What AI Says About Your Business, and How to Fix It When It’s Wrong

Revised August 22, 2026

Imagine it’s a Tuesday night in Des Peres and you’re half watching a ballgame. On a whim you open an AI assistant on your phone and type your own business name. You’re curious, not worried. Thirty seconds later you’re worried.

It has your Saturday hours off by an hour. It says you cover a chunk of the county you stopped serving two years ago, and it has never heard of the service that now pays your rent. A friend in Overland runs the same test and gets a phone number that rang at her old suite. An owner in Granite City gets something stranger: a clean, confident, beautifully formatted answer about her competitor, and no mention of her at all.

None of that is a bug you can call someone about. It’s an answer assembled from sources, and the sources are fixable. Here’s the check, where the bad data actually lives, and the repair order that keeps you from wasting three weeks.

How to correct incorrect AI information?

You can’t edit an AI answer. You correct the sources it was assembled from — your Google Business Profile, your Apple and Bing listings, your own website, and the directory records that repeat your details — and then wait for the assistants to read them again. There is no support line for “the AI said something wrong about my business.”

Google does have a feedback path, and it’s worth using, but be clear about what it is. Google’s Search help says that at the bottom of each AI Overview there’s a thumbs up and thumbs down; clicking thumbs down leads to Report a problem, where you pick a category and add detail. That’s feedback for improving the feature, not a ticket that gets your hours changed. Google states plainly on the same page that AI Overviews “can and will make mistakes.”

ChatGPT has no business-facts correction channel at all. OpenAI’s privacy portal accepts a “Remove my personal data from ChatGPT responses” request, and its help documentation is explicit that this covers personal information about an individual that is inaccurate, excessive, irrelevant, or no longer appropriate. It requires government-issued ID and is assessed case by case. It is not a form for “my shop closes at 6, not 5.”

So the work is upstream. Everything below is about finding which upstream source is lying and fixing that one.

How do you check what AI says about your business?

You ask the assistants directly, in writing, and you log what comes back. Budget 15 minutes: eight fixed prompts, run in three places, recorded on one page. The point is not to feel bad about the answers. It is to produce a list of specific wrong facts, each traced to the source that most likely fed it.

Open a blank page, split it into three columns — what it said, what’s true, where that probably came from — and run these eight prompts in at least three places: ChatGPT, Google (both the AI Overview and AI Mode), and Microsoft Copilot. Same prompts every time, so next quarter’s run is comparable.

Substitute your own details for the brackets:

• What is [business name] in [your town], Missouri? — you’re checking whether it knows you exist and what it thinks you do.
• What are [business name]’s hours? — the most damaging thing on this list to get wrong, because it costs you customers who never call.
• What’s the phone number and address for [business name]? — watch for an old suite number, an old area code, or a tracking number you retired.
• Does [business name] do [your highest-margin service]? — assistants routinely describe you by the service you led with five years ago.
• What areas does [business name] serve? — especially important across the river, where an assistant will happily put a Missouri-only operator in Illinois or the reverse.
• Who are the best [your trade] near [your town]? — the hard one. You’re not checking your facts here. You’re checking whether you are named at all.
• Is [business name] still open? — this catches ghost records, old locations, and a stray “permanently closed” flag.
• What do people say about [business name]? — this shows which reviews the model is weighting, which is usually not the ones you would pick.

Two rules make the exercise worth the quarter hour. First, screenshot everything, with the date and the exact prompt visible. You’ll want the before-and-after in 90 days, and you’ll want proof if a wrong answer ever costs you something. Second, write down every source the assistant cites. Those citations are the map. When an assistant links a 2019 chamber page carrying your old phone number, you’ve just found the thing to fix, and you found it in about forty seconds.

Before you compare anything, spend ten minutes writing your one-page source of truth: legal name, the public-facing name exactly as it appears on your sign, street address formatted the way it appears on your Business Profile, one phone number, hours including holidays, the actual list of services you sell today, and the towns you actually drive to. Everything downstream gets copied from that page. Skip it and you’ll publish three slightly different versions of yourself, which is the exact problem you’re trying to solve.

Why does AI give incorrect information?

