A woman in Studio City asks ChatGPT for the best facial in her neighborhood. It gives her three names, confident and nicely formatted. She does not book any of them.
She opens Google, then Instagram, then the reviews.
That little detour is the most important thing happening in marketing right now, and almost nobody selling "AI search optimization" wants to talk about it.
Here's the data. On October 23, Digiday and Modern Retail published a roundup of new shopper research, and the numbers are blunt. A Feedvisor survey found that 16% of shoppers start product searches on ChatGPT, but only 9% go there to actually buy. Only 5% purchase right after an AI assistant recommends something. Everybody else goes somewhere else to check.
Where do they go? PartnerCentric asked. After an AI suggestion, 33% of shoppers go read existing reviews. Another 22.3% check prices, 18.8% head to Amazon or another retailer, and 14.8% visit the brand's own site. And among the shoppers who trust AI the most, 42.2% still want reviews before they act.
Let that land. The people who love the robot the most still want a human to vouch for the answer.
PartnerCentric's founder and CEO, Stephanie Harris, put it neatly: "AI is a starting point, not a final verdict." Kieran Powell of Channel V Media made the same point with better comic timing. Paraphrasing: if a shopper won't let AI buy them a $40 shirt, they won't let it judge the shirt either.
The delegation numbers tell the same story from the other side. A Prosper Insights & Analytics survey in the same roundup asked people what they'd let an AI agent do on their behalf. Book a restaurant: 12.1%. Buy groceries: 11%. Pay bills: 9.8%. Buy apparel: 7.3%. Nearly nine out of ten people won't even let the machine reserve a table. These are not people waiting to hand over their wallets. They're people using a very fast intern to make a shortlist, then doing the real deciding themselves.
Which brings us back to Studio City. Our woman didn't ignore the AI. She used it exactly the way the data says everyone does: to get a shortlist. Then she went looking for proof, and the business that won her was the one with proof she could believe.
Most of that research was about retail. Now look at local service businesses, where the stakes are higher than a shirt.
BrightLocal's 2026 Local Consumer Review Survey, which polled 1,002 US adults, found that 45% of consumers now use ChatGPT or another generative AI tool for local recommendations. Last year it was 6%. AI has become the third most popular source for finding a local business, behind only Google and Facebook. That's the part the AI evangelists put on the slide.
Here's the part they leave off. In the same survey, 97% of consumers read reviews for local businesses, and 41% say they always do, up from 29% a year earlier. Seventy-four percent only care about reviews from the last three months. Thirty-one percent won't use a business rated under 4.5 stars, up from 17%. And 47% won't touch a business with fewer than 20 reviews.
So yes, the machine is in the room. It's just not the one signing the check.
For a med spa in Beverly Hills, a dental practice in Pasadena or an estate-planning firm in Century City, that changes the whole job. You're no longer marketing to a person. You're marketing to a person and to the AI that briefs them, and the AI is a terrible closer. It can get your name into the conversation. It cannot make anyone trust you with their face, their teeth or their mother's estate.
Something else does that. Something that has been sitting in your business, mostly ignored, for years.
Your clients.
The industry is currently selling the opposite story at volume. Every week there's a new deck about GEO, (generative engine optimization), promising to get your brand "cited" by ChatGPT. Some of it is useful. Structured information, clear service pages and accurate listings matter, and Paris Roche of Musely made a sharp point in the Digiday piece: the goal is accurate information, not just being mentioned. But the dominant pitch treats getting named by the robot as the finish line. It's the starting line. The research says so, over and over.
Now the turn, because there's a version of this argument that's too comfortable.
"Great," says every business owner reading this, "so I'll ask for more reviews." And then they send a robotic email blast begging for five stars, get a dozen two-line reviews that say "Great service!!" and wonder why nothing changes.
That isn't social proof. That's a participation trophy, and shoppers see through that shit instantly.
Reviews are where trust gets checked. They are not where trust gets built. A wall of generic praise doesn't close anyone. What closes people is specificity: a real person, with a real name and a real face, describing the exact thing they were afraid of and what happened instead. The bride who was terrified of looking overdone. The founder who'd been burned by two accountants before. The guy who walked into the boxing gym forty pounds heavier and couldn't finish the warmup.
That's not a review. That's a story. And stories are content.
So the strategy that actually follows from this data isn't "do AI." It's a system that turns client experience into proof, constantly. Ask for reviews at the moment of delight, not in a form email three weeks later. Turn the best of them into short, properly shot client stories, with permission and real people. Put those stories everywhere the shopper goes after the AI hands off: your Google profile, your Instagram grid, your site, the confirmation email. Keep them fresh, because three-month-old proof is already stale for three out of four people.
And then, here's the quiet bonus, that same material is exactly what the AI tools are reading anyway. Fresh reviews, consistent stories, real names and real outcomes give the machine something accurate to repeat. You don't win the AI recommendation by gaming it. You win it by being the business with the most credible evidence on the internet.
We build this for clients as a single loop, not two departments. The review ask, the shoot, the cut-downs, the placement. We'd rather have one client telling the truth on camera than a hundred AI-generated testimonials that look like they were made in a hurry, because they were, and shoppers can tell.
There's one more data point worth sitting with. Digiday also reported on October 27 that Anthropic, one of the companies building these AI tools, disclosed in its IPO filing that its advertising spend went from $33 million in 2024 to $147 million in 2025. The filing describes the brand as something that requires substantial spending, and says that spending will rise.
The companies selling automation are spending more on brand, not less. Make of that what you will. We know what we make of it.
AI is going to keep getting better at the first ten seconds of the customer journey. Fine. Let it. Let it do the shortlist, the comparisons, the "best facial near me." That was always the boring part.
The rest is still a human story. Somebody still has to look into a camera and say, "I was scared, and they took care of me."
Make sure that somebody is talking about you.
