Anyone searching for Anna Mae’s Bakery and Restaurant, a fixture in Millbank, Ontario, was met with a Google AI Overview announcing that the business would shut for good on July 31. The bakery has no such plans. Owner Amanda Herrfort spent the following week fielding calls, emails and walk-ins from customers who trusted the search result over her own storefront.
The erroneous summary appeared over the weekend of July 18 and 19. CBC News broke the story on July 23, and PPC Land’s account, published July 26, is the version in this newsletter’s window. The gap matters: a factual error can sit live in Google’s answer box for days before anyone outside the business notices.
Herrfort learned about the claim secondhand. She was in the car on her way to a wedding when her phone lit up with someone’s screenshot of what Google was saying. “That’s when I panicked,” she said, per CBC News. The likely cause, per CBC’s reporting, was entity confusion: there is a bakery in Illinois trading under the same name, that one genuinely is shutting down, and Google appears to have folded its record into the Ontario restaurant’s.
The confusion did not stay confined to a results page. “We have a lot of people coming in talking to our employees saying: ‘It says that you are [closing].’ And then the employees would say: ‘Well, no, we’re not,’” Herrfort said, per CBC News. For a seasonal restaurant, a fabricated closing date during peak summer travel is not cosmetic.
The more durable part of the story is what happened next. No dedicated channel existed for Herrfort to report a factual error in an AI Overview: no ticket system, no escalation form, no listed contact. She ended up typing her complaint into Google’s AI search bar. “I said: ‘I would like this fixed please.’ And then it wrote back saying: ‘Oh yes, I can see how this could be detrimental, and we will get this fixed as soon as we can and I will make sure the correct information is given going forward.’ It’s like, OK, I hope this works,” Herrfort said, per CBC News. That exchange produced an apology and a promise, not a ticket number, a service-level commitment or any way to confirm the correction happened.
A Google spokesperson told CBC News that “the vast majority of AI overviews are high quality and meet our high bar for helpfulness and accuracy,” adding that “issues can arise when there is an absence of high quality information on the web on a particular topic.” That framing places the cause on the open web, not on Google’s summarization layer, and measures the error against aggregate volume rather than the one business it named.
Christine Logel, a professor of social development studies at the University of Waterloo, offered a different read to CBC News. She said systems like AI Overviews “were released before being fully tested,” and that “means things like this are going to happen.” Once such tools are “integrated so much through the internet and in so many different products,” she added, “when they make mistakes, those have real consequences for real people.”
For local SEO teams, the mechanism matters more than the anecdote. Structured-data hygiene cannot stop an AI Overview from importing facts about a different entity that shares a name. Exposure concentrates wherever a name is not unique: franchise locations, chains operating across regions, and any trade name generic enough to be reused. The more candidates Google’s entity resolution has to choose between, the more chances it has to pick wrong. None of those businesses get a notification when a generated summary describes them wrongly, so the monitoring burden falls entirely on the operator.
The domain-level opt-out Google introduced this year does not close that gap. It clears a site’s own content from AI Overviews and AI Mode within one to two days, but it says nothing about statements Google generates about a business rather than from it. A restaurant that opts out can still have an AI Overview announce its closure.
The legal ground under these errors is shifting elsewhere. In May, the Munich Regional Court I issued an injunction against Google over false AI Overview claims about two German publishers, backed by fines up to 250,000 euros per violation. No comparable remedy exists for a village bakery in Ontario, where the correction ran entirely through a chat window.
Local and multi-location teams should treat this as a monitoring problem, not a schema problem. Run branded-query checks for every location name and its close variants on a recurring schedule, and keep a public-channel response ready on the website and social accounts, since a formal path to escalate a wrong AI Overview still does not exist.
Per PPC Land, reporting by Luis Rijo, published July 26, 2026.