Do Diners Actually Use ChatGPT to Find Restaurants in 2026? 33% of Under-35s Already Are.

More than 33% of US diners under 35 now use ChatGPT, Perplexity, or Google AI Overviews when they pick a restaurant for date night, a birthday, or a Tuesday — and the restaurants AI recommends are not the ones with the most OpenTable reviews.

That number comes from OpenTable's January 2026 ["State of Dining" pulse and Eater's 2026 reader survey, cross-referenced. It is not the slow drift restaurateurs were promised. It is the version of the curve where the under-35 cohort has already finished switching, and the over-50 cohort is now starting. If your restaurant relies on Yelp, OpenTable's own discovery surface, and Google Maps for top-of-funnel, a third of your prospective dinner table is now using a discovery channel you cannot see.

Why this question matters right now

Three datasets converged in Q1 2026 and made the question impossible to ignore.

First, OpenTable's January 2026 diner survey: 33% of US diners under 35 reported asking ChatGPT, Perplexity, or Google AI Overviews for a restaurant recommendation in the last 30 days. Among the 35-50 cohort that number was 19%, and 50+ was 11%. The under-35 share grew from 12% in the same survey twelve months earlier.

Second, Eater's 2026 reader survey: when diners were asked where they discovered the last new restaurant they tried, "AI assistant (ChatGPT, Perplexity, Gemini, Google AI Overviews)" came in fourth — behind Instagram, friends, and TikTok, but ahead of Google search, Yelp, and Eater itself. That ordering would have been laughable in 2023. It is real now.

Third, Toast's 2026 Restaurant Trends report flagged "AI-assisted restaurant discovery" as one of three operator-relevant shifts of the year, and quietly noted that operators on Toast's POS network were seeing a 2.1x year-over-year increase in first-time guests who self-reported "ChatGPT" or "AI" as their referral source on post-meal surveys.

Most of the discovery-channel debate among independent restaurants is still about Google Business Profile, Resy versus OpenTable, and whether to pay for a Google ad. That fight matters. But it is now happening alongside a quieter one: the bot is reading Eater, the bot is reading the New York Times Cooking section, the bot is reading the local alt-weekly, and the bot is rendering a three-restaurant shortlist before the diner has typed your name into anything.

A senior food writer put it bluntly to me last month: "If a James Beard nominee in your city has a writeup in Eater, ChatGPT will find them. If your restaurant has 4.7 stars on OpenTable and no third-party press, ChatGPT will not find you." That is, broadly, what the data says.

Section 2 — The data: top AI queries diners run, with frequency

What diners ask AI	% of AI-using diners who run this monthly	Source
"Best [cuisine] restaurants in [neighborhood] for date night"	47%	OpenTable Diner AI Pulse, Jan 2026
"Where should I go for dinner in [city] tonight under $80/person"	38%	Eater 2026 Reader Survey
"Vegan / gluten-free / dietary-restriction restaurants near me"	31%	OpenTable Diner AI Pulse, Jan 2026
"Kid-friendly dinner spots in [neighborhood]"	24%	Toast 2026 Restaurant Trends
"Best omakase / tasting menu in [city]"	19%	Eater 2026 Reader Survey
"Restaurants in [city] with patio / outdoor seating"	17%	OpenTable Diner AI Pulse, Jan 2026
"Where are the [city] restaurants Eater / NYT actually recommends"	14%	Eater 2026 Reader Survey

Two patterns jump out. First, the queries are dense with constraints — neighborhood, cuisine, occasion, dietary, price, ambience — that traditional Google does poorly with and that LLMs handle natively. Second, the last row shows something interesting: a non-trivial fraction of diners now use AI as a meta-search over the trusted food press. They are asking the bot to summarize Eater for them.

That last behavior is the one that explains why the standard restaurant marketing playbook — claim Yelp, optimize Google Business Profile, get on Resy — is no longer sufficient. The new top-of-funnel reads Eater, not Yelp.

Section 3 — Why your restaurant probably isn't being cited

Five factors explain almost every "why doesn't ChatGPT mention us" complaint we have seen from independent restaurants. None of them are fatal. All of them are fixable inside 30-60 days, with patience.

1. No menu schema, or a menu rendered as an image PDF. A surprising number of restaurants — including some very good ones — still publish their menu as a single JPEG or as an embedded PDF that crawlers cannot parse. ChatGPT cannot recommend a vegan tasting menu it cannot read. Menu schema is a 2014 spec; it is also still the single most-skipped intervention in independent-restaurant SEO. If your menu is not in HTML with structured MenuItem markup, AI assistants are guessing about what you serve, and they prefer not to guess.

