ChatGPT has my business hours or address wrong — how do I fix incorrect AI information?
TL;DR: You fix the sources, not the model. AI assistants assemble facts from your website, Google Business Profile, Yelp, and directory listings — when they're wrong, some source is wrong, stale, or contradictory. Correct every listing, state the facts in crawlable HTML and schema on your own site, and the answers follow within crawl cycles.
Why AI gets your facts wrong
Claim: Incorrect AI answers about your business are almost always a data-consistency problem you can trace and fix.
Evidence: Engines answering with live search cite their sources — expand the citations and you'll usually find the culprit: an old address on a directory you forgot existed, conflicting hours between your site and GBP, or a defunct listing from a previous tenant at your address. Engines answering from model knowledge alone can carry outdated training data; the countermeasure is making current facts so consistent across retrievable sources that search-grounded answers override stale memory.
Common wrong-fact causes
| Symptom | Likely source problem |
|---|---|
| Old address | Unclaimed or stale directory listings |
| Wrong hours | Site vs. GBP mismatch, or hours only in a JS widget |
| "Permanently closed" | A stale flag on a major listing platform |
| Wrong services or prices | Outdated pages still indexed, PDFs never updated |
| Confused with another business | Similar names; missing schema disambiguation |
Step-by-step: the correction protocol
- Reproduce the error in a fresh session and expand any citations — that's your source list.
- Fix your own site first: correct facts in plain HTML plus LocalBusiness schema with
openingHoursSpecification, address, andgeo. - Sweep the listings: GBP, Yelp, Bing Places, Apple Business Connect, Facebook, and industry directories. Claim, correct, or close each.
- Kill contradictions — one canonical NAP everywhere.
- Request recrawls via Search Console and Bing Webmaster Tools/IndexNow.
- Re-test weekly until the answer corrects. BookRails.ai's weekly scans across all four engines catch both the fix landing and any future regressions.
FAQ
Can I just tell ChatGPT it's wrong? In-chat corrections only affect that conversation. Durable fixes live in the sources engines retrieve.
How long until the correction shows up? Search-grounded answers can correct within days to weeks of sources updating; pure model-knowledge answers may lag until retrieval or a model update overrides them.
Is there a form to report wrong AI info about my business? No universal one. Feedback buttons exist per product, but source correction is the reliable path.
How do I stop this happening again? Consistency plus monitoring: one canonical fact set, and a recurring scan that flags drift the week it appears.
By Pinal Dave Last updated: 2026-07-23