GEO Pilot: we optimise your site for AI search in 30 days, $999.See what is included →

GEO

ChatGPT shows the wrong address, phone number or prices for my business: how to fix it

When an AI answer gets your contact details, services or prices wrong, the cause is almost always a wrong or old source somewhere on the web. This is a step by step way to trace the source, fix it and check the answer again.

Nitesh Dan CharanBy , Founder · 28 October 2026 · 6 min read
ChatGPT shows the wrong address, phone number or prices for my business: how to fix it
Key takeaways
  • Wrong details in an AI answer usually trace back to a wrong or old source: your own site, a directory, a review profile or an old article. Fix the source and the answer tends to follow, though no engine promises how fast.
  • Do not start by arguing with the answer. Start by recording the exact answer, finding where the wrong fact appears on the web, and correcting the pages you control first.
  • Use a single sheet of approved facts. Name, address, phone, hours, services and prices should be identical on your site and every profile.

Why does ChatGPT give the wrong details for my business?

An AI answer about your business is assembled from what the engine can retrieve and from what it learned in training. If it shows an old address, a closed branch, a wrong phone number or last year's price, there is almost always a source behind it: your own site, a directory listing, a review profile, a press article or a forum thread that still carries the old fact. Sometimes the engine also makes things up. We cannot always tell which, which is why the first job is to find out.

We cannot promise any particular correction time. OpenAI states that a robots.txt change can take about 24 hours to be picked up for ChatGPT search, but that is about crawler rules, not about corrections to an answer. Other engines publish no timelines at all. So the plan below fixes the causes you control and records the evidence so you can check again.

Step 1: Record the exact wrong answers

Before you change anything, save the evidence. Write down the exact prompt, the engine, the date and the full answer, and take a screenshot. Ask the same question five times, because answers vary, as we explain in why AI answers change between runs. Note how often the wrong fact appears. A fact that appears in four of five runs points to a persistent source. A fact that appears once may be noise.

FieldExample of what to record
Date and time28 October 2026, 10:15
Engine and modeChatGPT with search switched on, logged out
Exact promptWhat is the address of [your business] in [city]
Wrong factOld street address
Runs with the wrong fact4 of 5
Sources cited in the answerList every link shown

Step 2: Find where the wrong fact lives

Search the web for the wrong fact in quotes: the old phone number, the old street, the old price. Check these places in order.

  1. Your own website: contact page, footer, about page, old blog posts, PDF brochures, schema markup in the page source.
  2. Your Google Business Profile and any other map or business profile.
  3. Industry directories, chambers of commerce and trade association listings.
  4. Review sites and social profiles.
  5. Old press releases, news articles and partner pages.
  6. Forums and question sites where someone gave your old details.

Also look at the sources the engine cited in its answer. If the answer links to a directory with the old address, you have found a direct cause. If it cites nothing, the fact may come from training data, which you cannot edit. In that case the remedy is to strengthen the current, correct facts across many sources over time.

Step 3: Fix what you control, in this order

  1. Create a single sheet of approved facts: legal name, trading name, address, phone, email, opening hours, services, price ranges, founding year, registration numbers. This is the one source of truth.
  2. Correct your own site everywhere the fact appears, including structured data. The values in markup must match the visible text.
  3. Correct your Google Business Profile and other profiles you own.
  4. Update directories and listings. Many allow you to claim and edit a listing. Keep a record of each request and the date.
  5. Ask partners and publishers to correct old articles where it matters, politely and with the evidence.
  6. Publish a clear, dated page for the fact that changed, such as a page that says we moved to the new office on a given date.

The consistency work is the same discipline we describe in entity SEO and brand consistency. The approach to prices needs care: if you publish ranges or starting prices, label the date. A price with no date turns into a wrong price the day you change it.

Step 4: Report the error to the platform

Most answer engines provide a way to give feedback on a response, such as a thumbs down or a report option. Use it, and describe the factual error and the correct fact plainly. Do not expect a personal reply or a guaranteed change. Some engines also have a form for businesses or publishers. Check the current help pages for each engine, because routes change. Keep records of when you reported.

What we cannot do. We cannot edit another company's model or force a correction. No agency can. What can be done is to remove wrong sources, add consistent correct ones and measure whether the answers improve. Be wary of anyone who promises an instant fix.

Step 5: Check again on a schedule

Repeat the same prompts after one week, two weeks and a month, with the same engines and settings and the same number of runs. Compare the counts: the wrong fact appeared in four of five runs before, and now it appears in how many. Keep the log. If the wrong fact persists in answers that cite a source, look at that source again. If it persists in answers with no source, keep strengthening the correct facts and review again next month.

