Restaurant websites and AI search: menus, hours and bookings
Why AI search gets restaurant menus, hours and bookings wrong: PDF and image menus, booking iframes, Restaurant and Menu schema, and your Business Profile.
By Crawlito · 6 min read
In short
- A menu that exists only as an image or a scanned PDF gives crawlers little or no text to read; an HTML menu page with dishes and prices is the most reliable format.
- Opening hours, address and phone number should be plain text in the HTML of every relevant page, not only inside a map embed, an image or a JavaScript widget.
- Booking widgets usually load in iframes or scripts from the booking provider, so state how to book and link to the booking page in your own HTML.
- Google's LocalBusiness documentation requires name and address and recommends servesCuisine, menu, priceRange, telephone and openingHoursSpecification for restaurants.
- Google Business Profile has its own menu editor and menu link for food and drink businesses, and Google may transcribe menu data from your website, so keep the two consistent.
On this page
- What people ask, and where the answers usually live
- Menus: PDF, image or HTML
- Hours, address and phone number
- Reservation widgets and iframes
- Restaurant structured data
- Google Business Profile and your website
- A quick check for your restaurant site
- Frequently asked questions
- Check what crawlers see on your site
AI search tools answer restaurant questions such as "is it open on Sunday", "do they have vegan options" and "can I book online" from whatever they can read. On many restaurant websites, the answers sit in places crawlers handle badly: a menu uploaded as a photo or a scanned PDF, hours in an image, and bookings inside a third-party widget. The fix is to put the facts people ask about in plain HTML text, describe them with Restaurant markup, and keep your Google Business Profile in step with the site.
This guide goes through each part of a typical restaurant site, how to check it, and what to change.
What people ask, and where the answers usually live
| Question | Where it often lives | Readable by crawlers? |
|---|---|---|
| What is on the menu, and how much? | PDF, photo of the printed menu, or a script-driven menu widget | Often partly or not at all |
| When are you open? | Footer image, Instagram post, map embed | Often not |
| Where are you, and is there parking? | Embedded map | The map iframe content is not part of your page |
| Can I book? | Booking widget in an iframe or script | The widget itself usually not |
| Do you have gluten-free or vegan dishes? | Symbols on a menu image | Not as text |
| Are you open on public holidays? | A social media post | Not on your site at all |
None of these assistants document reading text inside images, and most AI crawlers are not documented as running JavaScript. Anything in the right-hand column that says "not" is invisible to them in practice.
Menus: PDF, image or HTML
Menu as a photo. A JPG of the printed menu has no machine-readable text. Some systems may try to extract text from images, but none of the major AI search operators document doing this for web pages. Treat image-only menus as unreadable. See text in images, canvas and video.
Menu as a PDF. Google lists PDF among the file types it can index, and a PDF exported from a design tool usually contains real text. A scanned PDF is an image inside a PDF wrapper and has no text layer. Even a good PDF is a separate document, often without a link back to the restaurant, opening hours or booking details. PDF documents and AI search covers how to check one.
Menu as HTML. A menu page with dish names, short descriptions, prices and dietary labels in text is the most reliable option. It is readable by every crawler, works well on phones, and can be updated without redesigning a file.
<section id="mains">
<h2>Mains</h2>
<ul>
<li>
<h3>Roast hake, brown shrimp butter</h3>
<p>Charred leeks, new potatoes. <span>Gluten free</span></p>
<p>£22</p>
</li>
</ul>
</section>If you want to keep a PDF for printing, keep it, and link to it from the HTML menu page.
Menus from ordering platforms
Many restaurants show a menu from an online ordering or delivery platform. If that menu loads by JavaScript from the provider, check whether any of it is in your page's HTML. Often the only thing a crawler sees is an empty container. A short HTML summary of the menu on your own page, with a link to order, solves that.
Hours, address and phone number
Put these in plain text, in the HTML, on the homepage and on a contact or "find us" page:
<address>
The Harbour Kitchen, 12 Quay Street, Falmouth TR11 3HH<br>
<a href="tel:+441326000000">01326 000000</a>
</address>
<p>Open Wednesday to Sunday, 12:00 to 15:00 and 18:00 to 22:00. Closed Monday and Tuesday.</p>Avoid these patterns as the only source of the information:
- Hours in a header or footer image
- An embedded Google Map as the only address
- Hours that only appear after a JavaScript widget loads
- Holiday changes that are only posted on social media
Seasonal and holiday changes are where AI answers most often go wrong, because the old hours stay on the site. Update them on your own pages and in your Business Profile at the same time. Holiday hours and location pages covers how to publish temporary changes.
