Your reviews decide whether ChatGPT names your business at all

What an AI assistant does with your reviews first

An AI assistant reads your reviews before it decides whether to say your name. Two numbers carry most of that decision: your star average, and the share of your reviews you have answered. Miss on either one and the assistant leaves you off the list, however close you are to the person asking.

That's the finding in SOCi's 2026 Local Visibility Index, which pulled performance data from nearly 350,000 business locations across 2,751 multi-location brands. Search Engine Land covered the report in January. SOCi says AI recommendations "consistently favor businesses with above-average sentiment," which in plain terms means reviews that read as mostly positive.

Sentiment works as a filter, and a filter runs before any sorting does. Come up short on it and nothing else about your business gets weighed, because you're already outside the group the assistant is choosing from.

A customer asks who to call about a water heater leaking on a Sunday night. Back comes a short paragraph with three or four names in it, a line of description each, and yours is missing. There's no second page to click through to. That paragraph is the whole result, and the customer picks a name out of it and dials.

One number is worth holding onto for scale. Brands in the study turned up in Google's local 3-pack 35.9% of the time. That's the three-business map box above the regular blue links, and it runs mostly on how close you are and whether your category matches. On the AI side, SOCi found that middling-rated locations are frequently excluded from the answer altogether. These systems would rather skip a business than risk putting a bad one in front of a customer.

Where the bar actually sits

The study puts numbers on it, and they differ by assistant. Locations that ChatGPT recommends average 4.3 stars. On Perplexity the average is 4.1, and on Gemini 3.9.

Those are averages of what each assistant actually picked, so read them as the neighborhood you want to be sitting in. If your rating is in the high 3s, you're already under what ChatGPT and Perplexity pick on average. Whatever else is true about your business, you're outside the group those two are choosing from.

The sharpest version of this showed up in one industry. Among the financial brands in the study, the ones sitting near 3.4 stars with review response rates under 5% were effectively invisible in AI recommendations. SOCi's own summary is that weak fundamentals translated directly into zero AI visibility. Those same financial brands also scored low on how accurate their listing information was.

That's financial services, one slice of the data, and nobody has published the same cut for dental or HVAC. What it gives you is a floor you can see. Among those financial brands, two weak review numbers together were enough to erase them from AI answers.

Worth being straight about who was measured here. The study looked at multi-location brands, chains with a marketing budget and staff pointed at exactly this. If those are the businesses coming up short on replies, a single-location shop with nobody assigned to it is unlikely to be doing better. The two numbers are the same two numbers either way.

The half most businesses are short on

The reply number is where most businesses are quietly failing. A SOCi benchmark published on August 12, 2026, drawing on the same Local Visibility Index data, found the average multi-location business answers 46.9% of its Google reviews and just 3.1% of its Yelp reviews.

Out of 100 Yelp reviews, about 3 get a reply from the owner. Half the Google ones go unanswered too, and most owners have never counted either number, because neither one sits anywhere you'd trip over it.

Replying is the cheaper half to fix. A star average is the weight of every rating you've ever collected, so it moves slowly even when the new ones are good. The reply rate is entirely yours to move. Two hours with a phone and a coffee clears a long backlog for most single-location businesses.

The number of customers doing the asking has moved fast. BrightLocal's 2026 consumer survey, a panel of 1,002 US adults, found 45% now use AI to find local business recommendations, up from 6% a year earlier.

People set their own bar long before an assistant gets involved. BrightLocal reported that 47% of consumers would not use a business with fewer than 20 reviews, and 31% would only use one rated 4.5 stars or higher. So the two numbers on your profile are being read twice on the way to a booking, once by software and once by the person it hands the shortlist to.

The five-minute check

I ask owners to do this on their phone before the first job of the day. Everything in it is on pages you already control.

1. Open your Google Business Profile and write down your star average.
2. Scroll your last 20 Google reviews and count how many carry a reply from you. Twelve out of 20 is 60%, and that's close enough for this purpose.
3. Do the same two things on your Yelp page.
4. Put your star average next to 4.3, 4.1 and 3.9, the averages ChatGPT, Perplexity and Gemini recommend at.
5. If either of your numbers is short, start with the replies, oldest first.

For a dental practice or a medspa, a reply is also a free place to put the words a customer would type. A patient who mentions the crown was finished in one visit gives you room to write back that same-day crowns are something you do every week. An HVAC contractor answering a review about a Sunday callout can say plainly that weekend emergency service is available and what the after-hours rate is. That text sits on a page the assistants read, in the language customers use.

Working through a backlog of replies goes faster than most owners expect. A lot of it is one-line thank-yous that take a few seconds each. Slow down on the complaints, because those are the ones a customer reads word for word. Name what went wrong in plain language and say what you've changed since. Then stop. A reply that argues back sits on the page doing more damage than the review it answers.

One caution on the star half. Asking only your happy customers for reviews is against the rules on the major review platforms, and worth reading up on before you build any request process. Ask everyone and answer everyone, then let the average settle where it settles.

Clearing this bar gets you into the group the assistant is willing to pick from. Everything else you do on your website and your listings competes inside that group. Plenty of businesses in every category are sitting outside it without knowing, and the two numbers that decide it are on pages you already own.