Vanishing Scarcity: Generative AI and the Redefinition of Musical Value

The contemporary sonic environment is undergoing a transformation comparable in depth to the invention of sound recording, yet unfolding at an incomparably greater speed. The platform Suno, founded by former Meta researchers, brings together more than one hundred million users and produces on the order of seven million musical works daily. For the sake of comparison, it is useful to keep the following proportion in mind: a volume equivalent to the entire Spotify catalogue is reproduced by this system in roughly two weeks of continuous operation. The company’s valuation reaches 2.45 billion dollars against an annual revenue of approximately 300 million, which permits one to speak not of a laboratory experiment but of a mature commercial sector embedded within the attention economy.

It is essential that production of this kind requires neither specialized education, nor prolonged preparation, nor even a distinct musical intention. A technically competent composition, containing elaborated harmonic layers, instrumental parts, and vocal lines stylistically compatible with the editorial playlists of major services, arises from a text prompt within an interval comparable to ordinary everyday actions. A peculiar intermediate zone emerges: the result can be characterized neither as outstanding nor as inadequate, and it is precisely this absence of extremes that turns out to be the most significant symptom of the shift underway.

A study conducted jointly by Deezer and Ipsos on a sample of nine thousand respondents from eight countries recorded a figure that calls for separate reflection: ninety-seven percent of participants proved unable to distinguish a machine-generated track from a work of human origin under conditions of blind listening. This kind of perceptual indistinguishability is confirmed by market precedents as well. The composition “Walk My Walk” by the project Breaking Rust, as well as the album “Floating on Echoes” by the collective Velvet Sundown, reached the upper positions of the Spotify and Billboard charts, which means that millions of listeners interacted with these works without possessing information about their nature, and detected neither aesthetic nor emotional dissonance.

This fact demolishes one of the industry’s implicit assumptions, according to which a trained ear is capable of recognizing a certain quality of “humanness” supposedly inherent to live performance. The empirical data testify to the opposite: the category of “humanness” in sound turns out to be less stable than had been assumed, and was to a significant degree sustained by extra-musical context — knowledge of the author, the circumstances of creation, the cultural frame of perception.

A theoretical support for interpreting what is occurring is provided by Walter Benjamin’s concept, according to which the aura of a work is composed of its rootedness in time and space, together with the viewer’s or listener’s awareness that a particular person expended a limited resource of their own life in creating it. Mechanical reproduction, in Benjamin’s account, detaches the work from this context, substituting quantity for quality. Generative systems carry this process to its logical limit: what they eliminate is not the copy but the original act of human labor itself.

The traditional path into the musical profession is described through the notion of filtering based on friction. Years of instrumental training, the mastering of theory, the accumulation of performing experience, work within an ensemble, the formation of an individual voice — all these stages performed a dual function. On the one hand, they produced technically competent performers; on the other, they served as indirect evidence of seriousness of intent, endurance, and the capacity for prolonged, concentrated effort. Technical competence thus functioned as a substitute for more fundamental qualities that cannot be measured directly. Under conditions in which a generative system yields a technically flawless result at zero friction, this indirect connection is severed, and mastery ceases to function as a sign of anything beyond itself.

The democratization of musical production is not a phenomenon that arose all at once. The appearance of synthesizers, and subsequently of drum machines, progressively lowered the thresholds of entry, redistributing technical capabilities between professionals and amateurs. The development of stock platforms, represented by such services as Shutterstock in the visual sphere and Splice in the audio sphere, continued this trajectory, converting previously unique elements into standardized building blocks. Generative models bring this logic to its conclusion, eliminating not individual barriers but the very structure of barriers as such.

As a result, the musical surface of culture acquires properties previously associated with wallpaper ornament: it fills space, provides an acoustic backdrop, but ceases to hold sustained attention. The iconic function — the capacity of a work to serve as an event around which perception organizes itself — remains attached predominantly to those artifacts that are connected to a traceable human biography and to the living presence of a performer.

Major market participants have developed a dual strategy, combining litigation against the unauthorized use of data for training models with simultaneous partnership with the most advanced platforms. Deezer has moved to actively identify machine-generated content, the volume of which in 2026 reaches seventy-five thousand uploads per day; such tracks are marked with special labeling and excluded from algorithmic recommendations. The media conglomerate iHeartMedia has launched the “Guaranteed Human” program, certifying the participation of live DJs and hosts.

In the legal dimension, the precedent of Thaler v. Perlmutter is telling, having established within U.S. jurisdiction the principle according to which human authorship constitutes a basic condition for the granting of copyright protection. Works produced by a generative system without substantive human contribution find themselves in a legal gray zone, which creates additional economic incentives for preserving at least a decorative human component within the process of creation. Projected losses in authorial revenue are estimated at twenty-four percent, or roughly ten billion euros, by 2028, which sets the lower bound for the scale of value redistribution within the industry.

As technical flawlessness ceases to serve as a reliable indicator of quality, a set of alternative grounds for evaluation advances to take its place. The first of these becomes provenance — the traceable history of a work’s origin, including an identifiable author, a documented process of creation, and a verifiable context. The second ground is live performance as a form of legitimation, since the presence of the performer’s body in a space and time shared with the listener does not lend itself to full simulation. The third direction turns out to be curation, understood as an independent craft: the capacity to select, juxtapose, and contextualize material under conditions of its practically unlimited supply acquires a value previously attributed to creation itself.

Alongside this, a fourth position is also taking shape, consisting in the rejection of authenticity as a relevant criterion. According to this perspective, the question of the human or machine origin of a sound loses significance, and evaluation reverts to the purely perceptual characteristics of the result. The clash between these approaches defines the content of the current discussion within the industry and around it.

Illustrative for understanding the bifurcated nature of generative systems is the contrast between two types of their use in the adjacent field of choreography. On the one hand, the neural interfaces employed in the project involving the paralyzed dancer Brianna Olson demonstrate the function of a prosthesis: the technology restores the will of the creator, recovering a lost capacity for the physical embodiment of intention. On the other hand, movement datasets such as Google’s AIST++ realize the function of stock, in which the living performer becomes superfluous and their labor is replaced by a statistical generalization of previously recorded performances.

The distinction between these two vectors carries fundamental significance for the musical sphere as well. In one case, the generative model expands the space of human expression, compensating for limitations of the body, of resources, or of training. In the other case, it replaces human expression as such, turning it into one of the parameters of an optimization problem. Both trajectories are technically identical and rely on similar architectures, which makes the question of application, rather than of the technology itself, central to ethical evaluation.

The totality of the processes examined leads to a question that extends beyond the bounds of industry analysis. If the value of a musical work, over a lengthy historical period, was sustained not only by its acoustic characteristics but also by its function as testimony to human feeling, then a situation in which the sound persists while the experience standing behind it disappears demands a revision of the basic definition. Is beauty a property of the sonic fabric itself, one amenable to statistical reproduction, or is it rather the result of the listener’s knowledge that behind the sound stands another human being with his or her own history of effort and suffering? The answer to this question determines whether effort and the resistance of material will prove to be necessary conditions of artistic legitimacy, or whether it will turn out that they were historical circumstances accompanying art without constituting its essence.

The institutional and aesthetic order presently taking shape inclines toward the first variant of the answer: the system of labeling, the legal entrenchment of human authorship, and the economic reorientation toward live performance and curation all point to an attempt to preserve the connection between the work and labor. The durability of this choice will be determined not by the technical capabilities of the systems but by the audience’s willingness to pay for the invisible part of the work — that very part which a generative model is, by definition, incapable of either reproducing or simulating, since its absence is precisely what constitutes its essential property.