How do I tell if AI systems have the wrong idea of what my business is?
Ask plainly and read the errors. Prompt an assistant to describe your company, list its services, name its service area, and compare it to a named competitor. Wrong category, wrong city, services you dropped years ago, or details belonging to a similarly named company are entity problems, not content problems. They come from inconsistent third-party records, and they are fixed at those sources rather than by rewriting your homepage.
What an entity is, in practice
An entity is a resolved thing — a specific company, in a specific place, with a category and a set of attributes — assembled from many independent sources that mostly agree with each other. Your website is one source among many, and not automatically the most trusted one.
That is the part people find counterintuitive. You can state a fact clearly on your own site and still lose to a dozen directories that say something older and disagree with you in unison.
The four failure modes
- Wrong category. You are described as a general contractor when you are a plumbing company, usually because an early listing chose the wrong primary category and everything downstream copied it.
- Conflation. Your details are merged with a similarly named business, often in another state. Shared owner names and generic trade names make this worse.
- Stale attributes. Services you dropped, a suite number you left, a phone number that forwards, an owner who retired.
- Split identity. Two half-correct records for one business — commonly caused by a legal name and a doing-business-as name that never got reconciled.
Trace the wrong fact to its loudest source
Search the incorrect detail itself, in quotes. You will usually find the handful of pages carrying it, and one of them will be the origin the others copied. Directories and data aggregators propagate quickly and correct slowly, which is why the same wrong suite number can survive a decade.
Correcting the origin matters more than correcting ten downstream copies, because the copies get refreshed from upstream. Keeping listings consistent across platforms is unglamorous and is most of the actual work.
Fix at the source, then make your own site unambiguous
Write a canonical fact set — legal name, public name, address, phone, categories, service area, founding year — and make every property match it exactly, including punctuation. Mark it up as Organization data with links to the profiles you control, so a machine can tie the records together.
Then expect lag. Records that took years to spread do not correct in a week, and no one can promise a date. Track it by re-running the same description prompt monthly and noting which errors have dropped out. That measurement discipline is described in the companion answer on measuring visibility, and it applies to entity work as much as to citations.
Topics: entities · local search · listings · diagnostics
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.