Search for a company's name plus the word "lawsuit" and you're already a little suspicious. You want to know if there's fire behind the smoke. In late 2024, anyone who searched "Wolf River Electric lawsuit" got a crisp, confident answer at the very top of the page: the Isanti, Minnesota solar installer was "currently facing a lawsuit from the Minnesota Attorney General due to allegations of deceptive sales practices," including misleading customers about cost savings, using high-pressure tactics, and tricking homeowners into binding contracts with hidden fees. It read like settled fact. It even cited sources.

There was no such lawsuit. Wolf River Electric has never been sued by the Minnesota Attorney General.

A real case, attached to the wrong company

Here's the part that makes this a textbook AI failure rather than a simple typo. There genuinely was a Minnesota attorney general action against solar companies. In 2024, Attorney General Keith Ellison sued four solar-lending companies, accusing them of concealing and improperly charging roughly $35 million in fees to Minnesotans. Real case, real defendants, real money. Wolf River was simply not one of them.

According to the complaint, when a user searched for Wolf River, Google's AI Overview cited four sources to support its claim that the company was a defendant. One was a Star Tribune article about Ellison's case against those four other companies; it mentioned Wolf River only in passing and did not say the company was being sued. The others reportedly included an Angie's List page and a news release from the attorney general's own office. None of them said what the AI claimed they said.

So the model did what these systems do when they're rewarded for sounding complete: it generalized. It had "Minnesota," "attorney general," "solar," "deceptive sales practices," and a company name sitting near a story about a lawsuit, and it fused them into a confident accusation that no human source had actually written. This is the difference between summarizing and inventing. A summary points back at something. The AI Overview pointed at four pages that, when you actually read them, contradicted the summary sitting on top of them.

Wolf River's complaint says employees discovered the false statements on or around September 2, 2024. The company alleges Google's autocomplete suggestions nudged people toward the same narrative, so the answer box and the search box reinforced each other. Once that machinery latches onto a false association, it doesn't quietly correct itself; it keeps serving the same confident answer to every new person who asks.

What it actually cost

Reputational harm is easy to wave away as abstract. This wasn't abstract. The complaint lays out a sequence of lost business in early 2025, right when prospective customers were doing exactly the due-diligence search that surfaced the false claim:

• On March 3, 2025, a customer canceled a solar contract worth nearly $40,000, citing lawsuits they'd found on Google.
• On March 5, a customer sent a Wolf River salesperson a screenshot of Google's AI output as the reason they were walking away from a $150,000 deal.
• On March 11, a nonprofit terminated a solar project and a lighting project together worth more than $174,000, referencing "several lawsuits in the last year with the Attorney General's Office."

There's a particularly grim twist buried in the fallout: the attorney general's office reportedly fielded complaints from Wolf River customers who had followed the AI's own suggestion to file a complaint against a company that had never been accused of anything. The hallucination didn't just sit there; it generated downstream actions against the victim.

For a regional installer that lives on local trust and word of mouth, an accusation of defrauding homeowners, displayed in Google's voice above the search results, is roughly the worst thing the most-used information source on earth could say about you. Wolf River and four of its officers filed suit, and the company has described its damages in the range of $110 million to $210 million. (The formal pleading uses Minnesota's convention of claiming damages "in excess of $50,000," which is a procedural floor, not the number they're chasing.)

"We didn't write that," says the company whose product wrote that

Google's defense is the interesting part, and it's why this case matters beyond one solar company. The company denies the allegations and has leaned on a familiar argument: it's a platform surfacing third-party content, the AI summarized what's on the web, and therefore Google isn't the author or publisher of the statement. In its filings, Google has characterized the AI Overview as describing documents and publications "the contents of which speak for themselves." Translation: blame the web, not us.

That framing runs straight into the awkward fact at the center of the case. The cited sources didn't say Wolf River was sued. There's no third-party publisher to point at, because no third party published the accusation. The AI assembled it. When Wolf River filed in Ramsey County District Court, Google moved the case to federal court, where Section 230 defenses (the law that shields platforms from liability for content created by others) tend to get a warmer reception. In January 2026, a federal judge sent the case back to Minnesota state court, ruling that Google's removal was untimely because it had notice of the case's value back in March 2025. That's a procedural loss for Google, not a ruling on whether AI Overviews can defame anyone.

William McGeveran, dean of the University of Minnesota Law School, summed up the hard question for Wolf River in one line: "The question in this case is, 'Who wrote the AI Overview?' Under current law, I think the answer, pretty clearly, is: 'Not Google.'" That's the crux. Defamation law was built around a human publisher who chooses to publish a specific statement. A model that generates novel text for each query doesn't fit that mold cleanly, and Section 230 was written for an internet of links and user posts, not synthesized answers.

Why this belongs in the graveyard

We already have the European bookend to this story. In June 2026, a Munich court treated Google's AI Overviews as Google's own statements and held the company could be directly liable when those summaries invent accusations about a business. Wolf River is the American version of the same collision, working its way through US courts with US defamation law and Section 230 in the mix. Same product, same failure mode, two legal systems poking at the same question: when the answer box makes something up about an identifiable company, who owns the lie?

The mechanism is what should worry every small business, not just solar installers. Traditional bad search results send you to a sketchy page you can evaluate. A bad AI Overview puts the false claim in the most authoritative real estate on the internet, in the platform's own voice, phrased as a conclusion, before anyone has a chance to object. "Grounding" the model in cited sources is supposed to prevent this, and it plainly didn't; the system cited four pages and then said something none of them supported. Retrieval is not the same thing as truth, and a footnote is not the same thing as accuracy.

Google's fallback, that AI Overviews are experimental and users should verify, gets the responsibility exactly backwards. The party that publishes a negative factual claim about a named business is the party that should verify it before showing it to millions of people, especially when the claim is "this company is being sued by the state for fraud." Telling the victim to go correct the record after the contracts are already canceled is not a remedy. It's an apology written in someone else's bankruptcy.

What's decided, and what isn't

To be precise about the status: this is a live lawsuit. As of mid-2026 the case is in its pre-trial phase in Minnesota state court after the remand. Nothing has been adjudicated on the merits. Google has denied the allegations, no court has found it liable, and the damages figures are Wolf River's claims, not awards. The facts that are not in dispute are narrow but damning: the Minnesota Attorney General did not sue Wolf River Electric, and Google's AI Overview repeatedly told the public that it did.

Whatever the court ultimately decides about authorship and Section 230, the engineering lesson is already on the table. If you ship a system that manufactures confident factual claims about real, named entities and displays them as answers, you've built a defamation machine with great UX. The only open question is who pays when it's wrong.