AI changes the score

Across China's music industry, artificial intelligence is no longer on the sidelines. It is shaping how songs are created, filtered, distributed, and valued.

That shift set the tone for the 11th Music Industry Forum in Beijing earlier this month, where researchers and industry leaders examined an industry expanding rapidly, but also reinventing itself at the same time.

The numbers remain strong. China's music industry generated 519 billion yuan ($76.64 billion) in gross revenue last year, up 5.3 percent year-on-year. But beneath those figures, a different story is unfolding:AI is not just speeding up production, it is changing the structure of music itself.

A separate industry report, White Paper on China's Digital Music Industry (2025), coproduced by Tencent Music Entertainment Group, TME Research Institute and the Digital Music Committee of the China Audio-Video and Digital Publishing Association, describes 2025 as a "year of acceleration". Not just in consumption, but in creation.

The data is striking. Tencent Music Entertainment Group's platforms now receive 107,000 new song uploads every day, up from 76,000 in 2024. By December 2025, AI-generated songs accounted for 36.2 percent of all new releases. On AI music platform Suno, users generate more than 7 million songs per day. Music creation is no longer scarce — it is abundant, continuous and increasingly automated.

Professional musicians, for the most part, have adapted. The report notes a shift from caution to adoption, with AI now widely used to generate lyric ideas, arrange drafts and assist production. Music is no longer created solely by humans or machines, but through collaboration between the two.

Yet, this explosion of content has not simply diluted music, it has reorganized attention.

Algorithms and short-video platforms now account for more than 80 percent of new music discovery. Increasingly, listeners are not searching for songs; songs are finding them.

As production becomes easier, value is shifting from making "more music" toward making "more meaningful music".

One of the report's most significant findings is the rise of "super-fans". Listening has evolved beyond streaming into an expression of identity, participation and emotional connection. In China's digital music economy, these highly engaged users generate outsized value, spending far more than average listeners and driving demand for experience-based services rather than simple playback.

That reflects a broader global trend. Growth in traditional streaming is slowing as the industry shifts its focus from expansion to engagement. New users increasingly come from emerging markets, while future growth will depend less on scale than on the depth of audience relationships.

In China, that engagement is becoming both more intense and more global. The country became the world's fourth-largest recorded music market in 2025, according to the International Federation of the Phonographic Industry, the organization that represents the recording industry worldwide, with revenue rising 20.1 percent. Yet, per-user revenue remains well below that of mature markets, leaving considerable room for growth. At the same time, Chinese music is increasingly reaching overseas audiences through its distinct cultural identity rather than translation.

Soundtracks from games such as Genshin Impact and Black Myth: Wukong, along with viral rap tracks, are finding global audiences through platforms where rhythm, tone, and visual identity often matter as much as language.

Creativity meets machine

"Artificial intelligence is rapidly reshaping the music industry, but its impact goes far beyond faster production or lower costs. It is challenging a basic assumption: that music and art must be created by humans alone," says Li Xiaobing, professor and head of the department of music artificial intelligence and music information technology at the Central Conservatory of Music.

"For years, many believed machines could not create art because they lack emotion. But that view is now being questioned. We understand only a small part of the human brain," says Li. Advances in AI composition and voice synthesis have already produced music and singing that many listeners struggle to distinguish from human performances — and in some cases even find emotionally moving, he adds.

The impact is evident across the industry. AI can generate compositions in seconds that once took hours and reconstruct how musicians performed historical works through data analysis. Music education, performance and production are all being compressed into faster, more automated workflows.

But key questions remain unresolved — especially around copyright, originality, and ownership in a world where AI can generate vast amounts of music instantly, Li says.

AI is now embedded throughout the production process — from recording and arranging to demo creation — and is already replacing parts of traditional workflows, according to Guo Kun, deputy secretary-general of the China Audio-Video Copyright Association.

One of the biggest challenges, she says, is scale and transparency. Not all AI-generated music is identified as such, and some tracks are even presented as human-made. As a result, no precise figures exist, though industry estimates suggest platforms receive anywhere from 50,000 to 150,000 new AI-generated tracks every day, she points out.

Streaming services have developed their own settlement systems. While copyright issues remain unresolved, AI-generated music is already generating revenue. Some payments are treated as royalties, others as platform incentives. In many cases, the same rules apply whether AI is involved or not.

Industry observers note a clear divide between China and some overseas markets. "In many foreign markets, the attitude is more negative, and most AI music does not receive income or formal settlement," she says.

China's platform ecosystem, by contrast, is taking a more pragmatic approach. The key question is not whether AI was used, but whether the music attracts listeners and succeeds commercially. "If a piece of music sounds good and attracts listeners, it can generate income," she says.

Even so, industry data point to an important distinction. Fully AI-generated music rarely performs strongly on its own. Most commercially successful projects still involve human creators, shifting the debate from whether AI will replace musicians to how deeply it will become part of the creative process.

Regulation remains unsettled."There is still no final legal judgment in this area," Guo says, pointing to the absence of binding case law on AI-generated music and copyright.

The human difference

But while AI is transforming the front end of music creation, its impact is not evenly distributed across the production chain.

In fact, one of the most striking effects of the AI boom is that not every role is disappearing. Some are proving remarkably resilient.

Mixing is one of them.

Mixing is the stage where vocals, instruments, and effects are balanced into a finished track. It is technical, but also deeply subjective. And that subjectivity is precisely what has made it resistant to automation.

The reason is partly economic. As AI reduces the cost of composing and arranging music, more production budgets are being redirected toward the final stages — where human judgment still carries the greatest value.

Cashmere Studios in Shanghai offers a glimpse into why.

For its founder, Kaka, mixing is less about engineering and more about taste.

"Clients don't come to a mixing engineer just for technical output," he says. "It's like asking a friend for fashion advice. You want aesthetic judgment, not a correct answer."

He argues that mixing resists automation because it depends on choices that are difficult to quantify: how warm a vocal should sound, how far an instrument should sit in the mix, or how emotionally a chorus should land.

There is also a structural reason AI struggles here. Much of professional mixing relies on proprietary tools, hardware-specific workflows, and tacit knowledge that is rarely documented. In industry terms, it is a "black box" process — difficult to capture, and even harder to train at scale.

While AI can now generate complete songs, it is still far less capable of finishing them.

This imbalance is shaping a broader reality across the industry: AI is not replacing music creation outright. It is removing the repetitive, standardized layers while increasing the value of judgment-heavy work.

That shift is already showing up in earnings.

Independent musicians report declining income from certain types of commercial work, especially functional music for advertisements, corporate events, and short-form video content. These are exactly the areas where AI performs best: fast, cheap, and "good enough".

One industry description captures this new aesthetic bluntly:AI-generated music often sounds like a "standard face" — technically correct, polished, but lacking identity.

At Cashmere Studios, Kaka has witnessed the shift firsthand. Corporate commissions have declined as companies increasingly generate music in-house using AI tools. Film and game studios, once reliable clients for licensed tracks and live recordings, are also experimenting with AI-generated demos before approaching human musicians.

"Before, they would buy music or hire musicians to test ideas," he says. "Now they often start with AI."

Still, the industry is not simply shrinking. It is splitting.

On one side, AI is rapidly absorbing functional, low-emotion content. On the other, demand is growing for work rooted in human interpretation, cultural understanding and emotional nuance.

The broader picture is not replacement, but rebalancing.

Even as production becomes more automated, the industry is becoming more dependent on human judgment.

That may be the AI era's greatest paradox: the easier it becomes to generate music, the harder it becomes to create something truly worth listening to.