Hours, address, menu: how contradictory data loses you the AI recommendation

Why AI treats a contradiction as a red flag, not a rounding error

Models like ChatGPT, Gemini or Perplexity pull their answers from many sources at once — your Google Business Profile, TheFork, Tripadvisor, menu portals, your own website. When those sources agree, the model trusts you. When they don't, it hedges, and a hedging AI recommends whoever's data holds together without contradiction. You don't get skipped over because of the food — you get skipped because the facts don't line up. Most restaurant owners miss this entirely, because they're thinking about the kitchen, not about their data.

That's the real shift from classic Google search. A guest used to click through five links and shrug off small inconsistencies themselves. Now the AI hands back one confident answer, and there's no room in that answer for '10 p.m. according to the website, 9 p.m. according to Google.' The model resolves the conflict by dropping you and naming a competitor whose information doesn't contradict itself. For you, that means consistency stops being a tidiness habit and becomes a direct revenue lever.

You feel this on questions like: which restaurant nearby is still open after 9 p.m. tonight and has a table for four? That's exactly where businesses split into two groups. List the same kitchen-closing time everywhere and you get named. List three different times across three platforms and you get dropped, because the AI won't gamble on sending someone to a locked door. It would rather protect the guest from a wasted trip than protect your revenue.

The three data points restaurants get wrong most often

Hours are the classic failure point. Few things in this business change as often — a day off shifts, the kitchen closes earlier than the bar, lunch breaks, holiday rules, seasonal patio hours. Every one of those details has to match everywhere it appears. In practice, the owner updates Google, a server enters something different on TheFork, and the website hasn't been touched since the last redesign. Now there are three different truths for one question, and the AI can't confidently pass any of them along.

The address looks trivial, but it isn't. A move to the next street over, a new building entrance, a listing with or without 'courtyard' or 'rear building' in the name, a typo in the house number — to an AI, 'Ristorante Bella, Marktplatz 3' and 'Bella Ristorante, Marktplatz 3a' can read as two different businesses. Reviews and signals then split across two ghost listings, and neither one is strong enough to get recommended. A clean, identical address — spelling included — is the baseline that lets every other signal roll up under one name.

The menu is the most underrated failure point. Do you offer a gluten-free pasta? A vegan menu, a weekday lunch special, weekend brunch, delivery? If that only lives on a chalkboard by the door, it doesn't exist as far as the AI is concerned. And if your website still shows last year's menu, the AI might recommend you for a dish you stopped making months ago — and the guest who shows up expecting it leaves a bad review.

A real example: the Friday night table that went to someone else

Picture a family restaurant open until 11 p.m. on summer Fridays. Google still shows the winter closing time of 9 p.m., because nobody switched it back in the fall. TheFork shows 10 p.m., pulled from its last reservation slot. The website doesn't list a time at all, just 'hot food served all day.' Three sources, three different answers — easy for a person to untangle, a red flag for an AI.

Now someone asks their AI assistant at 9:30 on a Friday for a family-friendly restaurant still serving food. This place doesn't come up. Not because it's closed — it's open — but because its data makes it look like it might be closed. The competitor whose 11 p.m. closing time is listed the same way everywhere gets the table for four instead. The revenue is gone, and the owner never finds out why the dining room was emptier than expected that night.

The frustrating part is that this restaurant didn't do anything wrong except neglect its data. The food was good, the service was warm, the prices were fair. None of that mattered, because the AI had already ruled it out before ever getting to the menu. That's exactly why keeping your data consistent isn't an IT task you can hand off and forget — it belongs on the same list as ordering groceries.

How to build one source of truth

The first move is a simple internal document that lives nowhere on the public internet: a master data sheet. It holds the exact business name, the exact address with correct spelling, the phone number, every opening and kitchen hour including days off and holiday rules, and your core offerings — vegan, gluten-free, lunch special, patio, dogs welcome. This one document is your source of truth. Every platform gets checked against it going forward, never the other way around. Without it, you're relying on memory each time you update something, and that's exactly how contradictions creep in.

From there, work through every platform once, carefully, and enter the identical values on each one. Prioritize by impact: Google Business Profile first, then reservation platforms like TheFork or OpenTable, then review sites like Tripadvisor and Yelp, then delivery apps, and your own website last. Watch the small stuff — special characters, abbreviated street names, inconsistent time formats. What looks like a minor detail to a person reads as a genuine discrepancy to a model.

Why your own website has to stay the anchor source

Platforms come and go, change their rules, and display your information however suits them. Your own website is the one source that's entirely yours to control at any moment. That's why it should be the anchor everything else gets checked against. When an AI system looks for an authoritative source in a moment of doubt, your domain is the natural candidate — which means the hours, address, and menu there have to be exactly right.

Structured markup does real work here. Machine-readable data using restaurant schema lets you specify hours, cuisine, price range, and address in a format search systems parse without ambiguity. This isn't decoration — it's the format machines trust most when reading your facts. A developer can set it up in a few hours, and after that you have a source the AI has a hard time misreading.

Keep the discipline going afterward. When something changes, update your website and your master data sheet first, then everywhere else follows. That keeps the anchor stable, so the AI never ends up trusting the one source that's out of date. A well-maintained website, in the GEO era, isn't just a digital business card anymore — it's the foundation your entire visibility rests on.

The holiday-hours trap

Restaurants live on exceptions, and exceptions are exactly what trips up an AI. Closed on Easter Sunday, lunch-only on Christmas Eve, shut down for a winter break in January, longer patio hours in summer. If those special rules only exist in your head, the AI falls back to your normal hours — and either sends a guest to a locked door, or misses that you're open on a holiday while the competition down the street is closed. Either way, it costs you: a bad review in one case, an empty table in the other.

The fix is keeping special hours current directly on the platforms. Google and the major reservation platforms both offer fields for holiday and seasonal hours. Enter them ahead of time, ideally for the whole year in one sitting. Pick a fixed date — January is a good one — to log every holiday and vacation period for the next twelve months. That single hour of work heads off dozens of frustrated guests and protects your recommendation rate on exactly the days that matter most.

How to check what the AI is actually saying about you

You don't have to guess whether your data is right — you can just check. Ask ChatGPT, Gemini, and Perplexity the same way a guest would. Try: 'Recommend a good Italian restaurant near me,' or something more direct: 'What time does Ristorante Bella on Marktplatz close, and do they have vegan options?' Compare the answers against your master data sheet. Every mismatch points to a source that's feeding the model bad information.

When you find a mistake, trace it backward. If the AI gives you the wrong closing time, check Google first, then the reservation platforms, then the website, until you find where the bad number originated. Often it's one stale listing poisoning the whole picture. Run this check every few weeks, and always right after you change your hours or your menu — that's how you confirm a fix actually took everywhere, not just where you made it.

Treat this the way you'd treat a health inspection: not a one-time project, but a routine. Run a full pass twice a year, and spot-check after every hours change, and you'll have a presence that reads as unambiguous to every AI system that looks at it. In the age of generative answers, that clarity is the currency recommendations get paid in.