Who Killed the Moldovan Eurovision Finals?
Political Intrigue! Technology Buzzwords! Conspiratorial Thinking! Graphs!
I know the last week or so was a tough time for a lot of folks in my communities, and I too have spent a good deal of time processing. We live in challenging times of transition, and this takes a toll as we try to make sense of the new world around us. That’s what today’s newsletter is about: developing an understanding of what’s happened so we can plan for the future.
Of course, the news I am talking about is Moldova’s surprising withdrawal from Eurovision!1 An unexpected and mostly unprecedented choice that left 12 announced finalists just kind of hanging there. And when I say I’ve spent time “processing”, I mean I created a data pipeline to process ~500 audio files of past Moldovan song entries to analyze musical and lyrical trends. (In my head this was something I could do in a few hours, but wow did it ever take longer than that!)
For those of you who don’t follow the qualifying round for regional finalist competitions, or event haven’t heard of Eurovision before, here’s a brief primer
What is Eurovision? It’s kind of like the Olympics of music. Though I guess the Olympics sometimes has music kind of? So imagine an Olympics of Music that isn’t the Olympics. A Non-Olympics Music Olympics. The Non-Olympics Music Olympics, even. But mostly just Europe. (Mostly.) It’s a song competition!
Why did they make it? A couple of reasons: (1) after World War II, the major countries of Europe were trying to find a way to interact without invading each other, and indeed Eurovision-participating countries almost2 never3 do4; (2) it was recognized that reality competition television was a low-cost way to create content for those big ad machines people were increasingly putting in their living rooms, about half-a-century before the US would exploit the same insight with American Idol and more.
Is it aligned with the European Union or the Eurozone? Neither! It’s operated by the European Broadcast Union, which has no association with the other European Union which was founded 43 years after it. So it’s aligned with “a European union”, just not “the European Union”. It’s the Eastern Hemisphere equivalent of how “America” is both a country and a pair of continents. Like, since we’re socially constructing geography, why not heighten the absurdity by using basically the same name for different territories.
How are the songs selected? It varies! In the UN, some representatives are selected through democratic processes and some are selected through dictatorial processes. Eurovision does this too! On the more democratic side of things, Sweden’s Melodifest 2023 had four heats with two rounds as qualifiers, then a semifinal round and then a final round -- in all creating a six-week direct democracy media event. On the more “centralized” side of things, the UK has an internal selection process and they just announce and promote the chosen artist on BBC.5
As for the focus of the recent Eurovision news, Moldova hosts its “Etapa națională” as a two-round song selection competition run by their broadcaster Teleradio Moldova (TRM) that includes a mix of direct voting and expert judges. So somewhere in the middle on the Eurovision democracy-autocracy index! And this “mixed” selection process is where, on Wednesday, January 22nd, the judges decided to end their participation in Eurovision, having already selected finalists and being on the hook for a financial penalty for withdrawing after the October 15th deadline.
Which brings us to the central mystery6: Why did Moldova really withdraw? Since we are data-driven pop culture detectives here at Middling Content, we will investigate! We will examine the evidence! And we will interrogate the suspects!
First, the evidence: testimony from “music specialists” with “reservations”! See below, a post from TRM, the public broadcast network responsible for the Moldovan song selection.
“We consulted music specialists, and most of them expressed serious reservations about Moldova’s participation this year, in exact agreement with the opinion of the jury that TRM selected for this edition. We respect the opinions of professionals and the general public, and we understand that we must make responsible decisions ,” Corneliu Durnescu noted.
Who are these music specialists? What were their… “reservations”?
And can I, an amateur musicologist, validate these concerns?
As for the suspects, there are three.
Theory 1: The Songs Were Bad 😵
Theory 2: AI is Destroying Creativity 🤖
Theory 3: Politics / Europop PsyOps 🕵️♀️
Let’s investigate each of them -- with graphs!
