Why does ChatGPT say my restaurant is closed when we're open?

Almost always it is a leftover 'permanently closed' or 'temporarily closed' flag on your Google Business Profile, usually left over from a past renovation, a rebrand, or a change of ownership. The AIs read that flag and repeat it. Open your profile, clear the closed status, and confirm your real day-by-day hours. It is the most damaging restaurant error and one of the fastest to fix.

Why does the AI recommend the place down the street instead of me?

Usually because that restaurant has more recent reviews, cleaner hours and menu data, or dietary and reservation tags filled in that you left blank. Between two similar spots the AI leans toward the one with the fresher, fuller picture. A LocalFox report shows you which competitors the AIs name in your place and the specific reasons they give, so you know what to close the gap on.

Do menu items in my reviews really matter for AI?

Yes, more than for most trades. The AIs read review text and pull out specific dishes. When reviewers keep naming a dish, that dish becomes a hook the AI can match to a craving question like 'best ramen near me'. You cannot write the reviews, but you can nudge happy diners to mention what they loved by name, and you can repeat the dish when you reply.

Can I pay to get my restaurant into AI recommendations?

No. There is no ad slot inside a ChatGPT or Gemini recommendation, and anyone selling guaranteed AI placement is selling nothing. What you can do is fix the sources the AIs read, your Google Business Profile, hours, menu, dietary tags and reviews, so you earn the spot. LocalFox shows you what those AIs say about you now and gives you the exact fixes, but it cannot and does not promise placement.

How to get your restaurant recommended by ChatGPT and AI

It is 8pm on a Friday. Someone two blocks from your door opens ChatGPT and types "good dinner spot near me that's still open." The AI thinks for a second and names three restaurants. If yours is one of them, you might get a table you never knew was up for grabs. If it isn't, that diner walks into a competitor and never knew you existed. This is happening every night now, on ChatGPT, on Gemini, in Google's AI Overviews, and the answer is not random. The AI is reading a handful of sources about your restaurant and repeating what it finds. This playbook is about what it reads, what it gets wrong, and the few things that actually get you into that short answer.

The questions diners actually ask AI

People do not ask AI the way they type into a search box. They ask the way they would ask a friend who knows the area. The wording is conversational, and it almost always carries a condition: a time, a craving, a group, a dietary need. The AI then matches those conditions against what it knows about each restaurant and returns the ones that fit.

If your restaurant fits the condition but the AI does not know it, you lose the table. A late-night kitchen that never tells anyone it is open until 1am will not get surfaced for the 'where to eat late tonight' question, no matter how good the food is.

"Best restaurant near me" and "where should I eat in [neighborhood]"
"Good dinner spot with kids" or "family-friendly restaurant that's not too loud"
"Where to eat late tonight" and "what's open right now near me"
"Restaurants with vegan options in [city]" and the same for gluten-free, halal, or dairy-free
Craving-driven asks: "best ramen near me", "where to get good tacos tonight", "a nice spot for a birthday dinner"
"Restaurant that takes reservations for 6" and "somewhere I can walk in without a wait"

What AI reads about a restaurant, and which fields matter most

AI assistants do not hold a private opinion about your food. When a diner asks, they read live sources and summarize. For restaurants the most-read source by far is your Google Business Profile, and Google's own AI Overviews lean heavily on it for local results. If that profile is thin or stale, every AI inherits the same gaps.

Not all profile fields carry equal weight for restaurants. A general local business can get by with name, address and hours. A restaurant lives or dies on a few extras the AI checks specifically to answer dining questions:

Current hours, broken out by day, including late-night close times and holiday hours. This is the single field the AI checks most for the 'open right now' and 'late tonight' questions.
Your menu, linked and readable as text, not locked inside a PDF or a photo. The AI reads dish names to answer cravings like 'best pad thai near me'.
Cuisine and the right primary category. 'Mexican restaurant' is not the same as 'Tex-Mex restaurant', and the AI uses this to decide which craving questions you belong in.
Dietary tags and attributes: 'vegan options', 'vegetarian options', 'gluten-free options'. These are the exact words the AI matches for dietary questions, so if the box is unchecked, you are invisible for that diner.
Reservation and seating signals: whether you take reservations, accept walk-ins, have outdoor seating, or are good for groups. The AI pulls these straight into 'somewhere I can book for 6' answers.
Recent, real photos of the food and the room. Photos help the AI and the diner trust that the place is open and the menu is current.

The wrong facts that hurt a restaurant most

When your sources disagree or go stale, the AI does not flag uncertainty. It states the wrong fact with full confidence, and the diner believes it. For restaurants, a handful of errors do real damage because they directly answer the question the diner asked.

The worst one is a 'permanently closed' flag left over after a renovation, a rebrand, or a change of ownership. It happens constantly. You reopen under a new name or after a refit, the old listing still says closed, and the AI tells anyone who asks that you shut down. You can be packed on a Saturday while ChatGPT is sending people elsewhere because it thinks your doors are locked.

