By Vienna Hayden, Head of Marketing & Strategy, Dine Savvy
TL;DR: The restaurant discovery industry rebuilt its interfaces and left the foundation alone. Better AI, friend graphs, editorial voices, all still drawing from aggregated crowd behavior. What's missing is a system built on demonstrated taste rather than accumulated traffic.
- Beli, Zesty, OpenTable Concierge, and Eater's relaunch are genuine improvements to the interface.
- All of them still pull from the same pool: what a large number of people did, clicked, rated, or posted.
- Popularity tells a restaurant it's visible. It doesn't tell anyone whether a specific room fits a specific person on a specific night.
- Austin is showing what the gap costs, and what closing it would look like.
A Thursday at 6:45
You want to go out. You open four apps. One shows you "popular near you." One shows pins and star averages. One shows available tables. One shows a prompt field.
You type something like "somewhere interesting, good wine, not too loud."
You get five options. One looks fine. You book it.
It is fine. The food's okay, the wine list is standard, the room's a little loud. You wouldn't specifically go back.
You didn't do anything wrong. The app gave you what it was built to give you: a statistically reasonable match based on what worked for people who typed something similar.
That's the part nobody in this industry says out loud.
The Interface Changed. The Foundation Didn't.
The move away from star ratings has been real and fast.
Beli passed 75 million restaurant ratings by replacing anonymous critics with friend graphs. Your feed shows where your friends went and what they thought, on the theory that a recommendation from someone you know beats a 3.8 from a stranger.
DoorDash launched Zesty in December 2025, an AI app that pulls from Google Maps, TikTok, and DoorDash's own data to surface personalized suggestions. It's a pilot, live in the Bay Area and New York rather than nationally, but the direction is clear.
OpenTable's Concierge arrived in July 2025, embedded directly in restaurant profiles across sixty thousand restaurants. It answers menu questions and surfaces venue details. OpenTable's own research found that 54 percent of Americans research a restaurant before booking, and the ones who do spend around twenty-one minutes on it. Twenty-seven percent have abandoned a booking because the information was too hard to find. Concierge solves for that friction.
Eater went a different way in March 2026, relaunching its discovery app around editorial curation. Recommendations from actual food writers, chef tips from Eric Ripert and José Andrés. The bet is that readers trust experienced voices more than any algorithm.
These are all real improvements over what came before.
They also all draw from the same foundational pool: what many people have done, clicked, rated, or posted about.
Bottom line: The interface evolved. The dataset didn't.
What Beli Got Right, and Where It Stops
Beli is the most interesting case here because it came closest.
The insight was correct. Anonymous reviews are close to worthless, and a recommendation from someone whose taste you know is worth more. Beli built around that from day one, and the result is a feed shaped by your actual social graph rather than an engagement algorithm.
The execution is good. The architecture still has a limit.
Your friends are a community. They aren't curators of your specific taste. Your closest friend might love a ramen counter you'd find ordinary. Someone you follow might have entirely different criteria for what makes a room worth returning to. A friend graph is a better signal than a star average. It isn't the same as demonstrated taste alignment.
And Beli still aggregates. The more popular a place is among its users, the more it surfaces. Which means the platform still undervalues the restaurant on a quiet street whose chef is doing something remarkable, whose guest list should include exactly you, and whose score hasn't built enough momentum to reach your feed.
The Editorial Bet Is Partly Right
Eater's move toward curation is the closest to what we're building, and it's worth taking seriously.
The premise is that readers trust curators. Not algorithms, not strangers, not even friends with different taste. Actual editors with reputation on the line and real knowledge of the restaurant.
That premise is correct.
The constraint is scale and specificity. Editorial curation works beautifully for the top fifty restaurants in a city. It works less well for a two-month-old counter in East Austin that nobody has written up yet but that matches exactly what a segment of the city has been looking for.
Editorial still requires a restaurant to earn enough outside attention to enter the pipeline. Before that happens, the room is running on whatever anonymous demand the algorithm sends.
What's missing is curation that operates at neighborhood level, across the whole city, before a place has accumulated recognition.
Austin Is Where You Can See It
Austin's dining culture right now is a working demonstration of why the standard stack falls short.
The city's identity has shifted from destination dining to neighborhood dining. Siti opened in East Austin in July 2025 doing Southeast Asian food with real precision and landed in the Michelin Guide within months. It isn't the highest-rated restaurant on any app. It's the right restaurant for a specific set of Austin diners who know what that means.
The walk-in sushi counters tell the same story. Shokunin and Konbini are winning against the reservation-heavy omakase model, not because they're more popular on the apps but because they match a diner who wants serious sourcing without the production. That diner exists. Every existing platform underserves them.
The research supports it. Business Insider reported in late 2025 that 73 percent of Gen Z and millennial respondents visited a restaurant because of a social media review, and nearly 44 percent go to social first for recommendations. People are already routing around the review platforms. They're just landing wherever the feed drops them.
The Architecture We're Building
Dine Savvy isn't a smarter layer on crowd data.
It starts with curators. People with demonstrated taste and real knowledge of Austin, who build a follow graph of diners sharing their orientation. Not influencers. Not writers with mass audiences. Curators whose taste is specific enough to mean something.
Restaurants earn DS CERT through quality and identity alignment, not review volume or proximity. The process asks what the restaurant actually is and who it's built for.
When the system routes a verified guest to a DS CERT restaurant, it does it on taste alignment. The guest follows a curator they trust. The curator knows the room. The restaurant knows something real about who's walking in.
That isn't a feature. It's a different category.
The Question Underneath
Everyone in this space is asking how to make better recommendations. That's a good question and it has produced real progress.
The better question is what a restaurant actually needs from the demand side to stay open.
The answer isn't better UX or a smarter chatbot. It's signal. Specific, accurate, trust-based signal about who's coming and why.
The platforms that win the next era will be the ones that close the loop between a diner's taste, a curator's trust, and a restaurant's identity.
The interface matters. The architecture matters more, and it's still open for someone to get right.
Vienna Hayden is Head of Marketing & Strategy at Dine Savvy. She is building the platform in Austin, Texas.