Half the titles in our Southeast Asia panel cannot be verified as Korean. Here is exactly how big that is.

Netflix does not say where a title was made. Anyone building a Korean panel has to decide that for themselves, and the usual method is to match a list of Korean works against the charts by name. That works until two countries make something with the same name, which they do constantly.

There is one thing in the published data that settles it. Netflix splits its global Top 10 by a title’s primary language, and a Korean work belongs on the Non-English side. If a title has reached the global chart, the label tells you which work charted.

Country charts carry no language field. And 204 of the 405 titles in our Southeast Asia panel have never reached a global Top 10. For those, the check cannot be run at all. They are in the panel on a name match and nothing else.

How much this actually costs you

Two numbers, and they point in different directions.

By title count it is half the panel. 204 of 405 — 50.4%.

By size it is small. The median title in that unverified group charted in one country for three weeks. The verified group’s median is 28 countries and 8 weeks. Whatever is wrong in the unverified half is concentrated in titles that barely moved.

Both numbers are true and neither is the answer on its own. If you are counting titles, half your rows are unconfirmed. If you are weighting by reach, the unconfirmed part is a tail.

What is actually in there

We looked. Among the unverified titles that did travel widely, foreign works with Korean-sounding company are plainly present — Tarot in 54 markets, Gossip Girl in 45, Mr. Robot in 43, Barbie in 37. None of those is a Korean work; each matched because Wikidata also records a Korean film or series with that exact name.

Others in the same group are real: Snowpiercer, Scent of a Woman, Watcher, Mine are Korean works that simply never broke into a global Top 10. We are not going to tell you what share is which, because we have not read all 204, and a guess presented as a figure would be worse than the gap it papers over.

Where the rule does work

On the titles that can be checked, it removes real errors rather than hypothetical ones.

36 titles matched a Korean work by name and turned out to sit on Netflix’s English charts — so they are the English-language work, not the Korean one. Seven more came out of this panel by hand after being read one at a time: Teach You a Lesson is Chinese, Hunger is Thai, The Empress is German.

The hand list is longer than seven — it is maintained once for every panel we build, and only the entries that actually appear in Southeast Asia’s charts show up in this count. The most recent addition, on 7 August 2026, was Friends: Wikidata carries a 2002 Korean drama of that name, but the title charting for 108 weeks across 36 markets is the American sitcom. It never charted in these six countries, so it is not among the seven above.

Why we print this at all

A panel that quietly includes what it cannot verify is more dangerous than one that says so, because the first kind gets quoted. These counts are now on /titles, /reach and /staying-power, read from the build rather than typed in, and a check fails our build if any limit the build measures stops being shown.

That check was written on 7 August 2026 after we found that the figure had been computed for weeks and displayed nowhere. Counting a limit and showing it are different things, and until that day we had been treating them as the same.