone.

The winning answer is a number, not a promise

We build and run AI phone receptionists for small businesses, so we have watched what happens when a machine calls a line a machine is answering.

The surprise: an AI caller is the easiest call our system handles. No background noise, no accent to parse, no small talk, no "hang on, let me find my glasses." It asks a short structured question and waits. A voice agent that struggles with a real human on a building site handles Google's agent perfectly.

The hard part was never the audio. It was the script.

Most receptionist scripts, human and AI alike, are built to defer. "We'd need to see it first." "Someone will call you back with a quote." "It depends on the job." Those are fine answers to a person, who will accept a callback because they have already invested five minutes in you. They are a failed call to an agent that is collecting price data for a comparison. It records "no price given" and your competitor's "$400 to $900 for a standard callout, next slot Thursday" wins the text message.

So we now brief every client on the same change: put a real number in the first thirty seconds. A band, not a quote. A range with the conditions attached is honest, defensible, and it survives the summary. "It depends" does not.

Every channel you own is getting an agent in front of it

Voice is not the outlier here. It is the third one.

Shopping went first: AI agents now read product data and decide what to surface, which is why we audited the structured data on our own Shopify stores instead of guessing. Messaging went second, with Meta putting business agents inside WhatsApp and Instagram. Now the phone.

The pattern repeats each time. A machine arrives at your front door with a short, structured question, and it does not wait. Businesses that answer in a format the machine can pass along get the customer. Businesses that answer with a promise to follow up get filtered out before the human ever sees them.

That is the whole shift, and it is not really about AI. It is about whether the thing you say in the first thirty seconds is usable by the person who asked. It always should have been. Now there is a scoreboard.

If your line is still going to voicemail at 2pm on a Tuesday, the fix is not complicated, and it is the same fix whether the caller turns out to be a person or a piece of software. Something has to pick up, know your prices, and be able to book. That is an AI receptionist's entire job description, and it now has a second reason to exist.