Theory #1: Art Is Dead
Nobody Wants To Songwrite Anymore
Probably the most obvious hypothesis to have in this scenario is that the songs and performances just weren’t very good this year. And you’re probably thinking, “Is there even much of a music scene there? I couldn’t even name a single Moldovan song!”
…But you’d (probably) be wrong!
If you are of my generation, you are probably familiar with the Numa Numa Song, made famous by early internet video celebrity, Numa Numa Guy. Numa Numa Guy was named VH1’s #1 Internet Video Icon, permeating pop culture. And the Numa Numa Song was sampled both in TI & Rihanna’s 2008 US #1 hit “Live Your Life” and then again in the 2024 David Guetta & OneRepublic song of the summer contender “I Don’t Want To Wait”. If I’m reading this Wikipedia page right, it’s 20th on the list of top-selling physical singles ever -- tied at 12M copies with The Beatles’ “I Want To Hold Your Hand” and The Village People’s “YMCA”. The Numa Numa Dance video foreshadowed the mass appeal of normal people dancing to high energy tunes, a format that launch short-form algo apps like Musical.ly and TikTok a decade later.
Numa Numa Guy’s real name is Gary Brolsma, and Numa Numa Song’s real name is “Dragostea din tei”, a europop earworm constructed and performed by O-Zone, a band from… Moldova!
So there’s a rich musical history! And it makes sense that Teleradio Moldova wanted to uphold standards.
“The decision [to withdraw from Eurovision] was not an easy one, but necessary in the current context. This year’s national selection highlighted certain challenges, including a decrease in public interest and the overall quality of the plays and artistic performances. This fact generated a negative perception among specialists and the public, underlining the need for a substantial reform in the selection process,” said the director of Moldova 1, Corneliu Durnescu.
Which raises the question… were the songs so much worse than to be expected?
It’s hard to get a standardized measure of quality, but I figured I would look at the YouTube views to get at this.7 The winning song will of course get a great many views, but those that don’t progress probably stay roughly in the same neighborhood. What does the 2025 batch of songs look like within this context?
We can see that the finalists (marked in gold) generally lead with views, which makes sense since those songs receive a lot of publicity from the broader Eurovision competition. The log10 scaling can hide the difference here -- finalists will often get 100X+ the views of other contenders.
The top two all-time top performing Moldovan Eurovision entries are by the same group, SunStroke Project. They are bolstered by the meme status of one of their members, referred to as “Epic Sax Guy”. And the third most-played Moldovan Eurovision entry song is by another recurring group, Zdob și Zdub, who were the first Moldovan artists to participate in Eurovision as well as record-holders for most times qualified for Eurovision from semi-finals (three times).
Looking at 2025, we see that only one song has 10k+ listens, and that is “Semafoare” by Bacho & Carnival Brain. The difficult question becomes: is this high or low? It’s more listens than even the 2008 selection “A Century of Love” by Geta Berlacu, but otherwise lower than other selections. But of course, the plays increase for a song when it is announced as the contender, and I am including all Eurovision-related plays in the number -- so the official competition plays are counting too. It’s better to compare it to the second-place finishers for a comparison. In 14 of the 19 available years for comparison, we find “Semafoare” outperforming the second-place finishers.
But what about the general pack of songs, described as “creating a negative perception in the public”? What do we make of that?
This seems more plausible. In 8 of the past 10 years, there are at least 2 songs that seem to break the 10k view barrier. This year really only had “Semafoare”. They were, of course, planning for a finalist round to be broadcast locally, and perhaps they didn’t feel they had the content to support that programming. Still, it feels like song quality alone doesn’t provide an explanation.
The Verdict: Plausible But Suspicious
It seems clear to me that Teleradio Moldova could have selected “Semafoare” but didn’t. Maybe the judges really had some quality-related objection to it.
But what if the issue was not in the songs, but in the songwriting?
