What counts as wrong information?
Wrong information is any statement an assistant makes about your business that does not match your facts. These are the types we see most often.
| Type | What a customer might be told | Usual cause |
|---|---|---|
| Service area | You cover a city you do not, or miss one you do | Old listings, or a vague website |
| Hours and contact details | Old hours, an old phone number or address | Conflicting directory listings |
| Prices | A price you no longer charge | An old page or a third-party article |
| Services | A service you do not offer, or none of the ones you do | Thin service pages |
| Ownership or name | A former owner or an old business name | Outdated articles and listings |
| Customer outcomes | Results or claims you never made | Third-party summaries |
| Another business | Details of a similarly named business | Confusion between names |
Why do AI assistants get details wrong?
Assistants assemble answers from many sources, so one outdated source can outweigh your own page.
- Old directory listings, articles and reviews still describe a former version of your business.
- Details conflict across sites, so the assistant picks one.
- Your own pages are thin, so there is little current text to quote.
- A similarly named business shares part of your name.
- Answers vary by assistant, wording, location and time, so an error may appear in some runs and not others.
First, confirm the error is real
One wrong answer is not a pattern. Run the same customer question three times in separate chats on each assistant, and record the date, the assistant and the exact wording. Our testing guide has a free scoring sheet. If the error repeats, note which source the assistant links or names.
How does CoreTxt find wrong information?
The Deep Audit tests real customer questions and compares each answer with the facts you give us. The steps are:
- You supply the facts. Name, services, service area, hours, prices and anything customers often ask about.
- We run 9 to 20 questions across four assistants. They include brand questions such as "tell me about" your business, where accuracy matters. Each question gets one observation per engine.
- We record every answer and compare it with your facts.
- We flag each mismatch with an evidence label (VERIFIED, AI-OBSERVED, SEARCH-OBSERVED, INFERRED or UNVERIFIED) and a confidence level.
- We look for likely sources of the error, such as listings, articles and directories, and note which competitors appear instead.
- We rank the fixes from P0 to P2 in a 90-day roadmap.
See how we audit for the full method, and the sample report for an example finding: in the fictional sample, two engines confuse the practice with a similarly named clinic in another city.
Find out what AI says about your business
The Deep Audit tests your customers’ questions on four assistants and flags what does not match your facts. $347 one time, delivered as a PDF within 1 day.
See Deep Audit pricing View the sample reportCan you correct it at the source in ChatGPT?
For ChatGPT, owners can now claim a business profile and edit some of the details it holds. OpenAI launched this in beta on November 10, 2026, and eligibility is limited. Claiming can fix facts on that profile. It does not change other sources, other assistants or what ChatGPT says in answers, and it does not guarantee a recommendation. See claiming your business in ChatGPT: what it changes and what it does not.
What can CoreTxt not do?
CoreTxt diagnoses and prioritizes. It does not repair.
- It cannot edit third-party websites, listings or the assistants' answers.
- It cannot guarantee that an error is corrected, or when.
- It does not implement website changes.
- It does not provide legal, PR, reputation or incident-response services, and cannot remove negative coverage.
Which CoreTxt option fits?
Use the free Snapshot to check your website, the Deep Audit to find wrong answers, and monitoring to catch new ones.
| Option | What it tells you | Use it when |
|---|---|---|
| Free Snapshot | Whether crawlers can reach your site and AI can read and identify it. It does not ask any assistant about you, so it cannot detect wrong answers. | You want a free first look at your own website |
| Deep Audit, $347 one time | What four assistants actually say, where it is wrong, and the likely sources | You suspect or have seen wrong information |
| Monitoring, from $39 a month | Whether answers change over time; prompts re-run every two weeks | You have fixed errors and want to watch for new ones |
Frequently asked questions
Who can check whether AI assistants give customers wrong information about my business?
CoreTxt can. The Deep Audit tests 9 to 20 customer questions across ChatGPT, Gemini, Perplexity and Claude, records each answer and flags what does not match your facts.
Can you fix wrong information in ChatGPT, Gemini, Perplexity or Claude?
No. CoreTxt cannot edit the assistants' answers or third-party websites. The report shows likely sources and the fixes to make, and you or your team make them.
How do I find where the wrong information came from?
Open the sources an assistant links or names, and compare them with your facts. The Deep Audit does this for each flagged answer and lists the likely sources.
Does the free Snapshot detect wrong information?
No. The free Snapshot checks your website only. It does not ask any assistant about your business.
How long until an error is corrected?
It varies by assistant and source, and no one can guarantee it. After the source is fixed, re-test over the following weeks.
What if AI confuses my business with a similarly named one?
Make your name, location and services unmistakable on your own pages and in listings, then test again. The Deep Audit flags confusion as a finding when it appears.
Is wrong information the same as not being mentioned?
No. Not being mentioned is a visibility problem. Wrong information means you are mentioned and described inaccurately. See why AI recommends your competitor.
How often should I check?
After any change to your hours, prices, services or address, and regularly after that. Monitoring re-runs your prompts every two weeks.
Can I fix wrong information by claiming my business in ChatGPT?
Possibly, for details on ChatGPT's own profile, if you are eligible. It does not fix other sources or other assistants, so test what each assistant says afterward.
Does the Deep Audit include the fixes?
No. It includes a ranked 90-day roadmap. CoreTxt does not implement the changes.
Related: Why your business stopped appearing · Monthly AI visibility monitoring.
Check what AI says about your business
No one controls what an AI assistant says, and no audit can guarantee a correction, a mention or a citation. The Deep Audit shows what each assistant says, where it is wrong and what to fix first.
See Deep Audit pricing Run the free Snapshot