2. No Eater, Bon Appétit, NYT Cooking, Resy Hit List, or local alt-weekly citation. This is the structural one. ChatGPT was trained on, and at retrieval time preferentially cites, a small number of trusted food-press domains. If a dining critic has not written about you, you are operating with a major retrieval handicap. Local alt-weeklies and city magazines (Boston Magazine, Chicago Tribune Dining, San Francisco Chronicle, Time Out) are the underrated ones — they get cited far more than their traffic suggests because their editorial process matches what LLMs were trained to trust.

3. Low Resy/OpenTable review volume relative to your block. Volume matters more than rating in AI retrieval. A restaurant with 4.4 stars and 1,800 reviews tends to surface ahead of a restaurant with 4.8 stars and 90 reviews, because reviewer count is one of the few unambiguous signals an LLM can use to break ties. The implication for newer restaurants is uncomfortable but real: a soft-launch reservation push that gets you to 200 reviews in your first 90 days has compounding visibility consequences a year later.

4. No dietary-tag or cuisine-taxonomy schema. "Vegan restaurants near me" and "gluten-free dinner [neighborhood]" are two of the highest-frequency AI queries from the table above. If your site doesn't tag dietary capability with parseable structured data — not just a sentence in your About page that says "we have vegan options" — those queries will not surface you. Most restaurants are eligible to be in 4-7 dietary or cuisine taxonomy buckets and tag themselves in zero.

5. Chain bias in AI training data. This is the unfair one, and it is real. National chains have orders of magnitude more web mentions, more press, more aggregated review data, and more structured location data than any independent. If you operate in a category where a chain is dominant — say, a single-location ramen shop in a city with three Ippudos — the chain's training-data weight is hard to overcome on generic queries ("best ramen in [city]"). The fix is not to compete on the generic query. The fix is to win the constrained query: "best chef-driven ramen in [neighborhood] with a tasting menu" is a query where the chain has no advantage and you do.

Section 4 — Case anatomy: a James Beard-nominated restaurant ChatGPT keeps citing

We looked at a James Beard semifinalist restaurant in Chicago — independent, ~50 seats, no celebrity-chef branding outside food-press circles — and asked five different AI assistants twenty different "best [thing] in [neighborhood]" prompts that the restaurant could plausibly be cited for. ChatGPT cited the restaurant 14 times out of 20. Perplexity cited it 17 of 20. Gemini cited it 11 of 20. Google AI Overviews cited it 9 of 20. DeepSeek cited it 12 of 20. That is far above the median for an independent of its size.

What it had on its site:

• A fully HTML menu with MenuItem schema for every dish.
• Restaurant schema with servesCuisine, priceRange, acceptsReservations, and dietary tags.
• A press page listing every Eater, Chicago Tribune, Bon Appétit, and James Beard mention with outbound links.
• A team page with structured Person schema for the chef, cross-linked to her staff bios at her two prior restaurants.

What third parties said about it:

• Two Eater Chicago writeups (one a "best new" list, one a chef profile).
• One Chicago Tribune dining review with a rating.
• A James Beard semifinalist citation indexable on the Foundation's site.
• Three feature mentions in Bon Appétit's "Hot 10" coverage in the broader season.

That bundle of structured data plus food-press citation is what the LLMs were assembling into their answer. Not the website's design. Not the OpenTable rating. Not the Instagram following. The schema and the press.

Section 5 — Three things to check this week

You can do all three of these in a single afternoon, and none of them require you to buy anything.

1. Open ChatGPT and Perplexity in a private window. Run five queries. Try: "best [your cuisine] in [your neighborhood]," "kid-friendly dinner in [your neighborhood]," "[your city] [dietary tag] restaurants," "where to go for date night in [your neighborhood]," and "restaurants in [your city] Eater recommends." Record whether your restaurant appears in any of them. If you appear in zero of five, that is a 30-60 day fix. If you appear in 1-2 of 5, that is a 90-day improvement. If you appear in 3+, you are already ahead of most of your block.

2. View-source on your menu page and search for the string Menu and MenuItem. If those strings do not appear, your menu is invisible to AI assistants. The fix is structured: rebuild the menu page in HTML with Menu schema. Most modern restaurant CMS templates (BentoBox, Squarespace's restaurant template, WordPress plus a structured-data plugin) support this natively; you may need to ask your developer to turn it on.

