Every operator has heard some version of “AI is changing search.” Most are still waiting for someone to show them the numbers. Over the past three weeks, that evidence arrived from more than a half-dozen independent research teams, and the findings all point the same direction: your review profile has become the price of admission to AI recommendations. Here’s the data and what it changes about how you run your reputation.
Where AI gets its answers
When a homeowner asks ChatGPT or Google’s AI Mode for a plumber, the answer gets assembled from sources the model can reach. One study logged nearly 2 million citations across 60,970 AI answer checks and found that Google Business Profile is the single most-cited source, accounting for 28.6% of all citations and appearing for 94% of businesses tested. Yelp ran second at 9.5%.
Notice what’s not at the top of that list: your website. Separate experiments tracking 775 citation events across six AI engines found that earned third-party sources produced 85.8% of mentions, versus 14% for content the brand owns and publishes itself. What other sources say about a business outweighs what the business says about itself, roughly six to one. For a contractor, those third-party sources are your Google, Yelp, Angi, and BBB profiles. And those profiles are made of reviews.
Volume is a gate, not a lever
Steady Demand analyzed 7,990 local queries across 50 metros and measured which businesses got cited inside AI answers. Cited businesses averaged over 1,100 reviews and a 4.75 rating. More interesting was the shape of the result: the difference between cited and uncited wasn’t gradual, it was pass/fail. Below the bar, businesses weren’t ranked lower, they were absent. And once above it, more reviews didn’t buy more citations.
A second study of 120,000 AI mentions points the same way: review volume predicted whether a business got mentioned better than star rating did in most categories tested. Businesses with high volume and lower ratings often got mentioned more than businesses with great ratings and thin volume.
AI systems appear to treat review count as evidence that a business is real, established, and currently active. The old target was protecting a 4.8. The new target is a count that keeps climbing.
Recency counts more than the pile
The freshest data point comes from an analysis of plumbing contractors published in Plumbing & Mechanical this month: a profile with 60 reviews that are all three years old consistently underperforms a profile with 25 reviews collected in the last six months. The same analysis found the median plumbing company holds 23 Google reviews while the companies in the top three map positions hold 87 or more.
That first finding deserves a minute. Most operators think of reviews as a total, a number that sits on the profile and only goes up. The data says Google reads the flow, not the pile. A profile collecting eight reviews a month reads as a business currently serving customers well. A profile with twice the total and nothing recent reads as a business that used to.
Activity signals extend to how you respond, too. Research comparing expanding and contracting multi-location brands found the growing brands answered 72.4% of their reviews versus 43.6% for the shrinking ones, and the growing brands were recommended by ChatGPT several times more often. These systems reward profiles that look actively managed.
Does getting named actually matter?
Fair question, and until this month there wasn’t a clean answer. New research measured what happens after an AI assistant names a brand: 14.2% of those users visited the named brand within seven days, versus 3.8% for the competitor. Across every category tested, the named brand won by roughly two to four times. Being the name in the answer roughly triples the odds the customer follows through with you.
There’s a second half to that journey. Consumer research this year found that only 19% of people contact a business straight off an AI recommendation. A third go read reviews first. So your review profile works two shifts: it gets you named in the answer, then it survives the verification visit that follows.
Why you won’t see any of this in your analytics
Here’s the part that trips up even sophisticated operators: AI-driven business rarely shows up labeled as AI-driven business.
Google itself acknowledged this month that Search Console’s AI search reporting doesn’t accurately represent how a site appears inside AI answers, with impressions counted for links nobody saw and positions reported for the whole AI block rather than your place in it. Meanwhile, most consumers who use AI in their research also use regular search, which means the AI-influenced customer usually arrives days later as branded search or a direct call. And clicks in general are draining out of the journey: researchers found AI Overviews cut search referrals even to Wikipedia, one of the most cited sources on the internet.
So if your traffic is flat while your reputation improves, you’re reading the wrong scoreboard. Calls, form fills, and booked jobs are the scoreboard now, and those can grow while sessions fall.
The new bar, in practice
Add it all up and the playbook is short. Build volume past your market’s threshold, because the gate is pass/fail and most contractors aren’t in the room. Keep velocity steady, because 25 recent reviews beat 60 stale ones and burst campaigns create exactly the profile that reads as inactive. Respond to what comes in, because answered reviews signal active management. And spread the work across more than one platform, since third-party review site authority ranks among the top factors for AI visibility.
Every one of those is a capture and consistency problem, which is to say a mechanics problem. Mechanics problems are solvable. You already do the work that earns the reviews. The bar that’s changed is what it takes for the systems deciding who gets recommended to see it.
Want to know where you stand against your market’s threshold? Free 30-day pilot, up to 5 locations, no commitment to start.
