AI gives wrong information about your business. Now what?

"According to ChatGPT you are still at the old location." "Gemini says you are open on Sundays." It happens more often than you think: an AI assistant gives information about your business that is not (or no longer) correct. Annoying, because customers assume it is true.

Why does this happen?

AI assistants rarely make things up out of nothing. Wrong information almost always comes from somewhere:

• Outdated listings in directories, on review sites or social media.
• Your own website, for example an old page or a footer that was never updated.
• Confusion with another business with a similar name.
• Old training data: what a model learned during training can be months or years old.

Sometimes a model also fills in missing information with an assumption. The less clear information there is online, the bigger that risk.

Step 1: find the source

Ask the same question to an assistant that shows sources, such as Perplexity or ChatGPT with search on. Look at which websites are listed under the answer. Also search yourself for your business name combined with the wrong information, for example your old address.

Step 2: correct it at the source

• Update your Google Business Profile and regularly check it for edits suggested by others.
• Update listings in directories and on review sites, or ask the owner for a correction.
• Go through your own website, including old pages, PDFs and the footer.
• Use structured data on your website with your correct name, address and opening hours.

Step 3: make the correct information stronger

It helps when the correct information appears the same in as many places as possible. Put your address, service area and opening hours clearly on your website, in your business profiles and in an llms.txt. If you are confused with another business, mention your town and full name explicitly.

Step 4: give feedback in the assistant

Most AI assistants have a thumbs down or feedback button on an answer. Use it and briefly explain what is wrong. It will not fix things immediately, but it helps the makers find errors.

How long does it take?

Answers based on live search results can improve within days to weeks once the sources are fixed. Information from a model's training only changes with a new model version. That is exactly why it matters that the correct information is easy to find online: a model is then more likely to rely on what it finds live.

Keep checking

A one-off check is not enough: errors can come back. With GEO tracking you read the answers ChatGPT, Gemini, Perplexity and Claude give about you every week, so you spot wrong information quickly.