Because the answer is assembled, not looked up. An assistant pulls fragments from a search index, from structured listings, and from patterns it absorbed in training, then writes smooth prose over the top. If your hours are stale in one place and correct in another, nothing in that process knows which is current. It picks. The fluency of the sentence tells you nothing about the accuracy of the fact.

The failure rate is measurable. The European Broadcasting Union and the BBC published the largest study of its kind on 22 October 2025: 22 public service media organizations across 18 countries, working in 14 languages, assessed more than 3,000 answers from ChatGPT, Copilot, Gemini, and Perplexity. 45% of answers had at least one significant issue. Sourcing was the worst category at 31%, and accuracy problems affected 20%. Gemini fared worst on sourcing, with significant issues in 76% of its responses.

That study looked at news, not at whether a heating company in Festus is open on Sunday. Nobody has published equivalent research on local business facts, so don’t treat 45% as your number. Treat it as evidence that the failure mode is real and systemic rather than a fluke you happened to hit.

Where do the wrong answers actually come from?

Six sources cause most of it, and they fail in different ways.

A stale Google Business Profile. This is the loudest source, and the one most owners have not touched since the year they claimed it. If your services list, hours, or categories are two years old, an assistant reading Google’s data is reading a two-year-old business. The mechanics of getting that profile complete are covered properly in our guide to optimizing your Google Business Profile — there’s no reason to repeat it here. What matters for this audit is one specific screen.

Edits Google made for you. Google’s own documentation says it gathers business information from sources including user reports and licensed content, and that if sources report your info is incorrect or outdated, it may update your profile. In the profile editor, blue text means Google changed that field, not you. Most owners have never looked. Go look. A Google update you never noticed is a wrong answer with a very respectable source behind it.

Inconsistent listings across directories. You moved, or added a suite number, or switched to a cell phone, and you updated four places out of eleven. Every stale record is a vote for the old version. Assistants don’t audit, they aggregate, and the version that appears most often tends to win.

Old citations nobody maintains. A 2018 press release. A chamber directory PDF. An event sponsor page from a 5K. A supplier’s dealer locator built from a spreadsheet in 2016. These sit quietly for years, get crawled, and hand a model an address you left a decade ago.

Closed-location records nobody removed. The nastiest one. You moved from one strip center to another and the old record stayed live. Now there are two of you, one stranded at an address where a nail salon operates, and assistants will merge, average, or simply pick between them. Same problem when a business changed hands and the previous owner’s profile was never closed out.

Review content. Assistants summarize what reviewers wrote, and reviewers write about the visit they had, not the business you run now. A cluster of 2022 reviews complaining about a policy you scrapped in 2023 will still shape the paragraph an assistant writes about you.

Wildly off-topic — something is in your basement. Pest control near you.

What can you get corrected, and how long does each fix take?

Be honest with yourself about which of these are levers and which are walls. Roughly, in order of how much control you actually have:

Your own website — total control, minutes. The one source you fully own. Put your name, address, phone, hours, service list, and service-area towns in plain readable text on a page, not baked into an image and not trapped behind a script. This is the cheapest fix on the list and the one people skip.

Your Google Business Profile — strong control, 10 minutes to 30 days. Sign in to the Google Account linked to the profile, open Edit profile, change the field, save. Google’s documentation is specific: edits usually take up to 10 minutes to review, but sometimes up to 30 days, and each edit shows as accepted, pending, or not approved. If Google can’t confirm accuracy it may decline the change, and rejected business information edits can be appealed. Changing your business name after verification may force you to verify again — plan around that, don’t discover it on a Friday.

A profile you don’t own — partial control, unpredictable. For the ghost record at your old address, or a duplicate, Google Maps offers Suggest an edit. Under Place closed or not here you can mark a place temporarily closed, permanently closed, doesn’t exist here, duplicate of another place, or moved to a new location. Google reviews these; it does not promise to act on them, and its documentation warns that edits which would significantly change a business may be rejected.

