The Overland Park Farmers Market did not hide. It did not run a secret pop-up in a cornfield. It had an official website, official social accounts, a published temporary location, and an opening schedule. Still, on April 18, 2026, more than 100 people reportedly ended up at a construction site instead of the market.

The culprit was not a bad flyer or an uncle with outdated directions. It was AI-generated search and social summaries pulling the wrong pieces of information together during a location transition.

The market was in a confusing but manageable year. Its longtime downtown site was being rebuilt into Clock Tower Landing, a new permanent home scheduled to open June 6. During April and May, the market operated at Matt Ross Community Center, a couple of blocks away. The future home existed, but it was not open yet. Humans can understand that distinction. "Future home" and "current location" are different concepts. Search AI apparently saw those words, nodded gravely, and sent shoppers to a fence.

What happened

According to the Kansas City Star, the Overland Park Farmers Market said more than 100 people went to the construction site at Clock Tower Landing on Saturday, April 18. That site was the future home of the market, but at the time it was still under construction. The temporary market was at Matt Ross Community Center, 8101 Marty Street.

A syndicated report quoted Overland Park Strategic Communications Director Meg Ralph saying the city received social-media messages about the problem and, by combining online questions with confusion at the market, determined that more than 100 people had been affected. Ralph said the city often has to correct outdated information online because the market is popular, and that she reached out to Meta to try to get the false information corrected.

The market's own social post was blunt: AI summaries are not always correct. It said more than 100 people had ended up at a construction site because of incorrect AI search results. The post showed examples where Meta AI returned the wrong address, Google AI summaries used the wrong image or location, and search results pulled stale season or vendor details into the current answer.

The market told customers to verify dates, times, and locations through the official city website and official market social media. That is good advice. It is also a little bleak that the public service announcement had to be "please do not let autocomplete decide where your strawberries are."

Why location is harder than it looks

This incident is a clean example of why AI summaries are risky for local business information. Local search is full of semi-current facts. A venue can have an old address, a temporary address, a future address, a construction address, a mailing address, and a promotional page celebrating the future address. All of those facts can be true in isolation.

The hard part is temporal context. On April 18, 2026, Clock Tower Landing was the future home of the market. On June 6, 2026, it was scheduled to become the active home. During April and May, the active location was the community center. A normal search result page can show multiple links, dates, and snippets. A user might click the official market page and read the current schedule. An AI summary compresses the messy web into one answer, which means it has to choose the right time-specific fact.

That is exactly where these systems still stumble. They often flatten "will be," "used to be," and "is" into one confident location. The result sounds helpful because it is short. The harm comes from the same shortness. The context that would have warned a human that a location was not open yet gets shaved off.

For a farmers market, the damage is not catastrophic. Nobody is alleging a legal judgment or a data breach. But the harm is concrete: shoppers lost time, vendors likely lost foot traffic, staff had to answer confused questions, and the city had to spend communications energy correcting machines it did not control. If more than 100 people showed up at the wrong place, this was not one quirky screenshot. It was an operational nuisance with real-world consequences.

The brand damage lands on the wrong people

One of the more irritating parts of AI summary failures is that the blame lands locally. A customer who drives to the wrong place is not angry at a retrieval model. They are annoyed at the market, the city, or the platform where they found the information. The farmer selling produce did not design the AI summary. The city communications team did not ask Meta or Google to blend outdated and future-location data into a misleading answer. They still had to clean up the mess.

This is the same pattern that shows up in AI-generated restaurant specials, bogus hours, fake policy summaries, and wrong customer-service answers. A large platform gets the convenience and engagement of answering directly. The business or public agency gets the angry person at the counter.

In Overland Park, the confusion was made worse by the fact that the wrong location was not random. It was plausibly related to the market. The construction site really was the market's future home. That makes the AI error harder for customers to detect. If a chatbot had sent people to a shoe store in Nebraska, users might have questioned it. Sending them to the shiny new market project sounds authoritative enough to waste a Saturday morning.

Official sources still matter

The practical fix is boring and obvious: for location, hours, transit changes, closures, election information, school schedules, and anything else where time matters, AI summaries need to privilege current official sources and show the date of the information they are using. If the system cannot reconcile "temporary location until June" with "future home opening in June," it should not collapse those facts into a single confident answer.

Local organizations also need to write for machines now, which is a sentence that should make everyone tired. Pages need explicit language like "current location," "future location," "effective date," and "not open until." Structured data helps. Repeated social posts help. But none of that should absolve platforms from the basic duty not to synthesize stale snippets into wrong directions.

The Overland Park incident belongs here because the harm is small enough to be mundane and large enough to be real. This is how AI search failures will often look in practice: not a robot revolution, just 100 people standing outside construction fencing because a summary box could not handle a calendar.

Search used to hand users a map and a list of sources. AI search increasingly hands users an answer. When the answer is wrong, the user does not merely learn a bad fact. They move their body through the world based on it. In this case, that meant shoppers went looking for flowers and produce and got a construction site instead. Very farm-to-table, if the table is plywood and the farm is a zoning update.