AI Visibility for Restaurants: Menus, Hours, Prices and Reviews

Diners use AI assistants the way they used search: first to shortlist, then to check details. Say you run a hypothetical Thai restaurant in Denver. The prompts that matter look like this:

• "Best pad see ew near Capitol Hill, Denver"
• "Thai restaurants open after 10pm near me"
• "Where can I get gluten-free Thai food in Denver?"
• "Is [restaurant name] good for a group of 12, and do they take reservations?"
• "What are the vegan options at [restaurant name]?"
• "Does [restaurant name] deliver, and on which apps?"
• "How much is dinner for two at [restaurant name]?"

The list holds two kinds of prompt. Discovery prompts ("best X near me", dietary searches) decide whether you're named at all. Detail prompts (hours, reservations, prices) decide whether a diner who already found you gets a true answer.

You can't optimize for the exact words a diner types. OpenAI's help documentation says ChatGPT search typically rewrites a query into one or more targeted queries before it searches. Cover the facts behind these prompts, not every phrasing of them.

As of October 2026, each assistant draws on its own crawler or index, so a healthy result in one tells you nothing about another.

• Google AI Overviews and AI Mode. Google says a page must be indexed and eligible to show with a snippet to appear as a supporting link, with no additional technical requirements. AI Overviews often don't trigger, and the two surfaces may use different models and return different links.
• Gemini. A separate product. Google-Extended controls whether your content is used for Gemini training and grounding.
• ChatGPT. Search answers depend on OAI-SearchBot. The app decides per prompt whether to search. Without search it answers from training data with a cutoff: OpenAI lists Feb 16, 2026 for the GPT-5.6 family and Aug 31, 2025 for the Instant path.
• Perplexity. Builds its index with PerplexityBot.
• Claude. Uses Claude-SearchBot for search. In Claude.ai, web search is a user toggle.
• Microsoft Copilot. Grounds on Bing's index, so your Bing Webmaster Tools status matters on its own.
• Meta AI. Meta states that allowing Meta-WebIndexer in robots.txt helps it cite and link to your content in Meta AI's responses.

Hotels face the same split across engines and booking sites, covered in AI visibility for hotels. The next move is the same: test each engine on its own.

Restaurant facts change faster than most businesses' facts. That's where answers break. The high-risk list:

• Hours, including holiday and seasonal hours, and kitchen-close times that differ from door-close times.
• Menu items that rotated off months ago.
• Prices, especially after a reprice.
• Dietary claims: gluten-free, vegan, halal, nut handling. A wrong "yes" here is worse than no answer.
• Reservation policy: walk-in only, deposits, party-size limits, which booking platform.
• Delivery and takeout: which apps, which radius.
• Location status: a closed second site still described as open.

Three mechanisms produce these errors. Training data is frozen at a cutoff, so a model answering without search may describe your spring menu in autumn. Retrieval can pull a stale third-party listing instead of your site. And an engine can't quote a fact it can't read.

You control the second and third. Nobody controls the first this quarter. For a structured pass over every fact, use the local business fact check. For prices alone, run the pricing accuracy check.

Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. Independent re-testing in June 2026 found the same for OAI-SearchBot, PerplexityBot, Claude-SearchBot and others. Googlebot is the main exception because it renders pages. A menu widget that loads dishes and prices through JavaScript is blank to most AI crawlers.

Image menus fail differently. A photo of a printed menu puts the text in pixels, not in the HTML. A crawler reading the HTML gets a file name and, at best, alt text.

The fix is to publish the key facts as plain text in the page's HTML:

• Hours by day, with kitchen close and holiday exceptions.
• Signature dishes with current prices.
• Dietary options as specific sentences, for example "every curry can be made vegan; the kitchen uses peanuts."
• Reservation policy in one line.
• Delivery partners by name.

Stating a specific fact on a page that can be fetched gives the engine a sentence it can quote. It's the only on-page content change with a positive experimental signal, and it still doesn't guarantee a citation.