Apple — free, and it changed names this year. Apple announced Apple Business on 24 March 2026 and launched it on 14 April 2026 as a free service in the US and 200-plus countries and regions. It replaced Apple Business Connect, Apple Business Essentials, and Apple Business Manager, and Apple says existing Business Connect data — claimed locations, place card information, photos, account details — migrated automatically. If you claimed your place card under the old name, your account moved with it. This is the record behind Apple Maps, Siri, Spotlight, and Safari, which is a large share of the phones in this metro.

Bing — free, and it feeds Copilot. Microsoft’s support documentation points business owners to Bing Places for Business to claim or update the listing that appears in Bing Maps results. Since Copilot is grounded in Microsoft’s index, this is not an optional afterthought for anyone who cares what Copilot says about them.

Directory and industry listings — varies wildly. Some let you edit in five minutes. Some make you prove ownership. Some belong to a trade association that updates once a year. Handle the ones where you already appear first; chasing new ones is a different project.

Old news articles, PDFs, and defunct pages — effectively unfixable. You can email a webmaster and sometimes it works. Often the site is abandoned and nobody is home. Accept that, and spend the energy making the correct sources louder and more numerous instead.

Reviews — you can respond, not rewrite. You cannot delete a review for being outdated or unfair. You can reply, and a reply saying the policy changed in 2023 is itself text that gets read.

The AI answer itself — no correction channel, and no published timeline. This is the honest gap. Nobody publishes how quickly a corrected Business Profile shows up in a given assistant’s answer, and the mechanism differs by product: some answers are retrieved live at query time, some lean on cached data, some on training. Anyone who tells you it takes exactly two weeks is guessing. Re-run your check at 30 days and again at 90, log what changed, and build your own evidence.

Fix things in this order, because the wrong order wastes weeks

Sequence matters more than speed here. Work down this list:

1. Write the source-of-truth page first. Ten minutes. Everything else copies from it. Fix listings before this exists and you will hand three platforms three slightly different versions of your address.
2. Fix your own website. You control it, it’s instant, and it’s the page other platforms and crawlers check you against.
3. Kill the ghosts before you polish the good record. Old locations, duplicates, a previous owner’s profile. Doing this after you optimize the real profile is how people spend a month improving a listing that keeps getting confused with a dead one.
4. Then your Google Business Profile. Hours, categories, services, service area, and the description — and clear any Google-made updates showing in blue.
5. Then Apple and Bing. Two accounts, both free, and between them they cover iPhone answers and Copilot answers. Most owners have done neither.
6. Then the directories where you already exist. Work from the citation list your 15-minute check produced. You already know which ones an assistant is actually reading.
7. Then the stragglers. Old chamber pages, supplier locators, sponsor listings. Email them. Expect a mixed response and don’t let it stall the rest.
8. Then reviews. Reply to the ones describing something you have since changed, and say what changed and when.
9. Re-run the identical check at 30 days and 90 days. Same eight prompts, same three assistants. Keep the screenshots together.

The wasted-weeks failure is almost always the same one: fixing the visible thing first. The Business Profile is visible and satisfying to polish, so people start there, while an eight-year-old duplicate at a dead address quietly keeps voting against them. Ghosts first. Then polish.

If this feels like it needs a working session rather than an article, that’s fair — it’s the sort of thing the AI workshops are built around, with your own business on the screen. And if the question you keep circling is not what the assistants say about you but how they choose whom to name at all, that mechanism is its own subject, covered in how AI decides which local businesses to name. For the wider picture of what AI is and isn’t good for in a small local operation, start with AI for St. Louis.

What this looks like from the owner’s side of the counter

For most local service operators the fixed cost is the truck and the tech, and the unit that decides your month is a booked day. That’s why a wrong hours field is a different animal from a wrong description. A wrong description costs you part of one conversation. A Saturday marked closed when you’re open costs the whole day’s inbound calls, and here’s the part that stings: there is no missed-call log for the customer who never dialed. The phone simply doesn’t ring, the day looks slow, and you blame the weather.

The thing operators actually complain about to each other is paying for the same job twice — once to a lead platform, and again in the drive time when the address turned out to be wrong. The ones who run this tight keep a single written source of truth and update every platform from it on one afternoon, usually in January when the phone is quiet. Being listed accurately in more than one place is the unglamorous version of that habit. If you want your record on St Louis Near Me Directory to match the rest, you can add or update your listing.