Theory #2: The AI Slop Has Come For Pop
Generating Song Submissions Is Easier; Singing Them Is Still Hard
Most readers will, I think, be familiar with the rise of GenAI as a force in the world of creative arts. It has critics! (I am, reservedly, one of them!) Specifically, many express concern about “slop” where large quantities of content can be created, occasionally causing problems downstream in content processes where humans need to review the content. This has, for example, led to overwhelming submissions at the SciFi magazine Clarkesworld in 2023, resulting in a temporary shut down of their open submissions process.
New tools like Suno.AI allow anyone to generate songs from a short prompt, so it seemed at least feasible that the concern was GenAI-related. Of three dropouts before the competition, at least one sparked rumors of being an AI-generated song submitted by a Ukrainian TikTok channel as a joke. (This song also happens to be the second most played of all the songs, and due to its patriotic lyrics was perhaps a pre-competition favorite.) When asked about it, the competition organizer stated that they do not test submissions for if AI was used in their creation.
How could we explore this possibility?
The Verge ran a story where an editor of Asimov’s Science Fiction shared that frequently AI-generated submissions would have the same title “The Last Hope”. While I haven’t done much exploration with AI-detection in music, I wanted to try to explore whether I could detect other similar homogeneities.
I processed the audio and lyrics for ~500 Moldovan submissions through harmonic and AI analysis, focusing (1) the potential “AI probability” of auto-transcribed lyrics and (2) how varied the harmonies / chord progressions work, using “bigram” analysis which is more common in NLP work.
This analysis is challenging and I honestly would like to work more on it. We do find that “Semafoare” has less variety in its chord progressions, but I honestly believe that is driven in part by the fact that Bacho, the performer, is generally more consistent with his pitch in a song that has a relatively straightforward melody.
Further, the high AI probability for lyrics are also misleading. This is obvious since clearly songs from earlier years also score high on this measure. GPTZero is optimized for English-language prose, so I expect that the confounds come from the general “predictability” of pop song lyrics and a lack of training material in Romanian.
There are a couple “soft tells” though. Generally if someone writes a song for their own voice, they are comfortable in the range the song is performed; in watching the semifinalists, it seems quite frequent that the melodies aren’t optimized for the singer’s voices. Additionally, in one of my favorite songs “Alexia”, the semifinalist is reading off of his phone the whole time; this too suggests that the song’s relationship with the performer lacks the intimate connection it would have with its songwriter.
With time, I’m sure GenAI music will get better at producing songs more easily performable by a given artist. I could imagine an app where people sing a few examples, their particular strengths are identified, and then a new song is created that plays to their strengths. This will create new challenges in the context of songwriting competitions. But perhaps other Eurovision selection processes will enforce at least some kind of Zoom screening before announcing the list of semifinalists.
The Verdict: Possible But Unverifiable
With the prevalence of this technology now available, it seems almost impossible that nobody submitting to Eurovision in Moldova would have used it. And if there were submissions without any detection attempts, it makes sense that some would make it through. But there’s really no way to know for sure.
But if we really want to delve into unknowables, we should investigate the potential that there’s a deeper geopolitical conspiracy afoot.
Theory #3: Little Green Melodians
Can the “Wind of Change” Blow Both Ways?
The geopolitics of Eurovision is one of the most interesting parts of the competition to watch.
For example, when Kiev hosted in 2017, Russia’s closed-door selection process just so happened to choose a singer who, due to a past performance in Russia-controlled Crimea, would not be legally allowed in Ukraine. And this contestant also just so happened to also be a highly photogenic sweet-looking young woman girl in a wheelchair. This of course forced a confrontation where the Ukrainian Security Council would “shockingly” not allow her to compete.
So despite Eurovision’s attempts to ban politics from the competition, there’s often political subtext that enriches my experience of the competition. So I did try to read in whether there might be a political angle here as well.
[Moldova’s broadcaster representative] stated that TRM aims to review the selection criteria and methodology, to encourage wider participation of talented artists and to provide a more consistent and competitive cultural representation on the international stage.
What might they mean by cultural representation? How representative have their performers been in the past?
