AI Music Can Be Difficult To Detect, But There Are Some Telltale Signs

Engadget says AI-generated music has become a real problem over the past year, with Suno-generated tracks filling streaming platforms and some being falsely credited to real artists on sites like Spotify. The piece is a practical guide to recognising such songs.

For listeners, it says vocals are usually the easiest giveaway: slurred or warbled delivery, odd emphasis, no natural imperfections and little change in inflection. Instrumentals tend to lack dynamic shifts, key changes, interesting bridges or non-Western instruments, and often resemble US pop such as Imagine Dragons or Taylor Swift.

The article also warns that detection is imperfect. 'Feels like AI' has become a common insult, leading some people to wrongly label songs as generated, and Bandcamp has occasionally flagged real music as AI by mistake. Context checks, such as searching for live shows, interviews and social profiles, are suggested as a complement.

The article frames AI music as a trust problem for listeners: millions of songs are churned out for AI training and slop anthems, and many listeners would prefer not to be duped into hearing a machine-made song. It cites Tyga's critically panned $TARFACE album and the works of Spalexma as examples for separating traits that point to AI from those that merely make a song feel off.

It notes AI music can mean different things: AI may generate the instrumentals, vocals, lyrics or all three, or serve as a smaller element such as a vocal filter or one instrumental part. Spotting tells therefore varies by component.

For checking an artist, it recommends using platform tools such as Spotify's Artist Profile Protection or Deezer's detector, then searching for live performances, social media profiles, interviews and contact info. It admits this method may not work for bands that only dabble in AI.