Skip to content

When AI gets it wrong

AI is saying the wrong thing about my business. What can I do?

Old opening hours, a branch that closed, a service you stopped offering: here's where wrong AI answers usually come from, and the practical way to correct them.

Why AI assistants get business details wrong

AI assistants don't make up your opening hours on purpose. They put an answer together from what they learnt in training and, often, from web pages they look up at the time. If those sources are old, inconsistent or about a different business with a similar name, the answer can be wrong in a confident tone.

The usual causes are ordinary:

  • an old directory listing with a previous phone number or address;
  • a branch that closed but still has a listing or a page;
  • a previous website still online, or old pages on the current one;
  • a services list that was never updated after you stopped offering something;
  • another business with a similar name in a nearby town.

How to correct it, step by step

  1. Write down exactly what was said. Copy the question, the assistant, the date and the wrong statement word for word, plus any sources or links shown with the answer.
  2. Find where it probably came from. Click the sources if there are any. Otherwise search for the wrong detail: old directory listings, a previous website, social media pages, old articles or a Google Business Profile that hasn't been updated.
  3. Correct it at the source. Update or remove the out-of-date listing or page. If you don't control it, use that site's own process to request a correction.
  4. Say the right thing clearly on your own website. Put the correct detail in plain text on the relevant page, so there's an up-to-date, first-hand source that agrees with your listings.
  5. Check again after a few weeks. Ask the same question in a clean chat. Assistants don't all re-read the web at the same pace, so corrections can take time to show.

See where you stand first. It's free.

We ask ChatGPT and Google's Gemini the kind of questions your customers ask, then email you whether you were named, roughly where, and who was suggested instead. A person reviews your check before we send it, usually within 2 working days.

Get my free check

What you can't control

Nobody outside the companies that make these assistants controls what they say, and anyone promising to "remove" a wrong answer overnight is overselling. What you can do is make the right information easy to find and the wrong information hard to find. That's usually enough over time, but it isn't instant and it isn't guaranteed.

If the wrong detail is harmful (for example, it says you've closed when you haven't), it's worth checking more often and fixing the most-visited sources first: your website, your Google Business Profile and the biggest directories in your sector.

Not sure what the assistants are saying?

The free check asks ChatGPT and Google's Gemini the kind of questions your customers ask and flags anything they said about you that looks wrong, alongside whether you were named and who was named instead. If you'd like the fuller picture, with a fix list in priority order, that's the AI visibility audit.

Questions people ask

Can't I just tell ChatGPT it's wrong?

You can, and in your own chat it may apologise and correct itself. But that doesn't change what it tells other people. Most assistants have a feedback button on each answer; using it is reasonable, but it isn't a reliable fix on its own. The lasting fix is correcting the information the assistant reads.

How long until the wrong answer goes away?

There's no fixed timetable. It depends on how often the assistant's sources are re-read and how widely the wrong detail is repeated. Our short answer on how long it takes to show up in AI answers explains what affects it.

The wrong information is about a business with a similar name. What can I do?

Make the differences obvious on your own website and listings: your full trading name, your town, what you do and your contact details. Consistent, specific details make it easier for an assistant to tell two businesses apart.

Will the free check find wrong information?

It flags anything ChatGPT or Google's Gemini said about your business in our questions that looks wrong. It isn't an exhaustive audit of everything every assistant says about you, and it doesn't cover Perplexity.

See what the assistants say about you.

Get my free check