An AI actor gave a CNN interview. We counted who people actually look up

Three Korean newspaper headlines landed on the same morning, 10 September 2026. An AI actor had given an interview to CNN and said it was not humanity’s enemy. In a second piece the same AI actor was asked to improvise a scene and, in the reporter’s word, creaked. And a studio announced it was starting co-production with the Korean film industry using generated performers.

The question underneath all three is whether synthetic performers are taking the place of human ones. That question gets argued. It rarely gets counted.

One thing about it can be counted, cheaply and repeatedly: how many people go and look something up. English Wikipedia publishes per-article pageview counts, the API filters out bots before it hands the number over, and the data is public domain. So we took nine virtual performers and six real K-pop groups and counted twelve months of lookups — September 2025 through August 2026.

The headline number, and why we do not trust it on its own

The virtual side wins. It wins by 36,092 lookups out of 8.3 million — 0.4%.

That is not a lead. That is a tie with a rounding error attached, and anyone who published it as “virtual K-pop has overtaken real K-pop” would be telling a lie with a true number.

Here is the real side broken out, so the total above can be added up rather than taken on trust:

We know exactly how fragile it is, because we measured the same comparison twice on purpose. For our fictional-vs-real page we used the window August 2025 to July 2026 and got 1.2 times. This piece uses September 2025 to August 2026 and gets 1.00 times. Nothing about the world changed between those two measurements. We moved the window by one month.

Last month the real groups won by 3.2 times

The film’s page fell 87.7% from its first month in the window to its last. BTS’s page rose 42% over the same period. One of these is a release; the other is a following.

Strip the film out and the picture changes completely

The film is one title with one opening. Virtual performers as a category have been operating for far longer — Gorillaz launched in 1998, Hatsune Miku in 2007, K/DA in 2018, Eternity and Isegye Idol in 2021, MAVE: and PLAVE in 2023. That is 27 years of shipped product, not a summer.

So we removed the film and its in-film group and counted the seven that have been running for years:

The biggest virtual performer that is not a 2025 film is a British band from 1998. It draws 0.32 times the six real Korean groups combined, and less than BTS on its own (3,025,276).

And the biggest Korean virtual idol group, PLAVE, drew 69,094 lookups in a year — 2.3% of BTS’s 3,025,276. Not a third. Not a tenth. Forty-four times smaller.

Two of the seven grew, five shrank

Direction is worth separating from size, because a small thing that is growing and a small thing that is fading are different stories.

• Isegye Idol rose 4.9% from its first month in the window to its last — the only virtual page here that ended higher than it started.
• Eternity fell 5.2%, Hatsune Miku 0.3%, MAVE: 22.2%, K/DA 41.5%, PLAVE 52.5%, Gorillaz 58.5%.
• On the real side BTS rose 42% while the other five fell.

What this does not tell you

Pageviews are lookups. They are not streams, tickets, ad revenue or a contract with a burger chain. A page can be read by someone who loves a group and by someone who has just heard the name in a headline and wants to know what it is, and this count cannot tell those two people apart.

The film comparison has a further problem we will not paper over: people open a film page to find the plot, the cast or the song they heard, and they open a group page because they follow that group. Putting the two side by side is useful for asking “what is being looked up” and useless for asking “who has fans.”

The AI actor from the CNN interview is not in this table at all. It has no English Wikipedia article inside our window, so there is nothing to count. We could have substituted something adjacent and called it a proxy. We would rather leave the row empty and say why.

Why we published the weaker version of our own finding

We had a cleaner number available. The August-to-July window gives 1.2 times, which reads as a real lead, and it is a number we measured ourselves and already published on an axis page. Choosing that window and staying quiet about the other one would have produced a better headline out of the same honest data.

The reason we did not is that a reader who repeats “virtual beat real by 1.2 times” and is then shown the September-to-August window has been handed a fact that dissolves. We would rather hand over a tie, two windows that disagree, and the arithmetic to check both.

The full table, both windows and the per-article numbers are on fictional-vs-real.

These are statistics, not you. This is not investment advice, and it is not a forecast about anybody’s career.