What happened
BGR reports that a 2025 study conducted by Deezer and Ipsos tested whether listeners could distinguish AI-generated music from human-made music, with only 3% reportedly succeeding. BGR also reports that Deezer said approximately 90,000 AI-generated tracks were uploaded daily in July 2026, representing more than half of all new uploads on Deezer. The article says Deezer tags AI music and excludes fully AI-generated tracks from algorithmic recommendations, while Beatport has banned AI-generated music. The study methodology, sample size, underlying data and independent confirmation are not provided in the source.
BGR reports that Deezer partnered with Ipsos in 2025 to study whether people could identify AI-generated music. According to BGR, only 3% of listeners could tell the difference between AI-generated and human-made recordings. The source does not identify sample size, participant demographics, listening conditions, genres, test materials, statistical uncertainty or publication venue. Those omissions make it impossible to determine how representative the result is or whether it applies to all types of AI-generated music. The finding is therefore best treated as a reported result from a study, not as a definitive measure of listeners’ ability in every setting.
BGR reports that Deezer said in July 2026 that about 90,000 AI-generated tracks were uploaded to its service each day, accounting for more than 50% of all new music uploads on the platform. This is the article’s clearest indication of scale, but the report does not provide a link to Deezer’s underlying data, explain how Deezer classified a track as AI-generated, or say whether the figure includes partially AI-assisted recordings. It also does not establish how the proportion compares with other streaming platforms. The claim is consequently attributed to Deezer as reported by BGR and is not independently confirmed in the supplied source.
The article describes several platform responses. BGR says Deezer launched an AI-music tagging system, excludes fully AI-generated tracks from algorithmic recommendations and withholds royalty payments from AI artists that commit fraud or manipulate streams. BGR also reports that Beatport has banned AI-generated music and partnered with Beatdapp, a company focused on detecting streaming fraud. Hangout FM, Audiomack and Napster are identified as other Beatdapp partners. The source does not provide the policies’ effective dates, enforcement details, appeal procedures or evidence of how often detection systems correctly identify synthetic recordings.
BGR includes practical listening advice drawn partly from Reddit users and a reference to Forbes. It says listeners may notice dull or low-fidelity high-frequency sound, although those characteristics can also result from ordinary compression or poor-quality audio. It says spectrograms may sometimes reveal clues and that unusually rapid popularity, a limited artist history or questionable credits may be more useful indicators than sound alone. These suggestions are not presented with controlled test results in the source, and BGR explicitly says listeners are unlikely to identify AI music by ear with complete reliability.
Read the primary source: bgr.com ↗
Why it matters
The figures, if accurate, suggest that ordinary listening may not reliably reveal whether a recording was made by people or generated with AI. That could affect music discovery, artist attribution, royalty systems and the ability of platforms to distinguish genuine audience interest from automated or manipulated activity. BGR also describes AI-generated music as a potentially profitable source of streaming revenue, but its examples use simplified minimum-revenue calculations and do not establish actual earnings or fraud.
If BGR’s account of the Deezer-Ipsos result is representative, it highlights an information problem in streaming: a listener can enjoy a recording without knowing who or what produced it. That matters because attribution is how audiences choose artists, how musicians build reputations and how platforms organize discovery. When the audio is difficult to distinguish, labels, credits and platform disclosures become more important. The source does not show whether listeners were told that some tracks might be AI-generated, whether they had access to artist information, or whether disclosure changed their judgments.
The reported upload volume also raises questions about catalogue quality and attention. BGR says AI-generated tracks made up more than half of Deezer’s new uploads in July 2026, but upload share is not the same as listening share. A large number of uploads could have little effect on what audiences hear if recommendation systems, search rankings and editorial choices filter them out. Conversely, if synthetic tracks receive substantial exposure, they could compete with human-made music for discovery and royalty pools. The article supplies no platform-wide listening data that resolves this distinction.
BGR portrays AI-generated music as financially attractive and gives streaming examples involving an AI-generated rock band and a track called “Dust on the Wind” by The Velvet Sundown. It calculates minimum possible revenue from listener or stream counts by assuming a payment floor, but the source does not provide contracts, payout statements, geographic breakdowns, repeat-stream data or evidence that the examples were entirely AI-generated. The figures should not be read as verified earnings. They illustrate the incentive described by BGR, while leaving the commercial scale uncertain.
The issue also intersects with fraud and consent. BGR says Deezer denies royalties to AI artists involved in fraud or stream manipulation, and that Beatport uses detection tools associated with streaming-fraud prevention. Those measures address platform integrity, but they do not answer questions about whether training data included copyrighted recordings, whether vocal or musical identities were imitated, or how creators should be credited. The article does not report a court ruling, licensing agreement or regulatory action resolving those questions.
BGR says AI-generated music is legal, but the source does not define the jurisdiction or distinguish between generation, distribution, copyright protection, voice imitation, training-data use and deceptive streaming. Legal treatment can differ across countries and fact patterns. The practical significance of the report is clearest as a question of transparency and platform governance, not as a settled statement of law. Readers should not infer that every form of AI-generated music is lawful everywhere.
What to watch next
The key questions are whether the Deezer-Ipsos result can be independently replicated, how the study defined AI-generated music, and whether listeners performed differently across genres, production quality levels or listening conditions. Platforms’ labeling, recommendation and royalty policies will also determine whether AI-generated tracks become visible to audiences or remain largely indistinguishable within catalogues. The source does not establish how widely Deezer’s measures work, whether other services use comparable systems, or how legal and licensing rules vary across jurisdictions.
The first priority is verification of the underlying study. Follow-up reporting would identify the Deezer-Ipsos methodology, number and selection of participants, track set, genres, listening equipment, instructions and success rates by subgroup. It would show whether the 3% figure measures unaided detection, forced-choice guessing or performance after repeated listening. Without those details, the result cannot be compared fairly with other research or generalized to current, higher-quality generation systems.
The second priority is independent measurement of supply and demand. Deezer’s reported 90,000 daily uploads and more-than-half share need definitions, time series and platform-level context. reporting should distinguish fully generated tracks from music made with AI assistance, uploads from tracks that receive plays, and legitimate releases from suspected stream manipulation. Comparable figures from services would show whether Deezer is an outlier or part of a broader shift.
Platform controls deserve scrutiny. BGR reports that Deezer tags AI music and excludes fully generated tracks from recommendations, while Beatport bans such music. questions include whether labels are visible before playback, how mixed human-AI works are handled, how detection errors are corrected, and whether artists can challenge a classification. The source does not report how systems perform, so their effect on discovery and creator income remains unknown.
The economics should be watched carefully. Stream counts and listener totals do not by themselves establish revenue, and BGR’s examples rely on minimum assumptions. Auditable information about payout rates, repeat streams, geography, platform fees, fraud investigations and removed tracks would be needed to assess whether AI-generated catalogues are genuinely profitable. It will matter whether platforms alter recommendation rules or royalty eligibility as synthetic uploads increase.
Finally, the public-policy questions remain open. The source does not report new legislation, a regulatory decision or a court case, and it does not establish a uniform legal standard. developments could involve disclosure requirements, copyright and training-data disputes, protections for recognizable voices or musical styles, and rules for synthetic performers. Until those questions are addressed with jurisdiction-specific evidence, listeners and creators should treat platform labels as useful signals rather than complete proof of how a recording was made.
