AI Is Saying Something Wrong About Your Brand
An AI platform states a wrong price, founding year or fact about you. How to find the hallucination alerts in your audit and correct the source.
ChatGPT, Perplexity or another platform told someone the wrong price, the wrong founding year, or that you were bought by a company you've never met. You want to know where it came from and how to make it stop.
Likely causes
- Your own site disagrees with itself. An old pricing page, a stale blog post or a press release still states the old fact.
- The right fact is only drawn by JavaScript. Most AI crawlers read raw HTML. If the price appears only after scripts run, they never saw it.
- A crawler can't reach the page. robots.txt or a firewall turns the retrieval bots away.
- The model learned it before you changed it. We ask ChatGPT and Claude without web search, so their answers reflect what the model already knows.
- Another site repeats the old fact. Directories and review sites often lag behind your own pages.
Find what was flagged
- Open the audit from Audits and click Show advanced.
- Read Accuracy Issues. Each flag is Critical, Moderate or Minor, with what the AI said and the fact it contradicts. The full text of each flag is on Starter ($49/mo) and above; Free shows the count.
- In Query Insights, click the keyword to read the whole answer and the sources it cited.
With "Accuracy alerts" on under Settings → Email Notifications, we email you when an audit finds a critical flag.
Fix it at the source
- Search your own site for the wrong value and correct every copy, including old posts and PDFs.
- Open the page with JavaScript turned off. If the correct fact disappears, ask your developer to render it in the HTML.
- Open Crawlability and click Check AI crawler access. Fix any retrieval crawler marked Blocked.
- Look at the sources the wrong answer cited. Ask those sites to correct the fact.
- Run a new audit. Platforms that search the web can change once they re-read your pages. ChatGPT and Claude, as we ask them, change only when the model is retrained.
V can do the first pass: pick "What are they getting wrong" on the V tab and it lists each error with a drafted correction.
What the check can and can't catch
The check compares answers with the brand facts you entered: dollar prices, founding year, location and features. It also flags acquisition claims and staff counts it can't verify. If a flag looks wrong, read A hallucination flag looks wrong.
How it works: Hallucination detection reference lists every check, its severity and the accuracy formula.
Still stuck?
Can't find where the wrong fact comes from? Send us the audit link and the flag.
In the dashboard, click the Help icon in the top bar and choose Ask support. It appears once you have set up a brand. The Help tab answers from these articles and from your own account.
Press Talk to a person and the conversation so far becomes a ticket. We reply by email to the address on your account.
Can't get into the dashboard? Email hello@99visibility.com or use the contact form. We don't promise a reply time on any plan; Pro tickets are answered first.
Didn't find your answer?
Use Ask support in your dashboard, or email hello@99visibility.com. We reply by email; Pro tickets are answered first.