What about wrong prices and services?

Wrong prices cause the most damage because they send buyers away or create disputes. Put the current pricing logic on a dedicated page in plain text: what you charge, what it includes, and the date reviewed. Remove old price lists, PDFs and cached pages that you control. If your prices are quotes, say that and publish the factors and a typical range if you can. For services you no longer offer, either remove the page with a proper redirect to the nearest current page, or keep it with a clear note that the service is discontinued and a pointer to the replacement.

What about businesses with several branches or names?

Multiple branches multiply the chances of conflicting details. Give each branch its own page with its own address, phone and hours, and make sure listings match page by page. If your company has changed its name, publish a page that states both names and the date of the change, and update listings. Our post on when ChatGPT confuses your company with another covers name collisions.

What evidence should you keep?

Keep a folder with dated screenshots of each wrong answer, the corrected pages with their dates, the emails or form submissions to directories, and the follow up test results. This record helps in three ways. It shows you whether your fixes worked. It lets you escalate to a platform or directory with proof. And it protects you if a customer relied on a wrong price or address shown by an engine and complains to you. You can show what was published by you, what was published by others and when each was corrected.

What should you do next?

Today, record three wrong answers with dates, and search the web for the wrong fact. This week, correct your own site and your top five listings and write your approved facts sheet. If you want a full check of how five AI engines describe your business, with the fixes carried out for you, the GEO Pilot is a fixed $999 and runs for 30 days. The broader method is in how to fix wrong information about your brand in AI, and the AI Visibility Checker is a free first look.

What the research says

Evidence behind this page

Checked figures from published studies, each with its source and date. More in our AI search statistics.

47.9%

Wikipedia accounted for 47.9% of ChatGPT's top-10 cited sources (7.8% of all ChatGPT citations); Reddit led Perplexity (46.7% of top-10 share) and Google AI Overviews (21.0%).

Sample: 680 million citations. Period: Aug 2024 – June 2025

Primary sourceProfoundJune 2025 (updated Aug 2025)
28.7 to 32.3

ChatGPT had the lowest overlap between its mention list and its source list: 28.7 to 32.3 shared domains out of 100 in the UK and 32.8 to 35.8 in the US.

Sample: 2.5M UK and 25M US prompts per month per platform; AI Overviews from 4.35M to 54.6M keywords a month. Period: January to July 2026, seven monthly measurements

Primary sourceSemrushOctober 2026
12%

Only about 12% (11.9%) of URLs cited by ChatGPT, Gemini, Copilot and Perplexity ranked in Google's top 10 for the original prompt; Perplexity was highest at 28.6%, ChatGPT in-text citations 8.0%.

Sample: 15,000 long-tail queries. Period: ChatGPT data early July 2025

Primary sourceAhrefsAugust 2025

Cite this guide

You may quote this page with a link. Where a figure comes from another study, please credit that study too.

GrowingAI. (2026, 18 October 2026). ChatGPT shows the wrong address, phone number or prices for my business: how to fix it. https://thegrowingai.com/chatgpt-shows-wrong-address-phone-or-prices-for-my-business/

FAQ

Questions people ask

Can I ask ChatGPT to correct its answer about my business?

Telling the chat does not change what other users see. Use the feedback tool and fix the web sources behind the wrong fact.

How long until the answers are correct?

No engine publishes a timeline. Fix the sources, then re check at one week, two weeks and a month and keep a log.

Why does the answer cite a directory I never heard of?

Engines use many sources. Claim and correct that listing, and check similar directories.

Does structured data fix wrong details?

It can help describe correct facts clearly, but it must match the visible page and does not override other sources.

Is it worth hiring someone for this?

If the errors affect sales and you lack time, an outside team can trace the sources and fix them faster. Check that they promise process, not outcomes.

Nitesh Dan Charan

Nitesh Dan Charan

Founder, GrowingAI

Nitesh is the founder of GrowingAI, a generative engine optimization (GEO) agency that helps brands get named and cited in AI answers on ChatGPT, Google AI Overviews, Gemini and Perplexity. He writes from client work in the UAE, USA and UK. Every guide is dated, sourced and reviewed before it is published. See our research standards.

Want this done for your site?

Send your site and five buyer questions. We test them in the major AI engines and send back a short, honest read.

  • Recorded baseline in five AI engines
  • Competitors named
  • Three first fixes, ranked

No card, no contract. Nitesh replies himself, usually the same working day. By sending, you agree we may contact you about this request (privacy policy).