Two cities wide
The short asks how big China gets on camera. The answer is narrower than the country: two cities wide.
Where the rest of the footage went
The video withholds the full distribution. Here it is.
We queried our travel index for China’s food lane with eight separate intent queries — street food, must-try dishes, where locals eat, cheap eats, night markets, breakfast, the famous noodle dish, and food tours. After capping each video at three matching chunks so no single upload could dominate, that returned 127 videos, which join to 101 distinct creator channels.
| What the footage names | Creators |
|---|---|
| Shanghai | 37 |
| Beijing | 37 |
| Sichuan / Szechuan / mala | 32 |
| Hot pot | 24 |
| A street-food stall | 23 |
| Dumplings | 22 |
Shanghai and Beijing tie exactly, at 37 each. Out of 101 creators making China food content, roughly three in four never name a third city at all.
Why the tie matters more than either number
A tie between two cities is not two destinations — it is one decision made twice. Both are first-tier airport cities with direct long-haul service, so the itinerary is set by the flight search rather than by the food. The order of operations inverts: you book the airport, then you look for the dish.
Sichuan is the interesting third entry. At 32 creators it is close behind the two cities, but it is not a city — the regex matches the province, the cuisine and the word mala together. It is doing duty as a flavour, borrowed to stand in for everything west of the two airports.
What we are not claiming
Two dishes you would expect are below our reporting floor and we are not reporting them: Peking duck has six creators in this pull, and hand-pulled noodles six. A fifteen-title floor is what separates a finding from an anecdote, and six is an anecdote. They may well be everywhere on YouTube; they are not everywhere in this corpus, and those are different sentences.
Reproducing this
The counts are not hand-tallied. The pull is a stored query set against our
travel index with a fixed top_k of 150 and a three-chunks-per-video diversity
cap, and the city counts are word-boundary regex matches over the retrieved text.
Change the lane and the shortlist changes with it.
More destination research: the travel section.