RPRED PALM

Home / Blog

When an AI assistant gets your business hours wrong

A customer asks ChatGPT or Gemini if you're open, and it says no. You can't edit the answer, but you can fix what it's reading. A step-by-step field guide.

2 min read#AI Search

More customers now ask an assistant instead of scrolling search results: “Is there a plumber open now near me?”, “What time does the vet on Main Street close?” When the answer is wrong about you (closed when you’re open, an old number, a service you stopped offering), the customer doesn’t argue. They call someone else.

You can’t log in and edit what an assistant says. But assistants don’t know your business personally; they assemble answers from public sources. Fix the sources and the answers follow.

Step 1: Make sure it’s really wrong

Answers vary from one try to the next, and by how the question is worded. Before you chase anything:

  • Ask the same question a few times, in a couple of different wordings.
  • Write down the exact question, the exact wrong claim, and the date.
  • Check your own record first. Is your website right? Is your Google Business Profile right, including holiday hours?

If the error shows up repeatedly and your own record is correct, it’s worth fixing.

Step 2: Find where it came from

Many assistants show the sources behind an answer. Start there. If they don’t, search the web for the wrong fact itself: your old phone number, the wrong hours in quotes, a service you dropped. Whatever pages still carry it are your suspects. The usual culprits:

  • An old directory or listing site you forgot existed.
  • A page on your own site that was never updated, like an old contact page or a PDF menu.
  • A third-party article, review site or forum thread.

Step 3: Fix it at the source

Update your own site and profile first, because they’re the sources you control and the ones assistants tend to trust. Then work outward: correct the listings you can edit, and contact the owners of pages you can’t. Put the facts in plain text on your site (hours, phone, service area) rather than only in an image or a widget, so they’re easy for any system to read.

Step 4: Report it, then re-check

Most assistants have a way to flag a bad answer, usually a thumbs-down or feedback button. Use it, then ask the same question again over the next few weeks. Assistants that search the web while answering can pick up a fix fairly quickly; facts absorbed during training can linger longer. Keep your notes from step 1 so you can tell when it’s actually fixed.

Don’t cry wolf

The worst outcome is spending an afternoon arguing with a listing that was already correct. One odd answer is an anecdote. An error that shows up again and again, checked against a record you know is right, is a finding.

How the engine handles it

The Decision Engine asks assistants the questions your customers ask on a schedule, and checks every claim about you against your own record, mechanically and without guessing. When it isn’t sure, it says “unverifiable” instead of raising an alarm. Real discrepancies arrive with the question, the quote and your record side by side, and corrections are made at the source with your approval. See AI Search Visibility, or read the full AI Search (GEO) Guide.

Green light. The race is on.

First snapshot the same day you sign up — actions in your decision log within 72 hours, your first report that Friday.

starting at $499/mo · month to month · cancel from your dashboard