AI music is often discussed as if there are only two positions available. Generative tools are either opening music to more people and giving artists new creative possibilities, or they are undermining musicians, replacing paid work and filling the internet with disposable content.
David Rovics does not fit comfortably into either position.
Rovics has spent decades writing, recording and performing songs, with political movements and current events forming a substantial part of his work. His official biography documents performances at protests across North America and Europe, while his current website presents Ai Tsuno as his AI-assisted musical alter ego, with Rovics describing himself as the lyricist, prompt engineer and producer behind the project.
He is enthusiastic about what generative music has made possible for him. He is also worried about AI, sympathetic to musicians whose work is threatened by it, and clear that something important disappears when a studio full of people is replaced by software.
That combination makes his experience more useful than another argument about whether AI music is simply good or bad.
Working with an AI band
Rovics does not usually describe Suno as software. He talks about working with an “AI band”.
That language partly reflects how the process feels to him. After years of recording with producers, engineers, session players and other musicians, he sees unexpected similarities between directing human performers and generating music with AI.
Neither necessarily gives him what he wants immediately.
“If you don't want that, and you have a real idea of the sound you're looking for, whether you're working with humans or AI, you have to know what you want and how to explain that to your band.”
Rovics writes the lyrics himself, then alters lyrics, prompts and musical directions while generating new versions. He says reaching a finished song can involve twenty, forty or sixty iterations.
For him, the valuable skill is therefore not simply knowing how to ask Suno for a genre. It is knowing when the result has arrived at the emotional effect he wanted.
A sad song should actually feel sad. A funny song needs to land as funny. An anthem intended to energize an audience needs to achieve that purpose. Rovics believes his years of songwriting, performing and recording give him a reference point for making those decisions.
This sounds familiar after several of our recent interviews. Vitalii Klimenko argued in AI Music and the New Burden of Choice that generation creates options while production creates decisions. Brian Myk similarly described in What Skills Should Artists Value in the Age of AI Music? how easier experimentation can make judgment and selection more important.
Rovics arrives at a similar idea through decades of recording experience. Generative systems may make polished sound much cheaper, but they do not necessarily make judgment cheap.
Why the binary breaks down
Rovics is particularly frustrated by the tendency to reduce the conversation to whether someone is for or against AI.
He does not dismiss the fears surrounding the technology. In our interview, he raised concerns about employment, environmental costs and the wider consequences of increasingly powerful AI systems. His argument is instead that rejecting the technology as an individual artist does not make those developments disappear.
“What I think we need desperately in the world, and especially on the internet, when it comes to AI and most other things is the capacity for nuanced thought.”
That does not mean every use of AI is equivalent.
Rovics challenges the idea that there is a clean border separating “AI music” from everything that came before it. He points to the extensive role technology has already played in modern recording, including tools for correcting pitch and timing.
The technical distinction is worth keeping clear. Traditional digital pitch correction or audio processing is not automatically artificial intelligence. But Rovics's wider point is useful because the label “AI music” can hide enormous differences between workflows.
A song with human-written lyrics, dozens of generations, detailed production decisions and AI-generated instrumentation is very different from accepting the first fully generated song produced from a short prompt. Both may end up carrying the same AI label.
The more useful questions are therefore about process. Where was AI used, what did the person contribute, which decisions remained human, and what work was automated?
Experience matters, but it may not be everything
Rovics goes further and argues that much poor AI music comes from people who lack the musical experience needed to recognize when something is not working. He contrasts that with experienced musicians who can use generative systems while bringing years of songwriting and production judgment to the process.
His own workflow gives that argument weight. Experience clearly affects what Rovics hears, rejects and keeps.
But it does not necessarily follow that good generative music must come from established musicians.
Our other conversations point in a different direction. Filmmakers, writers and other creative professionals may possess skills such as pacing, narrative structure, editing or visual identity that become unexpectedly useful once generative tools give them practical access to music production.
That tension is important. AI may reward existing musical experience while simultaneously allowing people with completely different backgrounds to discover abilities that were previously difficult to express through sound.
Both can happen at the same time.
AI does not make a weak idea strong
This becomes particularly clear when Rovics discusses political songwriting.
Rovics has spent much of his career writing songs connected to political issues and movements. His official biography describes a long history of performing at demonstrations, while his current work continues to use songs as a form of political communication.
His view of AI is not that better production automatically makes those songs better.
He argues that weak political writing remains weak even when a generative model surrounds it with polished vocals and instrumentation. Conversely, a songwriter who already has something effective to communicate can suddenly package that idea at a level of production that may previously have required a studio, musicians and a significant budget.
That distinction matters far beyond political music.
AI can reduce the cost of execution without reducing the difficulty of having something worthwhile to execute.
Political music also has a timing problem
For Rovics, lower production costs come with another advantage. He can move from writing a song to releasing a finished recording much faster.
He says that with his current AI workflow, he can finish roughly an album's worth of material during an average month. He contrasts a roughly $50 monthly spend on several AI tools with recording projects that he says could otherwise require tens of thousands of dollars to achieve comparable production values.
