What if ChatGPT is wrong about my company?
When ChatGPT is wrong about your company, the claim usually matches an old page on a third-party site, not your own site. Find that page by searching the exact wrong wording, correct it at the source with a dated line, then wait for it to be recrawled. Corrections take days to months to show.
People search "chatgpt wrong about my company" after seeing a wrong price, a missing product or a feature you never offered. The steps below trace where the claim came from and fix it there. They assume you have already looked at the answer, as in how to check what AI says about your company.
How do I record the wrong answer first?
Write down the wrong answer before changing anything. Note the date, the engine, the exact question, and whether you were logged out. Copy the wrong sentence word for word.
AI answers vary from one run to the next, so ask the same question a few times and keep each result. If the claim appears in only one of five runs, treat it as lower priority. If it appears in most of them, something it relies on is probably wrong.
How do I find the page the wrong claim came from?
Search for the wrong claim's distinctive wording in quotes on Google and Bing. The source page is often one that states the claim in nearly the same words as the answer.
Work through these steps in order:
- Pick the most specific phrase in the wrong answer, such as a plan name, a limit or a feature label.
- Search it in quotes, then search it again with your company name beside it.
- Open the top results and look at each page's date, if it shows one.
- Ask the engine for its sources, if it offers them, and compare the cited pages with your search results.
- Rank the candidates by how closely their wording matches the answer, and start with the closest.
An old page with the same wording as the answer is the likely carrier. A page that is newer but vaguer is less likely.
Which third-party pages are most often behind a wrong claim?
Third-party listings, old comparison pages and review sites are the usual carriers, because they are rarely updated after a company changes. Check each type in turn.
| Page type | Why it goes stale | Who can fix it |
|---|---|---|
| Software directory listing | Filled in once at signup | You, through the vendor profile |
| Review site profile | Pricing and feature fields copied from an old version | You, through the vendor portal or support |
| "Alternatives to X" or "X vs Y" page | Written by a competitor or a blogger at one point in time | The publisher, by request |
| Old blog post or press release | Describes the product as it was at launch | You, if it is on your domain |
| Your own pricing or docs page | Old versions left live or duplicated | You |
Check your own domain too. Old pricing pages, archived docs and duplicate landing pages sometimes stay live and contradict the current ones.
What if the page is about a different company with the same name?
The wrong claim can come from a page about another company with a similar name. I saw this in a run for Selah, my own pricing tool for creators. On 24 October 2026, Perplexity cited selah.so, an unrelated Bible-study app, on two questions about Selah. One answer opened with "Assuming you mean Selah.fyi". The shared name led the engine to guess which Selah was meant.
The fix I proposed was small. Describe the company the same way everywhere, such as "Selah, the brand deal pricing tool", in page titles and in the site's Organization schema. If your search turns up a page about someone else, the correction is to make your own pages unambiguous, because you cannot edit theirs.
How do I fix the source page?
Fix the source page directly and put a date on the correction. A page that says "Pricing as of November 2026" makes it clear to readers that the older figure is out of date.
- Your own pages: update the wrong claim, add the date, and remove or redirect any stale duplicates.
- Listings and review profiles: edit the profile, or send the site's support team the corrected facts with a link to your own page.
- Another company's page: write to the publisher with the exact sentence, the correct fact and a link to your current page. Keep it short and polite.
Keep the same wording for the corrected fact on every page you control. If your site, your directory profiles and your docs all say the same thing, an engine has fewer places to find a different answer.
How do I tell the engines about a correction?
Tell the engine vendors where they have a channel for it, but expect the corrected source page to do most of the work. According to eSEOspace's October 2026 guide, which summarizes statements from the vendors:
- Microsoft says IndexNow notifications help Bing reduce outdated or incorrect URL references in Copilot responses and grounding results.
- Google says it reviews feedback about knowledge panels from verified users within a few days and sends a confirmation email with a resolution update.
- OpenAI's content reporting webform covers content that may violate its Terms of Use or applicable laws. OpenAI also states that reported domains and other content may be reviewed by its Model Quality team, which may apply filters to help prevent ChatGPT from relying on unreliable sources.
A wrong price is not a clear violation of the Terms of Use, so I would not expect the webform to fix a stale listing. Google's AI Overviews help page also says AI Overviews can and will make mistakes, so some errors will happen even when every source is right.
How long does a correction take to show?
A correction can take from several days to several months to show. Google says recrawling of a fixed page can take that long before assistants fetch the corrected version, per the same eSEOspace guide from October 2026.
The number covers recrawling only. It does not say how long a given assistant takes to change its answer after the page changes. Re-ask your logged question every week or two and keep the dated results, so you can see whether the wrong claim fades or stays.
What should I not do when ChatGPT gets my company wrong?
Do not post fake reviews, and do not try to trick the model with prompt tricks or hidden text. Neither fixes the source of the error. Fake reviews can break the rules of the review site and damage buyer trust. Hidden instructions on a page are the kind of thing OpenAI says its Model Quality team may filter out.
Do not edit only your homepage and stop. If the wrong claim came from an old comparison page or a directory listing, that page still carries it.
What is the next step?
Record the answer, search the wording, rank the candidate pages and correct the closest match with a dated line. Then re-ask the question on a schedule.
The free teardown shows what AI tells your buyers about you today, so you can see which wrong claims to trace first. The method describes how the checking is run.
Sources
- What to do when AI says something wrong about your business — eSEOspace, 2026-10-13