First some context. Moldova is a country between Ukraine and Romania. It was a Soviet Republic for about half a century, before which it was a part of Romania, before which it was an Ottoman vassal state for 300 years. It has a couple million people and a couple autonomous breakaway states.
An aside: I was there for about 30 minutes in the midnight hours of November 26th taking a bus from Bucharest to Odessa. It was… dark out. But I was awake! I saw a little of it!
Like in many countries, the politics in Moldova are challenged by inflation and other instability. There are parliamentary elections this year, and the incumbent pro-European government is feeling pressure. Though their recent constitutional referendum on EU membership passed, it came in much closer than expected. (Reportedly in part due to a $39 million “cash for votes” scheme operated by a former Moldovan politician implicated in the “disappearance” of $1 billion from Moldovan banks.)
Below I’ve mapped past Moldovan Eurovision contestants to their birthplace and other areas of affiliation over top of a map of the recent EU membership referendum.
There is clearly a skew in Eurovision finalists in the areas opposed to EU membership. (Ultimately, support for EU membership seems to correlate strongly with proximity to the capital.) It’s nice to see such a mix of locales represented!
But something else is more interesting to me.
I mentioned earlier that there are two separatist states within the UN-recognized territorial borders of Moldova.
One is the southern region that is dark green and strongly opposed to EU membership — Găgăuzia. With only ~100k residents, they’ve never had a representative to Eurovision, but they did send representatives to all four seasons of Turkvision, the Eurovision-inspired competition that Turkey broke off to start to object to voting changes.8
The other separatist state, Transnistria, is the long strip on the border of Ukraine. This region is under effective control of its own government, hosts Russian troops, and despite holding only 13% of the region’s population has connection to 45% (5 in 11) of the non-foreign artists representing Moldova.9 This is nearly three times what we would expect from random chancew!
The skew is even more pronounced when we contemplate the success of the songs in terms of playcount. Of the top 10, only 2 are solely associated with official Chișinău-controlled Moldova territory; of the bottom 10, they have 9.
The Transnistrian connection in some cases can be even quite strong. I previously mentioned SunStroke Project, the act whose 2010 and 2017 performances are the first and second most listened to Moldovan submissions ever. Their Wikipedia page says that the name “SunStroke Project” is actually a reference to a member having a heat stroke… when they were serving in the Transnistrian army.
This puts their song in new context. Is this song somehow political? And if so, how? Is it a call for unity that former Transnistrian soldiers are competing under the Moldovan flag? Or is the 30-second highly memed sax solo designed to awaken our class consciousness so we support the separatists?
I listened to a good portion of this 10-hour loop yet the politics are still not clear to me.
The regional distribution of songwriters connects to this year’s competition as well. “Semafoare”, the leading song I’ve discussed already, is co-written by Bacho and Carnival Brain, and Carnival Brain contains one of the SunStroke Project members in the Transnistrian army heat stroke anecdote. While he’s relocated to Estonia, a political dimension to the decision seems at least possible, right?
I expect the political dimension is even simpler: life is very tough in Transnistria right now.
Transnistrians have been enduring the winter cold without heating since Russia halted their supply of gas on New Year’s Eve, meaning these semifinals operated with many suffering from the cold weather. Maybe a decision to stay out of Eurovision is an act of sensitivity? A performative prioritization of seriousness and austerity, intended to win over the eurosceptics north of the capital? Fortunately, the Moldovan government has been able to come to a deal to provide gas to the freezing Transnistria.
So it’s mercifully not all sad news in Moldova these days.
The Verdict: Curious, But Mostly Just Engagement Bait
There's likely no grand conspiracy here, but the fact that Transnistrian artists are so disproportionately represented in Moldova’s Eurovision entries is genuinely striking. I’m just not sure what to make of it yet.
I expect the musicians of the separatist Transnistria and those in the Moldovan capital Chișinău represent their governments about as much as American musicians feel aligned with US policy. Many of my favorite Chinese punk acts are from Beijing. The people who choose music tend not to fall into geographic political frameworks.
