Audio has always been the quiet backbone of brand identity. Jingles, sonic logos, and scored campaigns carry emotional weight that visual identity often cannot. For decades, the craft of music production served as a natural filter: expensive, time-consuming, and dependent on human expertise. That filter is dissolving faster than the industry anticipated.
A new study from Deezer and research firm Ipsos has put a precise number on the problem. In controlled listening tests, 97 percent of participants could not reliably identify AI-generated music as artificial. The sample covered a broad range of genres and listener demographics, making the result difficult to dismiss as an edge case. It is a signal that AI audio generation has crossed a perceptual threshold the industry was not fully prepared for.
The study placed listeners against tracks generated by current AI music systems alongside human-produced recordings. Participants rated authenticity and emotional resonance. Across nearly all categories, AI tracks scored within the margin of error of human compositions. The 3 percent who flagged AI content correctly did so inconsistently, suggesting guesswork rather than genuine detection ability.
Deezer responded by announcing it is now licensing its internal AI detection tool to third parties. The system, built to flag synthetic audio in the streaming pipeline, uses spectral analysis and generative artifact recognition to identify AI-produced content at the upload stage. The company has not disclosed pricing or the full client list, but the move signals that detection infrastructure is itself becoming a commercial product category.
The Perceptual Gap Has Closed
For most of 2023 and 2024, the creative industry operated under an assumption: trained ears could catch AI audio. Artifacts, unnatural reverb tails, rhythmic stiffness. That assumption no longer holds. The 97 percent figure means even engaged, music-literate listeners are operating blind. Agencies that once relied on acoustic authenticity as a differentiator cannot count on audiences to notice the difference.
What This Does to Audio Branding
If listeners cannot detect synthetic origin, the conversation shifts from detection to intention. Brands investing in sonic identity now face a legitimacy question that has nothing to do with quality. A hand-crafted score and an AI-generated one may sound identical to 97 percent of consumers, but the process carries different legal, ethical, and reputational weight. Audio branding studios will need to make the provenance of their work visible, not just the output.
Detection as Infrastructure
Deezer licensing its detection tool is a structural move, not a product launch. It positions the company as a standards layer for the audio industry, the way brand safety tools became infrastructure for display advertising after 2016. If major platforms and ad networks adopt a common detection standard, creative agencies will need to certify AI content before distribution, adding a compliance step to audio production workflows that does not currently exist.
The Licensing Race Tightens
Headline 11 in this same news cycle covers Sony, Warner, and Universal signing licensing deals with AI music startup Klay. The timing is not coincidental. When a study proves consumers cannot tell the difference, the commercial pressure on labels to license rather than litigate intensifies immediately. Agencies working with licensed music libraries should expect the terms and provenance of those libraries to become a more active point of negotiation within the next 12 months.
Creative Accountability Gets Harder to Dodge
The 97 percent figure will be cited in boardrooms as permission. It will also be cited in legal proceedings as evidence of scale. For creative directors and music supervisors, the practical consequence is that decisions made quietly, such as swapping a commissioned score for a generated one to meet a budget, become harder to reverse once they are normalized. The study does not resolve the ethics. It just removes the comfortable ambiguity that let the industry defer them.
Early reception among audio professionals has been a mix of resignation and urgency. Several independent music supervisors flagged the Deezer study on professional networks within hours of publication, with responses ranging from concern about client expectations to interest in the detection tool as a procurement requirement. No major agency network has issued a public response, but internal conversations are reportedly underway at several holding company creative studios.
The perceptual threshold has been crossed. The industry's next move is not technological, it is definitional. Agencies, platforms, and brands will spend the next two years deciding what authenticity means when the ear cannot arbitrate it. Those who build clear provenance standards now, rather than waiting for regulation to force the issue, will be better positioned when that reckoning arrives. The label licensing wave already underway suggests the commercial infrastructure is forming fast. The creative and ethical frameworks need to keep pace.