Part 2 of "Who Decides Where You Eat"
The first part of this series followed one argument through four institutions: a tire company's road guide, a couple's mail-in survey, a review site, and a map. Each of them won for a while. Each is still standing in some form.
But that lineage leaves out the fastest-growing part of restaurant discovery, and the part most people under forty now actually use. Short video. Neighborhood accounts with four thousand followers and better instincts than any algorithm. Recommendation apps. Creators with audiences larger than any newspaper food section ever had.
This layer gets talked about constantly and examined rarely. It is usually filed under one of two headlines — the death of criticism, or the future of everything — and both are slogans rather than descriptions. Neither survives much contact with what people are actually doing.
So it is worth slowing down and looking at it properly. What is this new thing good at? What did it arrive to satisfy that nothing else was satisfying? And how much of the old apparatus has it genuinely displaced, as opposed to simply joined?
Start with what it looks like in practice.
It is 6:40 on a Tuesday evening and a video is playing on a phone. Someone's hands are pulling a piece of bread apart, slowly, and the cheese inside is doing the thing that cheese does when a camera is close enough and the light is right. There is a location tag. There is a caption with too many exclamation points.
Nobody searched for this. No question was asked. But by the end of the eleven seconds, a decision has been made about dinner that a hundred and twenty-five years of guides, surveys, star ratings and review platforms had no mechanism to produce.
That is the behavior worth explaining. Not a review, not a search, not a decision anyone consciously made — and yet dinner is settled.
What is a video actually telling you?
Consider what the older tools were designed to produce. A star, a score out of thirty, an aggregate of four thousand ratings — all of these are attempts to answer one question with increasing accuracy: is this restaurant good?
An eleven-second video does not answer that question at all. It cannot. There is no methodology in it, no sample size, no comparison.
What it answers instead is: do I want this, right now?
Those are genuinely different questions, and for most of the history of eating out, only the first one had any infrastructure behind it. You could find out that a place was well regarded. Finding out whether it was the thing you were in the mood for on a cold Tuesday when you didn't want to think — that was left entirely to you.
Appetite is specific and situational in a way that ratings are structurally unable to capture. A restaurant with an excellent aggregate score is telling you what a large number of strangers concluded on average. It is not telling you that the bread is the good kind of chewy, or that the portions look like the portions you want tonight, or that the room appears to be the temperature of room you can tolerate after this particular day.
Video carries that information almost accidentally. It was never designed as a review format. It just turns out that watching someone eat something transmits a kind of data that a number cannot hold.
Why couldn't the old systems do this?
Because compression was the entire point of the old systems, and this information does not survive compression.
Every ranking mechanism in this history worked by taking a large amount of subjective experience and squeezing it into something portable — a symbol, a number, a short list. That was the achievement. A star travels. A thirty-point score fits in a book you can carry. An aggregate rating loads instantly.
But compression discards whatever does not generalize, and what you specifically want for dinner is exactly the sort of thing that does not generalize. It got squeezed out at every stage, not through negligence but by design.
The feed reversed the trade. It gave up portability and comparability — you cannot rank a video against another video — and in exchange it kept the texture. It is a worse instrument for determining quality and a far better one for determining fit.
Which explains something that otherwise looks irrational: people continue using ratings and videos side by side, for the same meal, without any sense of contradiction. They are not redundant. One is answering should I trust this place and the other is answering am I hungry for it.
How much has actually changed?
Less than the coverage suggests, and this is worth stating plainly because the exaggerated version has been repeated for years.
The widely circulated claim is that roughly forty percent of young people now look for lunch on social platforms rather than Google. That figure originated with remarks by a Google executive in 2022 and has been quoted ever since as evidence of collapse.
The more careful data tells a different story. In a survey of American consumers, among eighteen-to-twenty-four-year-olds, sixty-seven percent reported searching on Instagram, sixty-two percent on TikTok, and sixty-one percent on Google. The figures add up to well over a hundred percent because most people use all three. Broader surveys find that even among the youngest users, the overwhelming majority still use conventional search regularly.
That is not substitution. It is accumulation.
And the pattern holds when you look at what people do in sequence rather than in aggregate. A video creates the want. Then, very often, the map gets opened anyway — to check whether the place is open, how far it is, whether there is a wait, whether the rating is catastrophic. Discovery moved. Verification stayed put.
So why does nothing in this business ever die?
This might be the most useful observation in the whole series, and it becomes visible only once you stop expecting a winner.
Each new layer did not replace the previous one. It took over one specific job and left the rest behind.
The professional guides kept curation — the rationed list, the small number of names that mean something precisely because there are only a few of them. Aggregate ratings kept verification: the fast, boring, essential check that a place is not a disaster. Maps kept logistics, which turns out to be most of what people actually need most of the time. And the feed took desire, which nobody had ever formally owned.
None of those jobs disappeared. They just stopped being performed by the same object.
That is why the old institutions look simultaneously diminished and permanent. Their share of attention keeps shrinking while their function stays essential. A guide that names a handful of restaurants in a city still confers something no algorithm can manufacture, for the simple reason that scarcity cannot be automated. An aggregate rating is still the fastest way to avoid a mistake. Neither is glamorous anymore. Both are load-bearing.
What does the newest version look like?
The interesting development right now is not any single platform. It is recombination — formats borrowing each other's methods.
The clearest example is a style of review that has become enormous over the past few years: a person eating takeout in a parked car, delivering a score out of ten. What makes it work is that the food was ordered anonymously and paid for personally, which is the oldest verification method in the business, running inside the newest distribution format. Old rigor, new delivery. The result is something that answers both questions at once — is this good, and is this for me — which nothing else on this list has ever done simultaneously.
The category is also visibly institutionalizing, which is what a trend does when it stops being a trend. There are rate cards now. There are agencies. American regulators updated their endorsement rules in 2023 to explicitly cover this kind of content, including free meals as a disclosable connection. None of that is a scandal; it is a signature. It is what happened to newspapers, and then to television, and then to blogs. A recommendation channel becomes an industry, and industries acquire paperwork.
Meanwhile another layer is stacking on top of all of it. Summarized answers now arrive before any of these sources are consulted, assembled from all of them at once. Whether that becomes its own layer or a lens over the others is genuinely not clear yet. It is early.
So where does this leave a hungry person on a Tuesday?
With more layers than ever and no clear instruction about which one to use.
That sounds like a failure of the system, and it might just be an accurate description of what abundance feels like from the inside. There is no shortage of information about restaurants. There has not been for twenty years. What there is instead is a stack of tools that each answer a slightly narrower question than the last: is it good, is it near, is it open, is it for me, is it right now.
Nothing in the current evidence suggests one of them is about to swallow the others. The stack just keeps growing, and each new layer arrives claiming to replace everything below it, and then settles in as one more thing people check.
But every one of those layers terminates in the same place. The video ends, the summary resolves, the rating is glanced at and accepted — and then there is a door somewhere, and a room, and a table, and the entirely unautomated business of getting yourself there.
Which raises the last question in this series, and the one that keeps sounding more absurd every year without ever quite going away.
If everything about a restaurant can now be found, watched, compared and summarized from a couch — does "the location" still matter where the restaurant actually is?
Part 3 concludes the series.
Notes on sources
Search-behavior figures: remarks by a Google executive reported in July 2022; the 2024 SOCi consumer survey reported by Forbes; Adobe Express consumer research on social search. FTC Endorsement Guides, revised June 2023 (16 CFR Part 255). Reporting on car-review creators and their methods: New York Times, NBC News, Houston Chronicle, and Washingtonian.
Comments
Post a Comment