There used to be a practical limit to how much music the world could produce.
A song took time to write, rehearse, record, mix, master and release. Even a rough track represented hours of decisions made by someone who cared enough to finish it.
That limit is disappearing.
Generative AI can now produce convincing songs in minutes. One person can create dozens of tracks before an independent musician has finished recording a vocal take. The result is not a distant prediction about the future of music. It is already reaching streaming platforms at enormous scale.
In July 2026, Deezer reported that around 90,000 fully AI-generated tracks were being delivered to its platform every day. On peak days, they represented more than half of all new uploads.
This does not mean that half of all music being listened to is AI-generated. It does mean that the volume of available music is growing far faster than any audience could possibly hear it.
Music no longer has a supply problem. It has a discovery problem.
When music becomes infinite, attention becomes scarce
For independent musicians, releasing music has never guaranteed that anyone will hear it. AI did not create that problem. Years before generative tools became widely available, artists were already uploading songs into catalogues containing tens of millions of tracks.
AI changes the scale.
If music can be generated continuously and at almost no marginal cost, a platform can receive more new songs in a day than a person could explore in a lifetime. Human-made tracks, AI-assisted productions, fully synthetic songs, functional background audio and automated spam all enter the same crowded space.
The cost of making another track approaches zero. The cost of finding something worth caring about rises.
That shift matters because musicians are not competing only with other musicians. They are competing with an unlimited volume of content for a limited number of human minutes.
For an artist who has spent months on a release, this can feel brutal. The song may be thoughtful, original and beautifully produced, yet disappear before it has been given a meaningful chance. Silence after release is easily mistaken for rejection, even when almost nobody has encountered the music in the first place.
Those are not the same thing.
Recommendation algorithms cannot solve every discovery problem
Streaming algorithms are useful. They help listeners navigate catalogues that would otherwise be impossible to search, and they can make remarkably accurate recommendations based on listening history.
But prediction and discovery are not identical.
A recommendation system is usually strongest when it has signals: previous streams, saves, skips, followers, playlists, listener similarity and established patterns of behaviour. A new artist often has very few of those signals. The system has little evidence that people will respond, so exposure is limited. Without exposure, the artist cannot generate evidence.
This creates a familiar loop:
Music receives attention because it already has signals, and it gains more signals because it receives attention.
The problem becomes sharper as platforms absorb more mass-produced content. Algorithms must filter an expanding catalogue, while bad actors learn to imitate the signals those systems reward. Spotify said in September 2025 that it had removed more than 75 million spammy tracks during the previous twelve months and announced stronger protections against impersonation and mass-upload tactics. That figure refers to spam broadly, not to AI-generated music as a whole, but generative tools make industrial-scale production easier than ever.
The question is no longer simply whether an algorithm can find music similar to what you already enjoy. It is whether unknown music can receive a fair first encounter before popularity, marketing budgets and behavioural data decide its fate.
Human taste becomes more valuable as content becomes cheaper
When almost anyone can generate technically competent music, technical competence alone becomes less scarce.
Taste remains scarce.
People do more than detect patterns in sound. They bring memory, mood, culture, curiosity and personal experience to a song. They can be surprised. They can recognise an imperfect voice that feels honest, a strange arrangement that should not work but does, or an unfamiliar artist whose work deserves another listen.
That does not make every human judgement correct. Taste is subjective, inconsistent and sometimes influenced by fashion. Yet those qualities are part of what gives music social meaning. A song matters because somebody chooses it, shares it, returns to it or connects it to a moment in their life.
Listeners appear to care about that human dimension. In a global study covering more than 43,000 people across 26 markets, 79% of music fans said human creativity remains essential to the creation of music.
As automated output grows, the value of a real person’s attention may grow with it. A genuine listen, comparison or recommendation becomes a form of curation that cannot be created simply by producing more files.
This should not become a war against AI
The future of music will not divide neatly into “human” and “machine.” Artists already use software throughout the creative process: virtual instruments, pitch correction, generative effects, stem separation, mastering assistants and countless other tools. AI can help a musician explore an idea, overcome a technical barrier or create something that would otherwise have been impossible.
Using a new tool does not erase artistic intent.
The more useful distinction is between creation and extraction: between technology used to extend human expression and systems used to imitate artists without permission, flood catalogues, manipulate streams or occupy attention through sheer volume.
A musician using AI as one part of an intentional creative process is not the same as an automated operation generating thousands of tracks to exploit platform economics. Treating those situations as identical would be unfair and unhelpful.
Transparency matters. Consent matters. Intent matters. Most of all, listeners should have meaningful ways to decide what deserves their time.
The next generation of music platforms must rethink discovery
Adding more recommendation technology to an ocean of content will not be enough. Music platforms also need systems that create fair encounters between listeners and work that has not yet accumulated popularity.
That means designing discovery around a few simple principles:
Give unfamiliar music a genuine first chance
An artist should not need an existing audience before anyone is allowed to encounter the work. Early exposure should not depend entirely on follower counts, advertising budgets or placement in an established playlist.
Ask people to make active choices
Passive streams reveal behaviour, but active comparisons reveal preference. When someone listens carefully enough to choose between two tracks, the decision contains a different kind of information from background playback.
Separate popularity from the opportunity to be heard
Popularity can be the result of discovery, but it should not always be the prerequisite for it. Otherwise, platforms simply reinforce yesterday’s winners.
Make manipulation expensive and participation meaningful
Systems should reward attentive human involvement rather than mass uploads, fake engagement or automated listening. As content production becomes cheaper, proof of genuine attention becomes more important.
Why Kallitone puts people at the centre
Kallitone was built around a simple idea: new music should be evaluated through human choices, not introduced only through predictions based on existing popularity.
Listeners compare tracks directly and help determine which music rises. The artist’s follower count, marketing budget or industry connections do not have to dictate the first impression. What matters in that moment is the music and the decision made by the person hearing it.
Human voting is not magically free from bias, and no discovery system can promise that every worthy song will succeed. But it changes the source of the signal. Instead of asking only, “What will keep this listener behaving as expected?”, it can ask, “Which of these songs does this listener actually prefer?”
That difference becomes more important in a world where music can be generated without limit.
Kallitone is not based on rejecting technology. It is based on deciding where technology should stop making the choice for us. Software can organise comparisons, connect listeners with unfamiliar tracks and combine many individual decisions into useful rankings. The judgement at the centre, however, still comes from people.
Human discovery is not nostalgia
Putting people back into music discovery does not mean returning to record shops, radio gatekeepers or a time when only a small number of artists could release music.
It means using modern systems to distribute human judgement more fairly.
AI will continue to improve. More music will be generated, uploaded and recommended. Some of it will be remarkable. Some will be useful. Much of it will be ignored. Trying to stop the technology entirely is neither realistic nor necessary.
The real challenge is preserving a path between a person who created something meaningful and another person who might genuinely care about it.
When music becomes infinite, discovery cannot be measured only by how efficiently a platform serves more content. It must also be measured by whether unfamiliar voices can still be heard.
The future of music may involve more machines in the creative process. That makes human attention, human choice and human curiosity more valuable, not less.
The more music we can generate, the more carefully we must decide what deserves to be discovered.
Discover music through people
Kallitone gives listeners an active role in discovering and ranking new music, while giving emerging artists a chance to be heard beyond follower counts and recommendation loops.