ChatGPT Is Saying the Wrong Thing About Your Business: How to Correct It

An old price list, a closed office, a competitor named where you should be: when ChatGPT gets wrong information about your business, the fix is rarely a single message to OpenAI. It is usually a job of making the correct facts easy to find, consistent everywhere, and published on your own domain in a form machines can read. This article explains where those wrong answers come from and gives you a plan you can start this week.

Where ChatGPT gets its facts about your business

An AI assistant answers questions about a company from two broad sources, and knowing which one produced the error tells you how to fix it.

Training data

The underlying model learned from a large snapshot of the web taken at a point in time. If your prices changed or you moved premises after that snapshot, the model may still repeat the old version. Nothing you publish today changes that snapshot. Future models are trained on newer data, so correct pages you publish now feed into later versions.

Live web search

When ChatGPT searches the web before answering, it pulls in current pages and summarises them. Here the error usually comes from whatever it found: an outdated directory listing, an old review site, a forum thread, or your own website still showing last year's details on a forgotten page. This is the part you can influence quickly.

A quick way to tell the two apart: if the answer includes links, look at what it cited. Those sources are your first targets. If there are no links, the answer probably came from training data, and the plan below still applies, just on a longer timescale.

Why your own website is the main lever

Assistants lean on sources that are clear, current and consistent. When your site states a fact plainly and the rest of the web agrees, that fact tends to win. When your site is vague, or buries the answer in a PDF or an image, the assistant fills the gap from somewhere else.

Many small business sites never actually state the things people ask about. The address sits only in a map embed. Prices live in a downloadable brochure. The list of services is a row of icons with no text. A human can work it out; a machine often cannot.

We go deeper into how assistants pick sources in how Perplexity, Claude and Google AI Overviews choose their sources. The short version: be the clearest, most up to date page on the question.

Your correction plan for this week

Day one: record exactly what is wrong

1. Ask ChatGPT, and ideally Perplexity and Google's AI Overviews too, the questions a customer would ask: what you do, where you are, what you charge, whether you are any good, and who your alternatives are.
2. Save the answers and any sources they cite.
3. List each wrong fact next to the correct version and the source that seems to be feeding it.

Day two: fix your own site first

• Update every page that mentions old prices, addresses, opening hours or services, including old blog posts and landing pages you forgot existed.
• Write a clear About or Company page that states who you are, where you operate, what you sell and who you serve, in plain sentences.
• Put contact details in text in the footer, not only in an image or map.
• Add structured data, the small block of code that labels your name, address and phone number for machines. Most WordPress SEO plugins and Shopify themes can add basic organisation or local business markup.

If you want a structured sweep of the site, our 30-minute SEO audit covers the pages worth checking.

Day three: fix the listings that disagree with you

• Update your Google Business Profile, which feeds a great deal of local information.
• Correct Bing Places, since ChatGPT's web search has historically drawn on Bing's index.
• Check Companies House, Yelp, trade directories, industry bodies and any review sites the assistant cited.
• Ask the owners of outdated third party articles to update them. Not all will, but many will fix an address or price if asked politely.

Day four: publish pages that answer the questions directly

If the assistant recommended a competitor, it usually means you have no page that clearly answers the question it was asked. A local accountant with no page explaining their fees or specialisms will lose to one that has. Write the pages that match those questions: a pricing page, a page per core service, an FAQ built around what customers actually ask. Our guide to building an FAQ page AI assistants use shows the format.

Day five: get the changes crawled

Updated pages only help once search engines and assistants see them. Resubmit changed URLs in Google Search Console and Bing Webmaster Tools, and make sure your sitemap is current. Bing matters here more than most owners expect.

Feedback routes inside the assistants

These exist and are worth using, but set your expectations sensibly. They are signals, not edit buttons.

• ChatGPT: use the thumbs down on the wrong answer and explain the error. OpenAI also provides a privacy and personal data request route, which is aimed at individuals rather than businesses.
• Google AI Overviews: use the feedback option on the overview itself. Fixing the underlying pages and your Business Profile does more.
• Perplexity: flag the answer, and check which sources it cited, since those are what you need to correct.

Feedback tells the assistant it was wrong. Your website tells it what is right.

How long the correction takes

Answers built from live search can change within days or weeks of the sources being fixed and recrawled. Answers built from training data change only when a newer model is released, which is out of your hands. That is why the steps above focus on sources: they cover both cases.

Re-run your day one questions every few weeks and note what has changed. If one wrong fact keeps coming back, search for it in quotes to find the page still repeating it.

Keeping it right after it is fixed

The businesses assistants describe accurately tend to be the ones that publish regularly on their own site: current prices, new services, useful answers to real questions. A site last updated three years ago leaves the gap open for someone else to fill. A light routine works: update key facts whenever they change, and keep adding pages that answer what customers ask.

Where Helio SEO fits in

Helio SEO writes and publishes clear, current articles on your own domain, written to be quoted by assistants such as ChatGPT and Perplexity as well as ranked by Google, which helps keep the right facts in circulation. It will not fix your directory listings or old third party pages for you, so the correction steps above are still yours to do. If you would rather not keep the publishing side going yourself, here is how it works.