Why algorithmic discovery beat the food critic in 2026

Why algorithmic discovery beat the food critic in 2026

The food critic did not lose because the writing got worse. The critic lost because in May 2026, the average diner makes a where-to-eat decision in roughly 6 minutes, mostly on a phone, mostly while already walking somewhere, and a 1,200-word review built around restrained adjectives and a star count cannot survive that compression. I have watched this happen in real time across our data at GeoTok, and the story is not that taste collapsed. The story is that the format broke.

When the Los Angeles Times quietly closed its restaurant-critic seat in Q1, and the San Francisco Chronicle moved Soleil Ho off the pure-criticism beat the year before, the trade press wrote elegies. I read those pieces with the feeling you get reading an obituary for a job nobody had recently been hired into anyway. The For You Page had already eaten the role. The newsrooms were closing a door on a room that nobody was sitting in.

Here is the underlying claim of this post, and I will defend it for the rest of the piece: algorithmic restaurant discovery beat the food critic in 2026 not because TikTok creators are better writers, but because the unit of dining recommendation has been re-cut. The critic published one piece a week. The algorithm publishes one piece every swipe. When the supply of recommendations becomes effectively infinite and matched to your specific 4:30 PM Tuesday craving in your specific neighborhood, the weekly column stops being a meaningful comparison.

The format mismatch is the whole story

A traditional review is a long, finished argument. It assumes you will sit down with a coffee and decide where to eat next Saturday based on a writer's careful seven-paragraph case. That model worked when there were three places worth writing about per neighborhood per year, and a critic's job was to filter the city's roughly 26,000 restaurants down to a handful of weekly winners.

The For You Page made a different bet. It assumed you have 35 seconds, you are 11 blocks from where you will eat, and what you actually need is a moving visual of a dish you have not yet decided you want. @keith_lee built a 16-million-follower audience on TikTok using exactly that grammar: he points the camera at the food, eats it on screen, and gives you a numeric score in the last 4 seconds. The format does in 90 seconds what a critic does in 1,200 words, and crucially, it does it for the specific dish — not the restaurant as an abstract entity.

I want to be careful here because this is where the conversation usually slides into a cheap "kids these days don't read" argument. That is not what is happening. Gen Z reads plenty. What they don't do is read a review in order to decide where to spend 28 dollars on dinner tonight. They watch the place. They watch the dish arrive. They watch someone's face when they bite it. That is a different epistemology, not a worse one.

The other thing the FYP got right: it understood that the dining decision is geographic and time-bound. A New York Times review of a Brooklyn restaurant runs in the print edition that lands on doorsteps in 11 states. A TikTok of the same restaurant, posted by @foodbabyny, gets pushed to viewers within roughly 8 miles of the location, because the algorithm has learned that food content converts on proximity. The critic was, in a sense, broadcasting. The algorithm is narrowcasting per request.

The takeaway: the For You Page did not out-write the critic. It out-formatted the critic. The medium absorbed the dining decision before the words ever had a chance.

Trust got rebuilt around faces, not bylines

The second thing that broke in the old model is the trust mechanism. A newspaper critic earned trust through institution and tenure — you trusted Pete Wells partly because he wrote in the Times and partly because he had been doing it for over a decade. That kind of trust is slow to build and expensive to maintain, and it requires the institution to keep paying for it.

The TikTok creator earned trust on a faster and weirder schedule. @kalemebymyname does a recurring "would I go back" segment that closes most of her videos, and her audience knows within 12 seconds of a video starting whether she is going to recommend it. She is, in effect, publishing roughly 5 reviews a week to about 1.4 million followers, with the same yes/no verdict structure a critic might use, except she is doing it in a parasocial voice the audience already knows how to parse.

"I would not come back for that price. I would come back for half that price."

That is a 12-word verdict, delivered with the receipts visible on screen. A critic would need three paragraphs to set up the same conclusion, and the conclusion would be hedged. The hedging used to be a feature — it signaled rigor. In the new context it reads as evasion.

