What synthetic DJs do
YouTube Music and Spotify have introduced features that entrust the presentation of tracks within playlists to an AI-generated host, with brief interjections between songs. The Wired journalist, who uses YouTube Music as her main service and enables the experimental features available through YouTube Labs, recounts keeping "Beyond the Beat" active for months before turning it off. According to the account, YouTube launched the experimental DJ in 2025, while Spotify had offered a similar tool as early as 2023 and in the following November expanded the available personalities and the controls over the commentary.
The voice works, the text doesn't
The journalist's criticism does not concern the audio quality of the voices: both companies, she writes, produce convincing, human-sounding voices. Her rejection stems from the content of the commentary. A segment devoted to a Charli XCX track is described as a long paragraph of empty phrases, with the artist's name, moreover, mispronounced. In another case, the DJ explained in depth the meaning of a remix of Olivia Rodrigo with Chappell Roan, only to then play the original version without the second artist, contradicting what had just been said.
On the Spotify side, the journalist activated the premium subscription to try the beta version of the feature and listened to the introductory clips for the eight English-language voices. Several voices use filler words such as "you know" and "like," a choice that, according to her account, aims to make the speech sound more human but produces a sense of unease. Spotify's commentary seemed to her drier and more tolerable than YouTube's, though not pleasant enough to justify returning to use.
An adoption still to be verified
The journalist attributes to Spotify the statement, contained in the company's launch blog, that 80 percent of listeners return to the DJ week after week; this is a figure provided by the company, not independently verified. The material does not clarify how many users actually use the features or how they are distributed across the two platforms.
The article's conclusion is a personal choice: both features are optional and can be avoided, and the journalist decided to turn them off. The question of the real value of these synthetic hosts therefore remains open: the voice is credible, but the content of the commentary, between clichés and the simulation of human errors, adds nothing to the listening experience according to the test recounted.
