When AI Gets You Wrong: How to Fix Bad Information About Your Business

Sooner or later it happens to almost every business. A customer mentions that ChatGPT said you don't take insurance, or Gemini listed hours you haven't kept since 2023, or an AI answer described a service you stopped offering years ago. It's unsettling, partly because it feels like there's no one to call. You can't email an AI and ask for a correction.

What you can do is fix the sources. Every wrong answer has an origin somewhere in the public record, and finding it is usually easier than owners expect.

Where wrong answers come from

AI engines don't invent facts about your business out of nowhere as often as people assume. Most errors trace back to something real: an old directory listing, a stale profile, an outdated page on your own site, a news mention from a previous era of your business. The engine read something that was once true, or half true, and repeated it.

The second-most-common cause is a gap. When the engine can't find a fact it needs, it sometimes fills the hole with a guess based on similar businesses. Vague sources invite invention. The cure for both problems is the same: make the correct information easy to find and hard to contradict.

Step one: find the source of the error

Start with the engine that got it wrong. Ask it directly: "where did you find that information?" Perplexity shows citations by default. ChatGPT and Gemini will often name their sources when asked, especially when they browsed the web for the answer.

Then search your own trail. Google your business name plus the wrong fact. The old phone number, the discontinued service, the wrong address. Nine times out of ten, the bad data is sitting on a page you forgot existed. An old listing, a previous website still partially live, a profile someone set up in 2019 and abandoned.

Step two: correct it where it lives

Fix the source, not the symptom. Claim the outdated listing and update it, or use the platform's process to remove it entirely. Most major directories have a claim-this-business flow, and review platforms have correction processes for factual details.

While you're at it, check the big surfaces that engines lean on hardest: your Google Business Profile, your top review platforms, and your own website. If any of them agree with the error rather than with reality, that's your priority. An engine that sees the wrong fact in two places and the right fact in one will often go with the majority.

Step three: outvote the leftovers

Some bad data can't be deleted. A years-old news article, a forum thread, an archive you don't control. You don't need it gone. You need it outnumbered. Publish the correct fact clearly on your own site, keep it identical across every profile you control, and let recency work for you. Engines weigh fresh, consistent sources over stale, lonely ones.

This is also why the fix isn't instant. Answers built on live search can update within days of the sources changing. Answers leaning on older training data take longer to come around, and they come around faster when the current web speaks with one voice.

Make it a habit, not a crisis

The businesses that get burned worst by AI errors are the ones that only look when a customer complains. By then the wrong answer has been served for months. A twenty-minute monthly check, asking a couple of engines about your business and skimming for errors, catches problems while they're small.

And keep notes. The date you fixed a listing, the answer that was wrong last month, whether it's right this month. Corrections in this world are less like flipping a switch and more like steering a large boat. The turn happens, but only if you hold the wheel steady, and only if you noticed you were off course in the first place.