AI Cites Old Information About My Business: Old Prices, Old Address, Old Owner

Last year's business, quoted with confidence

Outdated answers are harder to spot than wrong ones, because every detail was true once. The customer who shows up at the old storefront does not think the AI failed. They think you did. The usual reports sound like this:

• "ChatGPT told me a tune-up costs $89." That was your 2023 coupon price, and it still lives in an old blog post.
• A customer drives to the address you left a year ago, because an AI answer gave it as current.
• The location you closed in the spring is still listed as open, with hours, in an assistant's list of options.
• An AI answer names the previous owner as the person to call, or describes the company under its old name.
• Your old menu, old service list or old phone number appears, word for word, as if nothing changed.

If the answer contains facts that were never true, that is a different problem, covered in AI made up facts about my business. If the details are current but scrambled, try AI has wrong business information. This page is for the past that will not stay past.

Why the old version outvotes the new one

A fact gets repeated in proportion to how many places state it and how established those places are. Your old address may sit on 40 directory pages, a newspaper article, three supplier sites and a chamber of commerce list built up over a decade. Your new address sits on your website and one profile. To a system that weighs agreement across sources, the old version looks like the consensus.

Freshness helps, but less than people assume. Ahrefs analyzed about 17 million citations in 2025 and found URLs cited by AI assistants averaged 1,064 days old, against 1,432 for Google's organic results. ChatGPT showed the strongest pull toward newer content; Google's AI Overviews showed almost none. So the AI-cited web is a little newer than the search web, and still mostly years old. Our page on content freshness covers what that means for your publishing habits.

Where the stale facts are hiding

Work through these in order. The first rows are the ones you control outright, and they feed several engines at once.

The correction routine, in the order that pays off

1. 1

   Catch the stale fact in the act

   Ask ChatGPT, Gemini, Perplexity, Copilot and Google the question that produced the old detail. Save each answer, the date, and any sources it lists. Our free AI visibility check runs a question through several engines at once; its API answers can differ a little from the apps.

2. 2

   Search your own site for the old version

   Search Google for site:yourdomain.com plus the old price, street name or owner's surname. Old blog posts, PDFs and forgotten landing pages turn up more often than owners expect.

3. 3

   Fix the site, and change the substance, not the date

   Rewrite or retire the stale page. Google warns against changing a page's date to look fresh when the content has not substantially changed, so update the facts first, then the date. Redirect closed-location URLs to a current page.

4. 4

   Update the three map platforms

   Correct Google Business Profile, Apple Business and Bing Places the same day, with identical name, address and phone. Google reviews profile edits before they go live; it says changes usually show in about 10 minutes but can take up to 30 days.

5. 5

   Notify the crawlers

   Request indexing for changed pages in Search Console's URL Inspection tool, and use IndexNow, which Microsoft recommends for reporting updates to Bing and Copilot.

6. 6

   Work the long tail

   Update directories, supplier and partner pages, and associations. Ask publishers of old articles for a correction or a note. Entity optimization explains why consistency across these matters more than any one listing.

7. 7

   Report the answer itself

   Use thumbs down and Report a problem on Google's AI Overviews, thumbs down in Gemini and Claude, and the feedback controls in other apps. Reports are one input among many, but they cost nothing.

Moved, closed, sold or repriced: the special cases

How long an outdated answer lingers

Nobody can give you a date. Google's own Business Profile help puts edits anywhere from minutes to a month. Directories and data partners refresh on their own calendars. Answers drawn from a model's training memory change only with a new model. Expect the engines that search the live web to catch up first, and judge progress across a monthly set of questions, not one lucky answer.

The web never deletes your past. It just needs more recent evidence than old evidence.

A standing date with your own facts

Stale answers come back whenever a change goes out to some places and not others. Keep a single fact sheet of name, address, phone, hours, prices, owners and locations, and treat every change as a checklist across the site and every listing. Do it again each quarter. In Athena CMS, which comes with every site we build, the Hestia agent keeps hours and contact details matching everywhere, and Delphi asks the AI engines what they say about you and flags outdated prices or claims.

Call for help when the old information keeps resurfacing after you have corrected the obvious sources, or when it is costing you real visits to the wrong door. A free GEO audit traces where each engine is getting the old detail and returns a fix list with a fixed monthly quote; the cleanup is part of our SEO and AI visibility programs. For other problems, start at AI search fixes, or look at how Siri handles business hours if Apple devices are the ones getting it wrong.