Indeed, these transpolitical vibes are a great deal of what Eurovision is about. I do find myself feeling more connected to the people of Transnistria after hearing the music of artists there; this is what can make music a powerful force of peace in the world.
Conclusion
I think it’s interesting how this early competition has a mix of the major stories of our time -- decreasing quality of services, rapid tech changes, and geopolitical upheaval -- all within it. I will look at a few additional country-level competitions to see if perhaps similar themes are present, even though I’m skeptical we’ll find anything conclusive.
To quickly summarize…
I don’t think the music was so bad that they needed to withdraw, but I do think the music had degraded, and the reliance on bringing back old acts for the the past five years already was a sign that they needed to be more proactive in identifying artists to participate; it’s possible that rewarding legacy acts even discouraged young talent from participating.
I don’t think all the songs were written by AI, but I do suspect that technology was used beyond just the act that withdrew. And I’m curious how AI will impact selection in other rounds and the overall submission rules in the future.
I don’t think there was some kind of nefarious psyop scheme by a hostile foreign power to infiltrate Eurovision, but I do think the political challenges of our times are making it harder for politicians to prioritize cultural initiatives. And it’s possible this will just get more difficult as the news marches on.
Concerningly, I can see all these factors will contribute to a sort of “democratic backsliding” in Eurovision. I mentioned the UK already moving away from direct election of participants in 2010, but it’s also true that The Netherlands ended their “Nationaal Songfestival” in 2012. Perhaps more countries will also follow suit.
A theme that has been going around lately is maintaining positivity by thinking less about systems collapsing around us and more time thinking what we may want to rebuild in the ashes. I do think there is a sort of creative democracy that lives on in the apps and platforms that let anyone compete for attention via algorithms. So in a way that wasn’t possible in 1956, it’s kind of Eurovision every day all the time all over the world. A challenge is that it’s unlikely that these algorithms operate fairly, and other than TikTok (perhaps disingenuously) and Twitter (by rogue employee leak) they have been only opaque. But these are tractable technical challenges, and I’m confident we can find the solutions.
Maybe in the future, a fair and open algorithm can serve as a feeder and qualifier, pulling the strongest songs into a concert/festival like Eurovision that can celebrate talent surfaced anywhere in the world. In the meantime, I’ll stay vigilant and do my best to provide data journalism to help readers like you connect with global pop music.
I did plan to discuss the other news of last week in a blog post, but I thought I’d write a short piece on Eurovision first. And then I found myself learning new audio processing libraries and designing a custom AWS data infrastructure and a week flew by. But I have thoughts and feelings about the state of the US that I believe I can frame everything within a media context.
see: https://en.wikipedia.org/wiki/Russo-Ukrainian_War
see: https://en.wikipedia.org/wiki/Nagorno-Karabakh_conflict
see: https://en.wikipedia.org/wiki/Russo-Georgian_War
The UK’s Eurovision entry selection process was centralized in 2011 when the Conservative Party had just taken the lead of the government; perhaps with the return of the Labour Party to power there will be direct election of Eurovision representatives again?
A comment from Ezra Klein in a recent podcast about “the opposite of doom” has stuck with me. Doom really captures attention. Hope less so. But “mysteries” are classic, and that’s a note I’m taking with me as I think through this blog.
I attempted to automate matching songs to their YouTube videos in a way that may have been imperfect, but it captures the broad strokes.
The Uyghurs of Xinjiang Autonomous Region were also scheduled to participate in Turkvision 2013, though according to Wikipedia they “later opted to not partake”. (Autonomously, one must presume.)
The 2014 Moldovan census put the population at 2,998,235 and the Transnistrian census was at 475,007. For more, see this Wikipedia page: https://en.wikipedia.org/wiki/Demographics_of_Moldova. For submitter statistics, I’m counting both Zdob și Zdub entries and both Natalia Gordienko entries as from individual artists even though they occasionally performed with collaborators.