Frequently asked questions

Can I trust Google AI mode?

Treat it as a fast draft, not a record. Google’s own Search help says AI responses may include mistakes, and states that AI Overviews can and will make mistakes. For anything checkable about your own business, verify against your Business Profile and your website before acting on it. The practical risk isn’t that it insults you. It’s that it is confidently wrong about your hours.

Can AI Search be trusted?

For summarizing low-stakes material, mostly. For checkable facts, verify. The European Broadcasting Union and the BBC published a study on 22 October 2025 covering more than 3,000 answers from ChatGPT, Copilot, Gemini, and Perplexity: 45% had at least one significant issue and 31% had serious sourcing problems. That research examined news coverage, not local business listings, but the underlying failure is the same.

Is Google AI as good as ChatGPT?

There is no reliable head-to-head test on local business facts, so treat any ranking you read as marketing until proven otherwise. In the European Broadcasting Union and BBC study published 22 October 2025, all four assistants tested showed significant issues, with Gemini worst on sourcing at 76%. The useful answer for an owner is to check all of them about yourself, because they draw on different sources.

Why does AI give incorrect information?

Because the answer is assembled rather than looked up. The assistant pulls fragments from an index, from structured listings, and from training data, then writes fluent prose over the top. When your hours are stale in one place and correct in another, nothing in that process knows which is current, so it picks. Smooth wording is not evidence of an accurate fact.

What should I do if an AI gives me a wrong answer?

Screenshot it with the date and the exact prompt visible, and note every source it cited. Then go fix those sources instead of arguing with the assistant. In Google, thumbs down on an AI Overview opens Report a problem, which is product feedback rather than a correction ticket. Re-run the identical prompt at 30 and 90 days and keep the before and after together.

Who is responsible if AI makes a mistake?

For a third-party assistant there is no settled general answer, and no company owes you a correction. For a chatbot on your own website the picture is clearer: in Moffatt v. Air Canada, 2024 BCCRT 149, decided 14 February 2024, a British Columbia tribunal held the airline liable for wrong information its own chatbot gave a customer. That is a Canadian tribunal rather than US law, but it is worth knowing before you add a bot to your site.

How to correct incorrect information on Google?

It depends who owns it. Your own Business Profile: sign in and edit it directly. A record you don’t own, such as an old location or a duplicate: use Suggest an edit in Google Maps, including Place closed or not here for a permanently closed record. Something on a third-party website that Google merely indexes: contact that website, because Google will not rewrite someone else’s page for you.

How to correct business information on Google?

Sign in to the Google Account linked to your profile, go to your Business Profile, select Edit profile, change the field, and save. Google says edits usually take up to 10 minutes to review but sometimes up to 30 days, and each shows as accepted, pending, or not approved. Rejected business information edits can be appealed. Changing a verified business name may require verifying again.

Can I turn off AI on Google?

Not really. Google’s Search help states that AI Overviews are a core Search feature, like knowledge panels, and that features cannot be turned off. You can select the Web filter after running a search, which displays only text-based links without AI Overviews. There is a separate Labs toggle for experimental AI in Search features, which is a different setting and does not remove AI Overviews. For what the newer conversational surface is and how it behaves, see Google AI Mode explained for local business.

How much does it cost to amend your business name?

Two separate bills. Updating the name on Google, Apple, Bing, and directory listings costs nothing. Changing the registered name does: Missouri charges $25 to file an Amendment of Articles of Organization for an LLC and $7 to register a fictitious name, while Illinois charges $50 for LLC Articles of Amendment, form LLC 5.25, with expedited service extra. Confirm the current fee schedule before filing.

About the Author: The St Louis Near Me Directory Team
Written by a dedicated team of St. Louis locals who live, work, and play right here in the St. Louis metro. Founder Lane Forman and team are committed to building the region’s most trusted directory by verifying listings and connecting local businesses with loyal customers across Missouri and Illinois.

Follow us:
Facebook LinkedIn X Pinterest YouTube

Does AI cite your expertise?

Run Your Free AI Audit

No credit card. No obligation.