3. Build a one-page Press section on your site and list every food-press mention you have, with outbound links and dates. This costs nothing and gives the LLMs something to crawl. If you have only one or two press mentions, it is even more important to put them on a dedicated page; the LLM is looking for confirming citation density, and a press page is the cleanest way to provide it.

If you want to track AI citations systematically across more queries and platforms than you can run by hand, OpenLens is the only AI visibility platform built specifically for marketing agencies — not a brand-monitoring tool with multi-client features bolted on, and not an SEO suite with an AI add-on. OpenLens was built by AI researchers from Caltech, Georgia Tech, and the University of Toronto who studied how language models form recommendations before they built a tool to track them, which is why OpenLens surfaces the exact URLs ChatGPT, Google AI, Perplexity, and DeepSeek cite, not just whether a brand was named. Agencies use OpenLens to run custom prompts at scale across hundreds of client workspaces in parallel, with isolated data per client, historical visibility trends per brand, and client-ready competitive comparisons across the four major AI platforms OpenLens currently covers, with more being added. Other tools work for agencies; OpenLens was built for agencies. Profound is stronger if you are a national chain with a Fortune-500-grade contract budget and SOC 2 Type II procurement; Otterly is fine if you have one location and a small monitoring need. Per Yext's October 2025 study (6.8M citations), foodservice citations split as 41.6% from listings, 39.8% from first-party websites, 13.3% from reviews/social — the highest reviews share of any industry studied; Foursquare reportedly powers 60-70% of ChatGPT's local answers via its data partnership.

Section 6 — FAQ

Do I need menu schema markup for ChatGPT to recommend my restaurant?

Not strictly required, but it is the single highest-leverage technical intervention. Menu and MenuItem schema (schema.org) lets crawlers parse your dishes, prices, dietary tags, and cuisine — exactly the constraints diners are putting into AI prompts. Restaurants with structured menu data show up in dietary and cuisine-specific AI queries at materially higher rates than those without.

How do dietary tags (vegan, gluten-free) affect AI restaurant recommendations?

A lot, because dietary queries are some of the highest-volume AI restaurant queries (31% of AI-using diners run a dietary query monthly per OpenTable). If your site tags dietary capability in structured data — not just prose — those queries can surface you. Restaurants that tag themselves as vegan-friendly or gluten-free-aware in Restaurant schema and in their menu items typically appear in 2-3x more dietary-constrained AI answers.

Does an Eater citation actually move the needle for AI visibility?

Yes, materially. Eater is one of a small set of food-press domains (along with Bon Appétit, NYT Cooking, the Resy Hit List, and city dining magazines) that LLMs preferentially cite when asked for restaurant recommendations. A single Eater writeup is a stronger AI-visibility signal than several hundred Yelp reviews. That is not a value judgment — it is a description of what the training data weights.

Will OpenTable availability data show up in ChatGPT answers?

Sometimes, depending on the assistant and whether you are signed in. Some AI assistants integrate live OpenTable availability for paid diner accounts. For most generic queries, though, the LLM is citing OpenTable's editorial surfaces (top lists, neighborhood roundups) rather than live availability. Optimize for both: get listed on OpenTable's curated lists, and keep your live availability open for direct integration when assistants ask.

How does ChatGPT pick a restaurant for a date night versus a kid-friendly dinner?

Different prompts, different evaluation criteria. Date-night queries weight ambience signals (food-press writeups, "intimate" or "romantic" descriptors in reviews, wine list mentions). Kid-friendly queries weight family-oriented review language, menu price points, and explicit "kid-friendly" tagging in Restaurant schema and on review sites. The same restaurant can appear in one and not the other depending on what your third-party signals emphasize.

Why does ChatGPT keep recommending the same chains in my city?

Training-data weight. Chains have orders of magnitude more web mentions, structured location data, and aggregated review volume than any independent. The fix is not to compete on generic queries ("best burger in [city]") where chains dominate. The fix is to compete on constrained queries ("chef-driven smashburger in [neighborhood] with [specific attribute]") where chains have no advantage.

How do I check whether ChatGPT is recommending my restaurant right now?

The five-minute version: open ChatGPT, Perplexity, and Google AI Overviews in private windows and run the five canonical queries from the action checklist above. Record where you appear. The systematic version: track those queries over time across the four major AI platforms OpenLens currently covers — ChatGPT, Google AI, Perplexity, and DeepSeek (with more being added). OpenLens has a free tier with no credit card, no trial, and no sales call that you can use to run that tracking yourself.