A two-minute test

Open the page source (not the rendered page) and search for a dish price. If it isn't there, most AI crawlers can't see it either.

When engines cite other sites instead of yours, being named accurately on those sites is the route into the answer. The evidence shows an association, not a guaranteed effect. For restaurants, check:

• Local listings and review platforms, including your Google Business Profile.
• Reservation platforms, where party-size rules and deposit policies live.
• Delivery apps, which often keep their own copy of your menu and prices.
• Local "best of" lists and food guides.
• Your social profiles, where an old hours post can contradict the website.

Look for contradictions, not volume. A delivery app showing last year's prices is a page an engine can retrieve. Fix the source, then ask the engine again.

Whether review counts move recommendations is a separate question, covered in do reviews affect AI recommendations. The short version: the published evidence is weak. Ranked "best of" lists behave much like affiliate review sites do in other categories.

Confirm crawlers can reach your pages before you rewrite them. robots.txt rules are per bot. GPTBot governs OpenAI training only. OAI-SearchBot controls ChatGPT search, and OpenAI says each setting is independent. A blanket AI block meant to stop training can silently remove you from ChatGPT search.

If you want to opt out of OpenAI training while staying eligible for AI search, the rules look like this:

User-agent: GPTBot
Disallow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Meta-WebIndexer
Allow: /

Then check the other gates:

• Firewall or CDN. OpenAI says eligibility needs robots.txt permission and a host or CDN that allows its published search bot IP addresses. A 403 to OAI-SearchBot is a silent failure. From 2026-09-15, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and existing free-tier customers, while Search stays allowed.
• Snippet controls. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers. Templates sometimes apply them by accident.
• Rendering, covered above.

The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away.

1. Week 1: baseline. Fill the prompt list above with your real dishes and neighborhood. Ask each engine while logged out. Note whether you're named, which facts are wrong, and whether the answer cited sources. An answer with no citations may come from training data. The free checker on this site gives a snapshot for up to three buyer questions.
2. Week 2: access. Check robots.txt per bot, firewall responses, snippet directives, and the view-source test on your menu and hours pages.
3. Week 3: your facts in HTML. Publish hours, prices, dietary options, reservation policy and delivery partners as plain text. Remove contradictions between pages.
4. Week 4: third parties. Update reservation, delivery and review listings. Re-run the week 1 prompts and compare.

After that, repeat the cycle with the quarterly AI visibility review, and after every menu or hours change. Answers that use live search can change within weeks. Answers from training data change only when a new model ships.

Does adding menu schema get my restaurant cited by AI? Not on current evidence. Google says structured data isn't required for generative AI search. Ahrefs tracked 1,885 pages that added schema and found no major uplift in citations on any platform, with AI Overview citations falling 4.6%. Those pages were already heavily cited, so the effect on less visible sites is untested. Existing markup does no harm. Put the menu facts in visible HTML text first.

Why does ChatGPT say we're open on days we're closed? There are three likely causes. The answer may come from training data frozen before your hours changed. It may come from a stale third-party listing the engine retrieved. Or your site may show hours only in an image or a JavaScript widget the crawler can't read. Check whether the answer cited sources, then compare each cited source against your current hours and fix the one that's wrong.

Should we allow GPTBot so ChatGPT recommends us? GPTBot governs OpenAI training only. ChatGPT search uses OAI-SearchBot, and OpenAI says each setting is independent of the others. Allowing or blocking GPTBot is a decision about whether your content is used for training. For ChatGPT search eligibility, allow OAI-SearchBot in robots.txt and confirm your firewall or CDN doesn't block it.

Can we rank higher in AI answers for 'best restaurant near me'? Nobody can promise that. Google reports only impressions for its AI features, with no position data, and no third-party tool has ranking ground truth for them. What you control is crawler access, the accuracy of facts on your own site, and the accuracy of your third-party listings. Whether an engine names you depends on its retrieval and its model, which you don't control.