← All use cases

Use case P36 · Human

Picky music listener uses Gemini to describe songs she likes and find new tracks for her playlists

Megan Ellis ·

The short version

What they did

Megan Ellis tried Gemini for two weeks to find individual songs rather than broadly similar artists. She turned a playlist screenshot into a table and used Gemini’s descriptions of what she liked to narrow later requests. She found new tracks she enjoyed, but also reports nine invented song recommendations and time wasted searching for them. This was her own listening experiment, not a Spotify integration she built.

Result (as claimed)

“Holy Mother in particular has become my new favorite track, which I listen to obsessively.”

“Through other prompts, I found a total of six more songs that I enjoyed, with another landing on my Liked playlists and the others ending up on other playlists.”

Prompt

Adapted from her method, not her published wording:

I’m a picky listener looking for individual songs, rather than broadly similar artists. I’ll attach [playlist screenshot or song list]. My notes on what I like and dislike are [voice, instruments, rhythm or mood].

1. Turn the list into a table of song titles and artists for me to check. Mark unreadable titles rather than guessing.
2. Ask about three songs, then describe the musical qualities I seem to enjoy. Let me correct that description.
3. Suggest five tracks based on those qualities, with an official artist or streaming page for each.

Keep the reasons short and use everyday words rather than unexplained music terms. Verify the title and artist of each suggestion. If you cannot confirm a song exists, say so and leave it out instead of inventing one.

Do it yourself

These steps are my suggested way in, not Megan’s setup.

What you need

  • A chat assistant that can read screenshots and search the web, such as Gemini, ChatGPT or Claude.
  • A screenshot or list of songs you like.
  • Your music app.

Start simple

  1. Upload a screenshot of a playlist and ask for a table of titles and artists. Check it.
  2. Pick three songs and say what you like about each: the voice, the drums, the mood.
  3. Paste the prompt and ask for five individual tracks with links.
  4. Search each one in your music app before you play it.
  5. Tell it which ones worked and why, and ask again.

Take it further

  • Keep a running note of what you like and don’t, and paste it at the start of each session.
  • Ask for songs that share one quality, such as a certain tempo, across genres you don’t usually hear.
  • Save the instructions in a project or Gem so you don’t repeat yourself.

Check your work

  • Search every title and artist yourself. Megan got nine songs that didn’t exist.
  • If a link goes to a different song, treat the recommendation as made up.

Source checked. The link, author, date and quote were confirmed against the source. The result itself has not been reproduced. Prompt adapted for this library, inspired by the case. Business size: Individual. Results are quoted as the source claimed them and never improved. Spot an error or a broken link? Email charlie@cbarmorecpa.com.

More in Music

New cases go out every Saturday in Inside AI.