Guide

How often do AI assistants update information about your business?

7 min read

An AI assistant told a customer your shop closes at 5pm. You close at 7pm. Now you are wondering how long you have to wait for it to correct itself, and whether there is anything you can do. The frustrating part is that AI assistants state wrong facts with the same confidence as right ones, so a customer has no reason to doubt it. There is no single update button and no fixed schedule anyone can promise you. But the way these tools get their answers is not a mystery, and once you understand it, you can see exactly which levers you control and which you do not.

There are two ways an AI assistant knows things about you

When an AI assistant answers a question about your business, it is doing one of two things, and the difference is the whole story.

The first way is its built-in model knowledge. The model was trained on a large snapshot of text up to a certain point, called a training cutoff. Anything it learned then is frozen until the company retrains or updates the model. That happens occasionally, on the company's schedule, and you have no control over it or visibility into when it lands.

The second way is live browsing. Many of the major assistants now search the live web when they answer a local question, then read what they find and summarize it. This path is not frozen. It is only as current as the pages it reads. If your Google Business Profile says you close at 7pm, an assistant that browses can pick that up quickly.

So the same question can get a stale answer one day and a correct one the next, depending on which path the assistant used.

Why you sometimes see old information

When you see a confidently wrong answer, it usually means the assistant answered from its frozen model knowledge instead of looking anything up. The fact it gave you was true at some point in the past, or it was pulled from an old page the model saw during training, and nothing has corrected it inside the model since.

This is why a brand new model can still tell people your old hours, your old address, or a phone number you stopped using two years ago. The model is not lying. It is repeating what it was taught, and it has no idea the world moved on.

Why you sometimes see current information

When the answer is right and up to date, the assistant most likely browsed. It searched, landed on your Google Business Profile or your website or a directory listing, read the current details, and repeated them.

This is the good news hiding in the problem. The browsing path reads public sources you can edit. You cannot reach into the model and change its memory, but you can change what the live web says about you, and the browsing path will read your corrections the next time it looks.

What you can actually control

You cannot force a model to retrain, and you cannot make a specific assistant browse on demand. Those levers belong to the AI companies.

What you do control is every public source the browsing path reads. That is the real work, and it is the same work that helps you in normal Google search too.

Fix the facts at the source first, then make sure nothing else out there contradicts them.

  • Correct your Google Business Profile. For most local questions this is the single most important source, so make the name, address, phone, hours, and category exactly right.
  • Update your own website so the same details appear there in plain text, ideally on a page that is easy to find.
  • Fix or claim your listings on the big directories such as Yelp, Apple Maps, Bing Places, and any industry-specific sites that cover your trade.
  • Remove or correct old listings that still show a former address, a closed location, or a disconnected number.

Why consistency across sources matters so much

Getting one source right is not enough if three others still disagree. When an assistant browses and finds conflicting facts, two from your old listings and one from your updated profile, it has to guess. It may hedge, it may average them in a way that makes no sense, or it may simply pick the wrong one.

When every public source says the same thing, there is nothing to guess about. The assistant reads a clear, consistent answer and repeats it. Consistency is not a nice-to-have. It is what removes the assistant's room to be wrong.

This is also why the cleanup step matters as much as the update step. A stray old listing with a wrong number can keep poisoning answers long after you have fixed your main profile.

How long a correction takes

Honest answer: nobody can give you a guaranteed date, because it depends on which path the assistant uses and how often it re-reads your sources.

For the browsing path, corrections can show up reasonably soon after your public sources are updated and consistent, because the assistant reads them fresh when it looks. For the frozen model path, the only thing that changes the answer is a future model update, which is out of your hands and unannounced.

The practical move is to fix everything at the source, get your facts consistent, and then re-check after a few weeks rather than the same afternoon. Give the browsing path time to re-read your pages, and give the search engines time to recrawl them.

How to check where you stand right now

Before you can fix anything, it helps to know what the AI assistants are actually saying about you today, and whether they are answering from memory or from your live profile.

You can do this by hand. Open ChatGPT, Gemini, and Perplexity, ask the kinds of questions a real customer in your area would ask, and write down what each one says, including anything that is wrong. Ask the same question more than once, because the answer can vary from one try to the next.

LocalFox runs that check for you. Its free check shows you what the major AI assistants currently say about your business and where they get it wrong, so you can see whether your facts are landing or whether you have cleanup to do before you spend time on it.

See what AI says about your business

Run the free check. Your name and city, no account, an answer in about two minutes.

Run the free check

Questions

Why does ChatGPT have old information about my business?+

Most likely it answered from its built-in model knowledge instead of browsing the live web. That knowledge is frozen at a training cutoff and only changes when the model is updated, which happens occasionally and is out of your control. The fix you can act on is to make sure your current details are correct and consistent across your Google Business Profile, your website, and the major directories, so that when the assistant does browse, it reads the right facts.

How do I get AI to update my hours?+

You cannot edit the AI directly. What you can edit are the sources it reads. Update your hours on your Google Business Profile first, then match them on your website and on directory listings like Yelp, Apple Maps, and Bing Places. Remove any old listing that still shows the wrong hours. When the assistant browses next, it will read the corrected hours from those sources.

How long does it take for AI to show my new information?+

There is no fixed timeline anyone can promise. If the assistant is browsing the live web, it can pick up your changes reasonably soon after your public sources are updated and consistent, since it reads them fresh each time. If it is answering from its frozen model knowledge, the information will not change until the model itself is updated, which is occasional and not announced. Update your sources, make them consistent, and re-check after a few weeks.

Will correcting my Google Business Profile fix what AI says about me?+

It is the most important single step, because Google Business Profile is one of the most common sources assistants read for local questions. But it works best when your other sources agree with it. If old directory listings still show different hours or a different address, an assistant that browses can still find the conflicting facts and get confused. Fix the profile, then make everything else match.

Why does the AI keep getting my address or phone number wrong even after I fixed it?+

Usually one of two reasons. Either the assistant is still answering from its frozen model memory rather than browsing, in which case it will not update until the model does, or there is still an old listing somewhere with the wrong details that the browsing path keeps finding. Track down stray listings on directories and old pages, correct or remove them, and the conflicting answers tend to fade once your sources line up.

Can I force ChatGPT or Gemini to recheck my business?+

No. You cannot trigger a model retrain or make an assistant browse on demand. Those are controlled by the AI companies. What you can do is control the public sources the browsing path reads, keep them accurate and consistent, and let the corrections get picked up over time. Re-checking after a few weeks is the realistic way to confirm it worked.

Keep reading