The 'lost two years' happened to all eight fourth-generation girl groups

A Korean entertainment desk ran a piece this morning under the headline “NewJeans’ lost two years.” The frame is familiar: a group at the top, a contract fight, and a long silence.

The frame may be right about Korea. Nobody had checked whether it is right about everywhere else.

So we counted. English Wikipedia records how many people opened an article each month. That is not affection and it is not sales. It is the number of people who wanted to know something. For a group whose audience is largely outside Korea, it is one of the few things about the last two years that can actually be counted.

What we counted

Every fourth-generation girl group with an English Wikipedia article — the set fixed by debut year, 2018 through 2022, so that the comparison is not our taste. For each one, monthly lookups from August 2022 through August 2026, then two twelve-month windows set side by side.

The result

NewJeans is at 0.61x. The median of the other seven is 0.66x. Two groups fell further than NewJeans did, and neither of them spent the window in court.

That is the finding. The lost two years are not a NewJeans story. Six of eight groups lost between a third and a half of their English readership over the same stretch, and the two that held on — aespa and NMIXX — are the exception that needs explaining, not the six that did not.

Against the rest of the catalogue, all eight are holding on

Six of eight falling by a third or more over two years reads like decay. Set against every Korean act we can measure, it is the opposite.

We keep a separate count of how long a group stays looked up after its peak: 774 Korean acts, each measured from its own best month. 731 of them have already halved. The median act took two months to do it.

Those two figures cannot be set directly against the table above, and the difference matters. The catalogue count compares one peak month with one recent month, which is the harshest test available. This article compares a twelve-month average with a twelve-month average, which is gentler by construction. A ratio from one is not a ratio from the other.

What the catalogue figure does establish is the shape of the distribution: sustained reading two years on is rare. Against a field where the typical act is halved within a season of its peak, eight groups still holding between 47 and 91 percent of a full year of reading are at the durable end, not the fading one. The headline framing has the direction right and the population wrong.

Which of the eight is read at all is a different question again. Our count of who the English-speaking world looks up when it looks up K-pop is where the monthly figures in the table above come from, and how lopsided that attention is explains why the gap between 74,031 and 7,686 in one table is normal rather than remarkable.

What this does not say

It does not say the dispute cost nothing. NewJeans still has the largest audience in the set by a wide margin, 74,031 lookups a month against a set median near 30,000, so a decline from a high base is not the same event as a decline from a low one. And it does not say the dispute caused the decline, because we did not measure causes. We measured eight groups over the same forty-nine months and put them in one table.

What the table rules out is the simplest version of the story — that something specific happened to one group. Whatever happened, happened to nearly all of them.

One number we got wrong first

The first run of this count put the other seven groups at a median of 0.74x, which would have supported the headline: NewJeans down 41 percent against a field down 26 percent. Two mistakes produced it.

One article title is written in full capitals, and the camel-case spelling we used returned a different, almost-empty page — 138 views over three months where the real article had 15,550. The other is that two groups renamed their article mid-window, and reading only the current name threw away the earlier half of their history, which removed one group from the table entirely.

Fixing both moved the field median from 0.74x to 0.66x and closed most of the gap the first run had found. The correction is recorded here rather than quietly applied, because the first version of this article would have been wrong in a way no reader could have checked.

A third one, caught by reading the numbers aloud

NewJeans’ figure for August 2026 came back as 1,343, against 69,473 in July. We went looking for a third article name to explain a 98 percent collapse. There was no third name. The request had been sent with the window ending on the first of August, and a monthly series asked to stop on the first of a month returns that month as a single day.

Every group’s recent average was being pulled down by the same one-day month. The ratios survived it because the error fell on all eight alike, but the averages printed above would have been wrong by roughly a twelfth. The window now ends at the last complete month, and the check that enforces it is part of the measurement rather than a note in a file.