Google Will Let a Free AI Agent Text Your Customers. Fix Your Business Profile Before You Say Yes

Smartphone lying on a warm wooden cafe counter next to a glass jar of cookies in morning light

6 min read

“Answer routine questions from customers, instantly and at no cost.” That is how Google describes the AI agent it is now offering to more and more businesses through their Google Business Profile, according to Search Engine Roundtable’s report from late October. For a solo owner who misses messages while working, that sentence is very tempting. Free, instant, and you can take over any time. What is there to think about?

Quite a lot, as it turns out. My view is simple: this is a feature worth turning on, but not this week, and not before you fix the one thing it depends on. Free is a great price. It is a bad reason.

What Google is actually offering

Here is what is confirmed so far. Google has been inviting Business Profile owners into an early access pilot of AI-powered messaging. French SEO blog Yes We Blog reported that invitation emails went out on October 20, 2026, and Search Engine Roundtable reported on October 29 that access appeared to be widening.

The setup works like this:

  • Customers who find you on Google Search or Maps can message your business from their phone’s own texting app.
  • An AI agent replies to routine questions on your behalf, such as hours, menu items, specials or reservation times.
  • When the agent does not know an answer, it hands the conversation off to you, and you can step in at any point.
  • All the messages live inside your Business Profile.

There is one fine-print detail worth knowing. The invitation footer, quoted by Yes We Blog, says the agent only answers customers on participating mobile networks whose phones support RCS (Rich Communication Services, the upgraded texting standard that adds typing indicators, photos and buttons to ordinary text messages). Everyone else’s messages go straight to your phone. So even with the agent on, some customers will still be waiting on you personally.

The agent can only be as right as your profile

The single most important line in the coverage came from a practitioner, not from Google: the agent can only answer as well as your profile does. In plain terms, the AI does not know your business. It reads what you have published on your Business Profile and repeats it confidently.

Now think about the state of a typical solo owner’s profile. Holiday hours that were never updated. A services list from two years ago. A price that changed in the spring. A photo of a menu item you stopped offering. None of that hurts much when a human reads it, because people expect listings to be a little stale and they call to check. It hurts a lot when an agent restates it in a friendly, instant reply that sounds like you.

That is the core of my skepticism. A wrong answer from a website is a stale page. A wrong answer from a chat agent feels like a promise. If the agent tells someone you are open on the Friday after Thanksgiving and you are not, the customer does not blame Google. They blame you.

Three reasons not to rush

1. It is a pilot, and pilots change

Search Engine Roundtable noted that Google had already made updates to the feature since it first appeared in June. That is normal for an early access program, but it means the behavior you test today may not be the behavior you live with in January. Pilots are for businesses that have time to watch closely. If you are heading into your busiest season, you probably do not.

2. The holiday season is the worst time to start an experiment

The weeks between now and New Year are when hours shift, stock runs out, shipping cutoffs bite, and customers ask unusual questions. That is exactly the kind of situation a routine-question agent is weakest at. Turning it on in late November means its first real test happens during your highest-stakes month, with the least attention you can spare.

3. “Free” hides a cost in your attention

Every hand-off from the agent lands on you, and you will want to review what it said in the conversations it did handle, at least at first. If you are not prepared to skim those conversations every day for the first couple of weeks, you will not catch the bad answers until a customer does. We have written before about how AI tools can quietly create more work when nobody owns the checking step. This is a textbook case.

The case for saying yes, eventually

To be fair, the upside is real. Many small businesses lose leads simply because nobody replies for hours. A free agent that handles “are you open today?” in seconds, and hands off anything harder, could recover some of those customers. It also lives where people already search for local businesses, which matters more every month as discovery moves into AI-powered answers. If you have read our beginner’s guide to GEO, you already know that a clean, complete profile is becoming your front door.

So this is not a “never”. It is a “not yet, and not like this”.

A calmer way to adopt it

If you receive the invitation, here is the order I would follow:

  1. Audit your profile first. Read every field as if you were a stranger: hours, special hours, services, prices, attributes, the business description, and your photos. Delete or update anything you would not want repeated back to a customer word for word.
  2. Add holiday hours now. Enter special hours for Thanksgiving, Christmas Eve, Christmas Day, New Year’s Eve and New Year’s Day, even if they are your normal hours. Blank holiday fields are where agents and humans both guess.
  3. Write down your “never” list. Decide which topics you never want handled without you: refunds, custom quotes, complaints, anything medical, legal or safety related. When the agent hands those off, reply yourself.
  4. Turn it on during a quiet week. For most solo businesses, that is early to mid January. Check every conversation daily for two weeks, then weekly.
  5. Keep a human reply time promise. State somewhere visible how quickly you personally reply to messages the agent passes along, and keep it.

One more caution. Any AI that reads text from the outside world can be nudged by cleverly written messages. Google’s agent is designed for narrow, routine questions, but if you have not met the idea yet, our beginner’s guide to prompt injection explains why “the agent said so” should never be the final word on anything involving money.

The bigger pattern behind this pilot

Google’s messaging agent is one more sign of a shift we keep seeing: AI systems are increasingly talking to your customers using the information you publish, whether that is your Business Profile, your website or your policies. The businesses that do well in this shift will not be the ones who switch on every free agent first. They will be the ones whose published information is accurate enough that any agent, from any company, gets the answer right.

That work is boring, and it is free, and it pays off no matter which AI wins. Start there. Then, when January is quiet, let the agent in and watch it closely.

Have you received the Google invitation yet? If you have tried the agent, tell us in the comments what it got right and what it got wrong.

Related reading

Leave a Comment

Scroll to Top