Deezer receives almost 75,000 AI-generated tracks per day, roughly 44% of everything uploaded to the platform daily, according to data cited in a review of listener attitudes published by unite. ai. The same review reports that a survey of 9,000 people found 97% could not distinguish music entirely generated by AI from music created by humans. The gap between those two numbers is the core of the story: production has become cheap and fast, while the audience's stated preferences have moved in the opposite direction.
What the streaming numbers show
Volume is not the same as demand. Many AI-generated tracks on Deezer receive few or no plays, which means that removing the cost barrier to production does not create an audience by itself. Deezer is one of the platforms that transparently tags AI-generated music, so the 44% figure rests on disclosed data rather than estimates. The study tracking attitudes toward AI in music creation ran from May to November 2025 and recorded a shift from -13% to -20% on a net comfort scale, meaning listener sentiment grew more negative over those months, not less. The same period saw the supply of AI tracks keep climbing.
The mechanism behind the negative sentiment is not the sound. In one experiment, participants rated music as significantly more likable, emotionally positive, engaging and higher in quality when they were told it was performed by a human, even though the audio was identical in both cases. A separate study found that before listening, most participants said they preferred human-made music, but after hearing both AI-generated and human-made songs they rated the two similarly. Even when listeners enjoyed the AI track, they remained less willing to endorse it. The reaction attaches to what the track represents, not to what it sounds like.
The debate is unfolding against a broader market shift. Experts predict generative AI could contribute up to $4.4 trillion annually to the global economy, and music is one of the creative industries where the tools have already reached commercial scale. Platforms are responding with labeling rather than bans, and the review notes that listeners want to know whether a track was made by a human, by AI, or by a combination of the two. That distinction matters because AI-assisted music and fully AI-generated music are different products: in the first case a human artist remains responsible for the main creative decisions, while in the second the system produces melody, lyrics, vocals and arrangement with little or no human involvement.
What this means for business
For companies that buy or commission music, the practical consequence is that disclosure is becoming part of the deliverable. A brand that licenses a track for advertising, a game studio that fills a soundtrack, or a retailer that runs in-store audio now has to ask where the audio came from, because a growing share of the catalog is machine-made and platforms increasingly mark it as such. The cost argument is real: AI tools can produce complete songs from simple text prompts within seconds, which compresses timelines and budgets that used to require session musicians, studios and licensing negotiations. For a small company this lowers the entry threshold to custom audio; for a large one it changes procurement, because the vendor list now includes tool providers alongside labels and composers.
The limits are less obvious than the savings. The 97% figure comes from a survey of 9,000 people and describes recognition, not preference: it does not mean audiences will accept AI music once they know its origin, and the same research shows they often will not endorse it even when they enjoy it. The attitude data from May to November 2025 points in the negative direction, so a buyer cannot assume that tolerance is rising. What the news does not establish is how any single track will perform commercially, and it does not settle copyright questions, which the review lists among the external concerns driving listener opinion. Before signing off on an AI-generated soundtrack, the questions worth putting to a vendor are whether the track is tagged on each distribution platform, who holds the rights to the output, and whether the tool was used for the whole composition or only for parts of the production process.
The marker to watch is the tagging data itself. If the share of AI-generated uploads on platforms that disclose it keeps rising while plays on those tracks stay flat or fall, the constraint on the market is audience attention and trust, not production capacity, and the value shifts to curation, labeling and human authorship. If plays start to track uploads, the economics of music production change for every company that pays for audio.