Stale hours are the next worst. If the AI thinks you close at 9pm when you now serve until midnight, you lose every 'where to eat late' query in your area. The same goes the other way: if it shows you open when you have cut back hours, a diner shows up to a dark window and leaves a bad review. Outdated menu items and prices cause the smaller, daily version of this: a diner asks for a dish you stopped serving, or quotes a price you raised a year ago.

A leftover 'permanently closed' or 'temporarily closed' flag from a past renovation or rebrand
Hours that do not match what you actually serve, especially late-night and holiday hours
An old phone number or address that no longer matches your website
Menu items the AI still lists that you removed, or prices that are out of date
A wrong cuisine or category that drops you out of the craving questions you should win

Reviews, and the review themes AI surfaces for restaurants

Between two similar restaurants, the AI almost always names the one with more reviews and a higher, fresher rating. Review count and recency are the clearest signal that a place is real and busy. Thirty reviews from the last two months read as more alive than three hundred from 2021.

For restaurants there is a second thing going on that other trades do not get. The AI reads the text of your reviews and pulls out specific dishes and occasions. When reviewers keep naming a dish by name, that dish becomes a hook the AI can match to a craving. If 'the carbonara' shows up in twenty reviews, you start getting surfaced for 'best pasta near me'. If reviewers keep saying 'great with kids' or 'easy to get a table for a big group', you get pulled into those exact questions.

You cannot fake this and you should not try. What works is a steady habit: ask happy diners for a review before they leave, make it one tap, and reply to the ones you get. You do not get to write the reviews, but you can nudge what they talk about. A server who says 'glad you loved the short rib' plants the dish name that ends up in the review, which ends up in the AI's answer.

Specific dish names: "the brisket", "their margaritas", "the tasting menu". These feed craving questions.
Occasion fit: "good for a date", "great with kids", "perfect for a big group", "quick lunch"
Dietary mentions: "lots of vegan options", "they have a real gluten-free menu"
Service and timing notes: "no wait on a weeknight", "open late", "easy to book"

The three highest-leverage quick wins

Most of the value comes from a short list. If you do nothing else, do these, in this order.

First, kill any stale 'closed' flag and fix your hours. Open your Google Business Profile, confirm there is no leftover closed status from a past renovation or rebrand, and set your real hours day by day, including how late the kitchen runs and your holiday hours. This is the cheapest fix with the biggest payoff, because a 'closed' flag silently removes you from every recommendation.

Second, get your menu and dietary tags in as readable text. Link a real menu the AI can read, not a flat image or a PDF that hides the dish names. Then check every dietary and attribute box that honestly applies: vegan options, vegetarian, gluten-free, outdoor seating, reservations, good for groups. Each checked box is a question you become eligible to win.

Third, start a simple review habit aimed at dishes. Ask every happy table for a review and prompt them gently about what they enjoyed by name. Reply to reviews, and when you reply, repeat the dish: 'so glad you loved the duck ragu'. Over a few weeks this builds both the volume the AI rewards and the dish-level detail it reads to match cravings.

Make your own site easy to read

Your website is where the AI double-checks the facts it found elsewhere. Two things help most. Put your name, address, phone, hours and cuisine in plain text on the page, not buried in a header image, and keep your menu as real text rather than a single photo of a chalkboard. Then add LocalBusiness structured data, a small block of code that states your name, address, hours and cuisine in a format machines read without guessing.

If your site is one image-heavy splash page with the address only inside a graphic and the menu as a downloadable PDF, you are making the AI work to find basics, and a working AI that has to guess is an AI that gets it wrong. The same consistency rule applies everywhere: pick one exact version of your name, address and phone, and make your site, your Google profile and every directory match it letter for letter. Conflicting facts are the fastest way to make an AI hedge or name a competitor instead.

Check where you actually stand

After you fix the profile and the hours, find out whether it landed. Ask the AIs the questions a diner would, in your own city: 'best dinner near [your neighborhood]', 'where to eat late tonight in [city]', 'restaurants with vegan options in [city]'. Ask each one a few times, because answers move run to run. A single check tells you almost nothing. What tells you something is your mention rate across repeated runs: out of nine asks, how many named you, and which competitors came up instead.

This is the step nearly every owner skips. They fix the listing, then never confirm whether ChatGPT changed its answer. Measuring closes the loop and shows which fixes worked. It also catches the thing you cannot see from the kitchen: the exact wrong fact an AI is repeating about you, quoted word for word, and the competitor it keeps recommending in your place.

A LocalFox report does this part for you. You enter your restaurant name and city, and it runs the real diner questions across ChatGPT, Gemini, Perplexity and Google AI Overviews three times each, shows you your visibility score and the single biggest problem for free, then gives you the full picture: every wrong fact quoted as the AI said it, which competitors get recommended and why, and a copy-paste fix kit with review-request wording, a Google Business Profile description draft, and a LocalBusiness schema block. It is a one-time $39 report, no subscription and no card kept on file, and it includes one free re-scan within 60 days so you can confirm your fixes worked. There is no way to pay an AI to recommend you and nobody can promise placement, but you can see exactly what it says about you today and fix the inputs it reads.