For political songwriting, speed has particular value.
A song responding to a current event can lose much of its immediate relevance if the recording process takes months. Generative production can compress the distance between an event, a written song, a finished recording, a captioned video and publication.
That makes AI more than a production technology in Rovics's workflow. It also becomes a communications technology.
His own website describes the appeal as being able to create what he calls “real-time musical commentary”, supported by quickly produced videos for social platforms. Rovics says one recent month of Ai Tsuno material received more than one million views across Facebook and Instagram, although those audience figures are his own reported metrics rather than independently verified platform data.
In our interview, he similarly said the Ai Tsuno releases often travel further than his conventional guitar and vocal recordings.
The interesting question is therefore not simply whether AI makes the songwriting better. It may make the same songwriter considerably more capable of competing for attention.
The studio becomes available all the time
There is an obvious reason why that prospect appeals to an independent musician.
Rovics describes working with Ai Tsuno as having access to a band that is available around the clock and capable of moving faster than he can write. He can ask for instruments such as oud or shamisen and explore combinations of sounds that would be difficult or expensive for him to organize conventionally.
He also likes the uncertainty.
Rovics says he does not always know exactly what the system will return, and compares that to working with musicians who bring their own interpretation to a recording. He knows the destination he wants, but the route can surprise him.
That does not mean he sees the two experiences as interchangeable.
What disappears when the musicians disappear
When asked what AI cannot replace, Rovics's answer changes the tone of the conversation.
“What can't be replaced, among other things, is the camaraderie, the banter, the friendships.”
He says he remains a strong supporter of human musicians and would like to spend much more time recording with them. The problem, according to Rovics, is that the budgets which once made those sessions possible had already become increasingly difficult for him to access long before generative music arrived.
This complicates one of the simplest narratives around AI music.
It is tempting to describe the process as an AI system replacing musicians who would otherwise have been hired. Sometimes that will undoubtedly happen. But Rovics describes another scenario in which the paid studio session had already become financially unrealistic, and AI entered a space where the alternative might have been a much simpler recording or no full production at all.
That distinction does not remove the labor issue.
If an AI system can generate convincing performances of instruments that previously required session musicians, the economic value of some of that work may fall. Rovics acknowledges this directly. He describes declining need for session players and engineers as sad even while arguing that the same change gives writers and producers access to capabilities they could never previously afford.
The same technology can therefore look very different depending on where someone stands in the production chain.
One artist gains access while another loses work
For a songwriter like Rovics, generative music can feel like an expansion of creative freedom. He can write more, experiment more often and hear arrangements that would previously have required money and coordination he did not have.
For a session musician, the same technology may represent fewer calls.
That is not a contradiction that can be solved by deciding whether AI music is good or bad. It is an economic tradeoff.
The recording industry has always divided creative work among different roles. Songwriters write, performers perform, producers shape recordings, engineers capture and process sound, and labels or independent teams distribute the result. Generative systems can collapse several of those jobs into a much smaller workflow controlled by one person.
That creates significant leverage for some creators. It can also reduce the amount of paid human labor required to make a recording.
Both observations belong in the same conversation.
What becomes more valuable
Rovics believes this shift increases the importance of writers and producers.
His reasoning is straightforward. If convincing instrumental performances, vocals and polished production become easier to generate, simply having access to those capabilities becomes less distinctive. What matters more is knowing what the song should say, how it should sound and when it is finished.
This again echoes the pattern emerging across our recent interviews.
Klimenko has emphasized judgment. Myk has emphasized selection. In The Story Behind AI Music Artists You Might Have Missed, we explored how Doniell Clay approaches identity and perspective across fictional performers.
Rovics adds experience as a producer and performer. He believes years spent writing for audiences and recording alongside musicians give him a better sense of when an AI-generated performance actually works.
The exact balance will differ from artist to artist, but the wider pattern is becoming clearer. As generating sound becomes easier, more of the distinction between projects may move toward the decisions surrounding that generation.
AI music is not one outcome
Rovics's experience does not resolve the arguments around AI music. It makes them harder to simplify.
AI can lower the cost of creating a record while reducing demand for some of the people who used to be paid to make it. It can give independent artists access to production they could never afford while making an already crowded music market even more crowded.
It can help a political songwriter respond to events faster while leaving legitimate questions about the technology, labor and infrastructure behind that new capability. It can even give someone an always-available virtual band while making them miss the actual people who used to be in the room.
That is why “AI music is good” and “AI music is bad” both fail as useful conclusions. They compress a set of artistic, economic and cultural changes into a judgment that is too simple to explain any of them.
Rovics is benefiting from AI music while simultaneously describing some of what it threatens. That is not inconsistency. It is what technological change often looks like when viewed from inside an industry rather than from either side of an argument.
The more useful conversation is about who gains access, who loses work, which human skills become more valuable, what becomes cheaper and what kind of music comes out the other side.
AI music is not black and white. The interesting part is everything happening in between.
Follow David Rovics
Find David Rovics through his official website, Substack, Patreon, Bluesky, Threads, Bandcamp, SoundCloud, Instagram, X and Rumble.
