AI Gets Your Business Hours Wrong. Nothing Tells You.

Quick read: SOCi's 2026 Local Visibility Index, based on an analysis of roughly 350,000 business locations, found that business profile information surfaced by ChatGPT was about 68% accurate and Perplexity about 68% accurate, compared to roughly 100% for Gemini, which pulls directly from Google Maps (Search Engine Land). Separately, a UK vendor audit covered by Search Engine Journal in July 2026 found at least one false fact — most often a wrong postcode — in answers about 64% of the high-street retailers it tested (Search Engine Journal). Neither AI answer engine flags these errors to the business it's describing.

Someone's standing outside a shop at 6:10 on a Tuesday, phone in hand, because ChatGPT told them it closed at 7. It closed at 6. They didn't check the website. They didn't call. They asked an AI assistant a simple question, got a confident, specific, wrong answer, and acted on it like it was fact — because it was phrased exactly like one.

That's not a hypothetical. Per the numbers above, it's closer to a coin flip's ugly cousin: roughly a third of the time, what ChatGPT or Perplexity says about a local business's basic details doesn't match reality. Not "vague" or "outdated by a year." Specific and wrong, delivered with the same flat confidence as when it's right.

What's actually happening here

Here's the part that trips people up: these systems don't know when they're guessing. This sits at the sharper, riskier edge of Answer Engine Optimization — getting an AI system to describe your business accurately is a different, harder job than getting it to find you at all. Ask ChatGPT if a business is open right now, and it isn't checking a live clock against a live database the way Google Maps does. It's assembling an answer from whatever mix of training data and retrieved web pages it has on hand — your website, a directory listing, a review site, sometimes a stale cached copy of any of those — and it fills in gaps with the most plausible-sounding answer rather than pausing to say "I'm not sure." Even when every one of those public pages agrees with each other, the model can still misread, misdate, or flatly invent a detail, because agreement between sources isn't the same problem as the model reading them correctly.

This is a different failure mode than a mismatched address on Yelp. A business can have perfectly consistent name-address-phone information everywhere and still get its hours, holiday closures, or service list stated wrong in an AI answer, because the model's job in that moment isn't "look up a fact" — it's "generate the most likely-sounding sentence," and those aren't the same task even when the output looks identical to a correct answer.

Why this matters more for a small, single-location business

A multi-location chain has a marketing team running audits and a support line to absorb the occasional confused customer. A single-location small business doesn't get a warning. Nobody emails you when ChatGPT tells someone you're closed on Sundays and you're not. There's no dashboard, no notification, no dip in a chart you'd think to check. The only "signal" is a slightly quieter Tuesday you'll never connect to the cause, because the customer who got turned away never called to tell you they tried.

I ran a dog grooming salon for years before any of this, and the thing that would have driven me up a wall is exactly that silence. A wrong hour on your own website is something you'd catch. A wrong hour inside somebody else's AI assistant, recited to a stranger who never lands on your site at all, isn't something you catch — it's something you'd have to go looking for on purpose, and most owners don't know that's a thing you now have to do.

What actually helps

None of this is fixed by hoping the model gets smarter. A few things genuinely move the odds:

• Put your hours in LocalBusiness schema using openingHoursSpecification, not a plain-text sentence. The structured version supports validFrom and validThrough fields for holiday hours and seasonal changes — the plain-text openingHours field doesn't, which is exactly where a lot of "wrong hours" answers originate once a holiday schedule quietly expires and nobody updates the sentence version.
• Keep your Google Business Profile hours current, especially around holidays. Gemini's near-perfect accuracy in the SOCi data isn't a coincidence — it's grounded directly in Google Maps, which pulls from your Business Profile. That's the one dataset you have direct, immediate control over.
• Test it yourself, periodically, by just asking. Open ChatGPT and Perplexity and ask what your hours are, whether you're open today, and what you offer. It costs nothing, and it's currently the only reliable way to catch a wrong answer before a customer does.
• Treat this as recurring, not a one-time fix. Hours change for a holiday, a slow season, a staffing gap — and every change is a fresh chance for a stale version to be the one an AI system happened to retrieve. The signals these systems lean on decay by default, which is why staying accurate is upkeep, not a project with an end date.

What won't help

• Assuming a "verified" badge on a directory fixes this. Verification confirms you control a listing. It doesn't stop a different model, reading a different page, from getting a detail wrong somewhere else.
• Writing your hours once and never touching the sentence version again. Plain-text hours with no expiration mechanism are exactly what goes stale first, especially around holidays.
• Blaming the customer for not calling first. They didn't get a wrong answer because they were careless. They got a wrong answer because a system built to sound confident did, and they had no way to know it might be guessing.
• Buying an AI-visibility monitoring subscription instead of fixing the underlying data. A dashboard that flags the error after the fact is worth less than data clean enough that the error doesn't happen in the first place — monitoring is useful, but it's a complement to accurate schema and profiles, not a substitute for them.

FAQ

Why does ChatGPT get my business hours wrong even though my website is correct?

Because ChatGPT isn't checking a live database against a live clock the way Google Maps does — it's generating an answer from a mix of training data and retrieved pages, and it can misread, misdate, or fill a gap with a plausible guess rather than flagging uncertainty. A correct website reduces the odds of a wrong answer; it doesn't guarantee one.

How often does AI actually get local business information wrong?

Per SOCi's 2026 Local Visibility Index, business profile information surfaced by ChatGPT and Perplexity was about 68% accurate, against roughly 100% for Gemini. A separate UK audit found at least one false fact in answers about 64% of the retailers it tested. These are two different studies with two different methods, but they land on the same conclusion: this isn't rare.

Does fixing my Google Business Profile fix this everywhere?

It helps most directly with Gemini and Google's own AI Overviews, since those draw on Google Maps data. It doesn't guarantee ChatGPT or Perplexity get it right, since those systems read a different mix of public web pages and don't check your Business Profile directly.

Is this the same problem as inconsistent business listings (NAP)?

No, though they're related. NAP inconsistency is about your name, address, and phone disagreeing across different public pages, which can get you excluded from an answer entirely. This is about a model stating a specific wrong fact — often hours — even when your listings agree with each other, because the error happens in how the model generates the answer, not in a conflict between sources.

Will schema markup alone stop this from happening?

It reduces the odds significantly, especially for holiday and seasonal hours, but it isn't a guarantee. Schema gives the model a clean, structured fact to draw from; it doesn't control which sources a given AI system retrieves or how it phrases what it finds.

How do I know if this is happening to my business right now?

Ask. Open ChatGPT and Perplexity yourself and ask what your hours are and whether you're open today. There's no notification system that tells you when an AI answer about your business is wrong — checking directly is currently the only reliable way to find out.

Is this worth worrying about for a small, single-location business?

More, not less, than for a big chain. A large brand can absorb a handful of confused customers. A single-location business runs on the walk-ins and calls those exact minutes generate, and there's no support team catching the ones who simply didn't show up.

Checking this by hand every time your hours change isn't most owners' idea of a good use of a Tuesday, which is part of what ongoing website Care is for — keeping the facts an AI system might repeat about your business actually current, instead of finding out they weren't from a customer who never called.