There is also a structural reason creators won the trust game in 2026. Their video output is unfalsifiable in a way critic prose is not. When @markiethedoer says the line at a place is 40 minutes long, you see the 40-minute line. When a critic writes that a restaurant has "energy," you have to take their word for it. The video format collapses the gap between claim and evidence to zero frames.

I want to flag the obvious counter, because I think about it constantly: creators are paid. Some are paid by the restaurants directly. The disclosure rules are loose, the FTC enforcement is uneven, and there are absolutely creators who have posted glowing content about restaurants that comped their entire meal without saying so. I am not pretending that does not happen. But the critic system had its own pathologies — long-running grudges, reviewers who hadn't paid for their own dinner in 19 years, restaurants that closed before the review ran. The choice in 2026 is not between a pure system and a corrupted one. It is between two imperfect trust mechanisms, and one of them ships recommendations 200 times faster than the other.

The takeaway: trust in 2026 is verified by what the camera sees, not who the byline belongs to. Creators won partly because their evidence is visible in the same frame as their verdict.

What this means for how you actually pick a place

I run GeoTok, so I am not a neutral party. We build the index that sits between TikTok food content and the city you happen to be in tonight. I will tell you what we see, which is that the diner of May 2026 does not think of "restaurant discovery" as a single act. They do it in three layers, and the algorithm-vs-critic war was really only about the bottom layer.

The top layer is taste calibration — the slow, ambient process of figuring out what kind of food you like, who you trust, what cities matter to you. That layer is now run by the FYP. People learn what laab moo looks like from @pailinscookingchannel before they ever order it. They learn what omakase costs by watching @sushiandeats price-react to bills. The critic used to own this layer too, and they lost it cleanly. It is over. The middle layer is curation — taking the wash of FYP content and converting it into a shortlist for a specific city or neighborhood. This is the layer GeoTok plays in, alongside saved-folders inside the TikTok app itself, alongside Beli for the people who want a sortable database, alongside the better Substacks that hand-curate. Critics still partly own this layer, especially the ones who pivoted to newsletter formats. The Infatuation does too, although they read more like a city guide than a critic now.

The bottom layer is the live decision — the 4:47 PM Tuesday choice of where to actually walk. Maps used to own this layer through star ratings. In 2026, star ratings are increasingly background noise. What people want is the dish in the video matched to a place that is currently open within a 9-minute walk. That layer is almost entirely algorithmic, and the critic was never really competing for it.

The mistake the dining press made was treating this as a single war. They wrote thinkpieces about whether TikTok creators were good critics. They mostly are not, in the way critics meant the word. But the war was not about who writes a better review. It was about which system delivers the right dish to the right person in the right neighborhood at 4:47 PM, and that war the algorithm won years ago. The critic was, very late, defending a layer of the stack that nobody was attacking.

The takeaway: in 2026 you should expect your dining decision to be a layered one — slow taste-building from the FYP, hand-curated shortlists for your city, and a fast live match when you actually need to eat. The critic, when they still exist, is now a narrow specialist on the middle layer.

The honest position

Some of what was lost is genuinely lost. A great critic could tell you something about a kitchen that a 35-second video physically cannot transmit — the technique behind a sauce, the lineage of a dish, why a particular plate is the move-the-form thing the chef has been chasing for 7 years. That kind of writing was good for the culture and good for the chefs who deserved it. It is not coming back at the scale it used to exist, and pretending otherwise is sentimental.

But the format that replaced it is not worse. It is differently good. It is more democratic, faster, more honest about what it is doing, and more accurate about the way a real person makes a real dining decision on a real Tuesday at 4:47 PM in May 2026.

If you want to feel less anxious about all of this, the move is to stop comparing the FYP to the critic and start using each for what it is actually built for. Use creators to calibrate your taste. Use a curated index — ours or someone else's — to convert that taste into a city-specific shortlist. And when you are 11 blocks from dinner and you want the dish you saw last Thursday, that is what we built GeoTok for: the saved-from-TikTok place, ready when you are, without you having to remember which creator's video it came from.

The critic's job did not disappear. It got unbundled, and the pieces went to different surfaces. May 2026 is just the first month enough people noticed for the trade press to write the obituary. The For You Page was already running the funeral.