Reservation widgets and iframes
Booking systems are usually embedded as an iframe or a script from the provider. The content of an iframe belongs to the provider's page, not yours, and a crawler reading your HTML sees at most the iframe tag. Iframes and embedded content explains how different crawlers treat them.
That is fine for the booking itself. What matters is that your own page says, in text, how booking works:
<h2>Book a table</h2>
<p>Book online for up to 8 people. For larger groups, call us on 01326 000000.
Walk-ins welcome at the bar.</p>
<p><a href="https://booking.example.com/harbour-kitchen">Book a table online</a></p>
<iframe src="https://booking.example.com/widget/harbour-kitchen" title="Online booking"></iframe>A plain link to the booking page also gives people a fallback if the widget fails to load.
Restaurant structured data
Google's LocalBusiness structured data documentation requires name and address, and recommends properties including geo, telephone, url, priceRange, openingHoursSpecification, servesCuisine and menu. It asks you to use the most specific type, so a restaurant uses Restaurant.
{
"@context": "https://schema.org",
"@type": "Restaurant",
"name": "The Harbour Kitchen",
"url": "https://www.harbourkitchen.example/",
"telephone": "+441326000000",
"priceRange": "££",
"servesCuisine": ["Seafood", "British"],
"menu": "https://www.harbourkitchen.example/menu",
"acceptsReservations": "https://booking.example.com/harbour-kitchen",
"address": {
"@type": "PostalAddress",
"streetAddress": "12 Quay Street",
"addressLocality": "Falmouth",
"postalCode": "TR11 3HH",
"addressCountry": "GB"
},
"geo": { "@type": "GeoCoordinates", "latitude": 50.15421, "longitude": -5.06927 },
"openingHoursSpecification": [{
"@type": "OpeningHoursSpecification",
"dayOfWeek": ["Wednesday", "Thursday", "Friday", "Saturday", "Sunday"],
"opens": "12:00",
"closes": "22:00"
}]
}A few details from the documentation and from schema.org:
- Coordinates need at least five decimal places, according to Google.
- Seasonal closures can be marked with
validFromandvalidThroughon anOpeningHoursSpecification. acceptsReservationsis a schema.org property ofFoodEstablishmentthat accepts a Boolean, text or a URL. Google's LocalBusiness page does not list it, so treat it as descriptive rather than as a Google feature.menuandhasMenu. Google's page usesmenuwith a URL. On schema.org,hasMenusupersedesmenuand can hold a full structuredMenuwithhasMenuSectionandhasMenuItem. Marking up every dish is optional; the HTML menu matters more.
The markup must match the page. If the hours in your JSON-LD differ from the hours in your footer, you are publishing two answers. Local business schema covers the type in more depth.
Google Business Profile and your website
For Google's own surfaces, your Business Profile is a separate source of truth with its own menu features. Google says the menu editor is available to food and drink businesses, that you can upload menu photos or a PDF, and that you can set a menu link under your profile's contact information. It also says Google may transcribe menu data from your business website onto your profile, and that menu updates can take 24 to 48 hours to appear on Maps and Search.
So the website and the profile feed each other. Keep the menu link pointing at your HTML menu page, keep hours identical in both places, and check the profile after you change the site.
How other AI assistants use business listings is not documented in detail, so do not rely on the profile alone. Your website is the source you fully control.
A quick check for your restaurant site
- Fetch your menu page and search for a dish name:
curl -s https://www.example.com/menu | grep -c "Roast hake". Zero means crawlers cannot read your menu. - Search the homepage HTML for your opening hours as text.
- Confirm the booking page link is a normal
<a href>in your HTML. - Paste the homepage into Google's Rich Results Test and check the Restaurant markup matches the visible page.
- Compare hours, phone and menu link with your Business Profile.
Frequently asked questions
Can ChatGPT read my restaurant's PDF menu?
OpenAI does not document how its crawlers handle PDFs. A text-based PDF is readable in principle, but an HTML menu page is the safer choice, and a scanned PDF has no text to read.
Is a photo of the menu enough?
No. An image has no machine-readable text, and none of the major AI search operators document reading text from images on web pages. Publish the menu as text.
Do I need Menu schema for every dish?
No. Google's LocalBusiness documentation recommends a menu URL. Full Menu markup is optional; the dishes and prices being in your HTML matters more.
Why does Google show different hours from my website?
Google Business Profile is a separate source. Update hours on your site and your profile at the same time, especially for holidays.
Check what crawlers see on your site
Restaurant sites often look complete in a browser while the facts people ask about are locked in images and widgets. To see the text AI crawlers receive from your menu or homepage, run the free check.
Filed under Strategy and audits. Topics: structured-data, json-ld, html, google, content