How to Correct Wrong Information in ChatGPT

• An outdated source — when ChatGPT searches, it repeats what the pages it reads say. An old price list or a closed branch page is enough.
• Training data — without a search, ChatGPT answers from what its model learned, which can be months or years old.
• Name confusion — another company with a similar name, or a former business at the same address.
• Mixed-up details — facts from different locations, products or plans combined into one answer.

Record the question, the full answer, the date and every source link shown. Ask the question a few times, in a temporary chat, to see whether the error is consistent or occasional — a recurring error is worth more effort than a one-off.

Open the cited links and search them for the wrong fact. If none contains it, search the web for the wrong value itself — the old price, the old address — to find pages still stating it. If nothing turns up and the answer shows no sources, the error probably comes from training data.

• Update or remove pages with old facts; redirect pages for discontinued services or closed locations.
• Keep one clear page for the facts people ask about — prices, locations, hours, what you do and don't offer — and show when it was last updated.
• State the facts in plain text, not only inside images or PDFs.
• If you use structured data, make it match the visible page (see structured data for AI search).

Update your directory listings and profiles, claim the ones you haven't, and ask publishers of outdated articles for a correction — with the correct fact and a link to where you state it. Prioritise sources that ChatGPT actually cites for the question.

If ChatGPT mixes you up with another company, make your identity unambiguous: use your full business name consistently, state your location and category near the name, and link your official profiles from your site.

Answers built from web search can change once the sources are corrected and re-crawled — often within weeks, but there's no guaranteed timeline. Errors from training data can persist until newer models are released. Re-check the same questions on a schedule, and use ChatGPT brand monitoring to keep the answers in view.

• Telling ChatGPT it's wrong in a chat — that conversation doesn't change answers for anyone else.
• Publishing many near-identical pages repeating the correct fact — one clear, current page beats ten thin ones.
• Hiding text for crawlers that visitors can't see — it breaks search engine guidelines and helps nothing.