How to tell if a song is AI

You cannot reliably hear the difference any more. You can still check the paperwork. Here is the order we would go in, from quickest to slowest.

1. Look at when the music started

Open the artist on Spotify, Apple Music or Deezer and sort the discography by date. An act that did not exist eighteen months ago and now has hundreds of thousands of listeners is worth a closer look. Hits from nowhere do happen, but they come with a story: a TikTok moment, a label push, a famous feature.

2. Count the releases in the first few months

This is the strongest single tell. AI projects tend to upload in bursts: three albums in a month, or forty singles in a season. Writing and recording that much music by hand is close to impossible. Our reports draw this as a timeline so the burst is easy to see: compare Xania Monet or IngaRose, both documented AI projects, with a human band like Sleep Token.

3. Read the songwriter credits

On Spotify, open a song, then the three-dot menu, then View credits. Human songs credit human writers, often several. AI projects usually credit nobody, or only the act's own name. A credited human writer does not prove a song is human, since some AI projects are steered by a real person who writes the lyrics, but a total absence is a signal.

4. Look for a life outside streaming

Masks and anonymity are not evidence of AI on their own: Sleep Token and Angine de Poitrine hide their faces and are very much human, with tours and documented members. Search for live dates, interviews, a local scene, a hometown. A real band leaves traces: gig posters, photos from different angles, other musicians who have played with them. Check MusicBrainz and Wikipedia. AI acts usually have a polished profile picture and nothing else.

5. Watch the photos and videos closely

Generated images still slip at the edges: hands, teeth, jewellery, text on clothing, hair that moves strangely. Several acts that turned out to be AI, like Sienna Rose, were first spotted by fans noticing glitches like these.

6. Listen for the generator's habits

Many generated tracks share a sound: very clean, slightly smeared high end, vocals that never breathe, lyrics that rhyme generic images (fire, desire, night, light) without saying anything specific. These are hints, not proof. Plenty of human pop sounds this way too, which is exactly why audio detectors struggle. See our detector test.

7. Check what the platforms say

Deezer tags albums it detects as AI-generated content in its app. Some artists disclose AI use in their biography. Spotify has said it will show AI disclosures supplied by labels and distributors in song credits.

The quick version

Paste the artist into the checker at the top of this page. It does steps 1 to 4 for you and shows the evidence.