The Stub That Already Looks Finished
A wiki stub for a small town usually has a name, a state, a population figure with a census year attached, maybe a county and an elevation. That's the whole page. It reads like the first paragraph of something, the setup before the real content arrives. It isn't. It's the whole transaction, and the transaction was never about the town.
A census doesn't produce a description of a place. It produces a count of the people in it, on a specific day, for a specific bureaucratic purpose: apportionment, funding formulas, congressional districts. The number gets attached to the place because the place is the unit the counting happened in, but the number itself belongs to the census, not the town. When that number lands on a wiki page with nothing else around it, what you're looking at is a complete, faithful record of an act of counting. It says: this many people were tallied here in this year. That sentence is finished. It doesn't need a school district or a founding date or a Main Street to be true.
The trouble is that it doesn't look like a finished record of a count. It looks like an unfinished record of a place. Same shape, wrong reading. A stub has a title, an infobox, a citation, the visual furniture of every other article on the site. Nothing about the rendering distinguishes "this is all there is to say" from "this is all anyone has said so far." The page has already done the thing pages are for: it exists, it's indexed, it answers the query. Once a container looks occupied, the eye slides past it. Nobody stops at a filled box to ask whether it's actually full.
Compare that to a place with no article at all. That's a gap you can see, because the search returns nothing, or a red link sits there daring someone to click it. Absence has a shape. A census stub has no shape you can point to, which is worse for the town, not better. The place that got fully skipped is at least legibly missing. The place that got a stub is marked as handled, filed, done, and it will sit at "population: 4,213 (2020)" for a decade because there's no visible wound to draw an editor's attention. Disappearance by omission gets noticed eventually. Disappearance by a satisfied-looking record does not.
I don't think this is unique to Wikipedia, it's just unusually visible there because the format makes the mismatch legible. Any system where a record's existence gets read as a proxy for the record's completeness has the same failure mode. A ticket with a status of "logged" looks handled next to a ticket that's actually still open in every way that matters. A changelog entry that says "fixed" tells you a commit landed, not that the bug is gone. In each case the small true thing that got recorded (a count, a status flip, a diff) borrows the visual authority of the big true thing it resembles, and the resemblance is what stops anyone from doing the rest of the work.
The fix isn't "write longer stubs," because most census-stub towns don't have more to say, or nobody who knows the town has bothered to say it yet, and inflating the page with adjacent facts just moves the disguise one layer deeper. The fix is noticing that a record answering one narrow question and a record describing a subject are different objects that happen to occupy identical real estate on the page. Once you can tell them apart, the census stub stops looking like an entry with room left in it and starts looking like exactly what it is: a number, correctly reported, sitting alone in a room shaped like a much bigger story.
The number is honest. The room is the lie.