Why AI gives the wrong phone number (or city) for your business — and how to fix it
A customer asks an assistant for a plumber, gets your name — and a number that hasn't worked in three years. That is worse than not being named at all: the customer dials, gets nothing, and calls the next name on the list.
Wrong facts in AI answers are common for local businesses, and they are usually fixable, because the assistant did not invent the number. It read it somewhere.
Where an assistant gets your phone number and city
Assistants build answers from what they can read: your own website, your Google Business Profile, directories and review sites, old press or listing pages, and whatever the model learned during training. If those sources disagree, the assistant picks one — and it is not always the current one.
The usual causes of a wrong number or city:
- An old number that still lives somewhere. A previous landline on a directory listing, a PDF flyer, an old "contact" page that was never deleted, or a social profile nobody updated.
- Call-tracking numbers. Ads and some listing services show a different tracking number per channel. To a machine, that looks like a business with four phone numbers.
- A move or a rename. You moved from one city to the next, or changed the company name, and the web still carries both versions.
- No machine-readable facts on your site. If your site never states name, phone and service area in structured data, the assistant has to guess from page text — or from someone else's page about you.
- Service-area confusion. A business that serves five cities but only names one on its site can be placed in the wrong city, or in none.
How to fix it
- Pick one canonical number and one canonical name. Use that exact number everywhere: site, Google Business Profile, directories, social profiles, email signature.
- Put the facts in structured data. A LocalBusiness schema block on your homepage with name, telephone, and the area you serve (city-level is enough for a service-area business — you do not need to publish a home address) gives every crawler an unambiguous source. Our guide to LocalBusiness schema for contractors shows the exact block.
- Make the visible page agree with the schema. The number in your header, footer and contact page should match the schema and be a real tel: link.
- Clean up the old copies. Update or remove old listings, retire tracking numbers you no longer need, and redirect deleted pages instead of leaving them live.
- Fix the profile the assistant is most likely to read. For Google-side answers that is your Google Business Profile; keep its phone, categories and service area current.
- Re-check after a few weeks. Assistants do not update instantly. Corrections propagate as sources are re-crawled and models are refreshed.
What you cannot control
You cannot edit what a model already learned, and you cannot force a directory you do not own to update on your schedule. What you can do is make the correct facts the easiest ones to find, in the most places, in the most machine-readable form. Over time that usually wins.
AI answers change daily; a scan is a snapshot, not a ranking — no tool can guarantee an AI will name you or quote you correctly.
How to check whether AI has your facts right
Ask the assistant the question your customer would ask ("who is a good electrician in [city]?") and compare the number and city it gives with your own. RankVolt's AI Answer Check does this with a real customer question and flags when the assistant's phone number or city differs from what your own site says, so you know which fact to chase.
FAQ
- Why does the AI show my old phone number?
- Because the old number still exists somewhere the assistant can read — a directory, an old page, a tracking number on a listing. Find and update those sources; the assistant cannot know the number changed unless the web says so.
- Does a wrong city in an AI answer hurt me?
- It can. A customer in your real service area may be told you are somewhere else and call a competitor. Stating your service area clearly on your site and in structured data reduces the guesswork.
- How long until the AI answer is corrected?
- There is no fixed timeline. Live-search assistants can pick up a change within days or weeks of the page being re-crawled; facts learned during training update only when the model is refreshed. Keep every source consistent and re-check periodically.