AI Answers Are Getting Businesses Wrong in Both Directions
Local Falcon analyzed 10,000 US restaurants across all 50 states in May 2026 and found that 74.9% never appear in Google’s AI Overviews, against 19.6% invisible on Google Maps. Last week in Willoughby, Ohio, a bookshop ran into the opposite problem: Google’s AI summary said it was permanently closed. It isn’t. Put those two together and you get a decent picture of how AI answers currently handle small businesses.
What happened
The Local Falcon study ran in May and resurfaced this week in coffee trade press. It compared visibility on Google Maps against visibility in AI Overviews, and the gap is the whole story. Three-quarters of the venues Maps lists happily never show up in the AI answer.
Reputation rescued nobody. Restaurants with over 1,000 Google reviews were still invisible in AI 70.9% of the time, which is almost exactly the rate for places with far fewer. Star ratings barely mattered either. The best band was 4.5 to 4.7 stars, and near-perfect ratings did worse than that.
Brand recognition didn’t help at all. McDonald’s appeared in zero AI recommendations despite 96% visibility on Maps. Starbucks turned up in under 5% of searches.
Coffee came off worst. Of 926 coffee shops, cafés, and roasteries in the dataset, 77.2% didn’t appear in Google’s AI answers, against 13.2% missing from Maps. Chains were left out 95.7% of the time, independents 68.2%.
“Coffee has more to lose here than any category we looked at,” Local Falcon CEO David Hunter said. “These are businesses that turn up almost every time on Google Maps, that have spent years building their ratings, and then a customer asks AI where to get a good espresso and three in four of them aren’t in the answer.”
The other failure mode is happening to Chanel Bush, who co-owns the Viridian Bookshop at 4081 Erie Street in Willoughby, Ohio. The shop opened less than a year ago. A customer told her that Google’s AI summary described it as permanently closed, at 4041 Erie Street, a different address a few doors down. Bush had already noticed traffic dropping without knowing why.
It doesn’t show up consistently, which is what makes it so hard to fight. Some searchers get the false summary, others get the truth. Bush reported it and asked customers to report it too, thought it was fixed, then watched it come back. News 5 Cleveland asked Google for comment and got nothing before publication.
“Don’t always trust the first thing that AI pops up,” Bush told the station.
Why it matters
Both stories run on the same mechanic. AI answers aren’t ranking your business the way search results did. They’re assembling a claim about it, open or closed, in stock or not, worth recommending or not, from whatever machine-readable signals they can find. The trust metrics that took a decade to build aren’t part of that calculation.
Willoughby is the sharper version because it shows what happens when the assembled claim is simply false. Nobody clicked through and decided not to buy. The AI answered on the shop’s behalf and the customer never arrived.
What this means for Shopify merchants
The study looked at restaurants, not online stores, so treat the numbers as a signal rather than your own scoreboard. What transfers is the mechanic. You don’t have a storefront on Erie Street, but you do have an ecommerce version of “permanently closed,” and it’s the availability data attached to your products.
When an AI agent or an AI Overview answers “where can I buy X,” it’s checking whether X is actually purchasable. That check runs against structured data: your product schema, your Merchant Center feed, the catalog Shopify now exposes to agents. It does not run against your brand, your reviews, or how well your product page reads. The Local Falcon numbers are the evidence for that. A thousand reviews bought these businesses essentially nothing.
So keep the availability value in your product schema current. A restocked item still marked OutOfStock is the same failure Bush is dealing with: you exist, you’re sellable, and the answer layer has no idea. Don’t delete or 404 sold-out product pages either. A dead URL takes you out of the index entirely, and getting back in is slower than staying in. Keep the page, mark it accurately, restore it on restock.
Then there’s the shopper who does make it to your site. Pushy covers that end of it, dropping out-of-stock products to the bottom of the collection when stock hits zero and returning them to position when you restock. The pages and URLs stay live and indexed, which is the state you want them in when a machine comes looking to confirm you’re open.
“These are businesses that turn up almost every time on Google Maps, that have spent years building their ratings, and then a customer asks AI where to get a good espresso and three in four of them aren’t in the answer.” — David Hunter, CEO, Local Falcon
The bottom line
A decade of reviews didn’t get three-quarters of these restaurants into an AI answer, and it didn’t stop one bookshop from being declared closed. What decides the answer is structural and fairly boring: whether the record is current, whether it agrees with your page, and whether it says you can be bought from today. Get that wrong and you probably won’t notice until a customer mentions it, which is how Chanel Bush found out.