The current state of the art is retrieval + summarization, all that knowledge it was trained on still exists. When performing a search having no reference for the knowledge is a decent signal that it may not exist at all and you may be talking to a liar.
Never used Spotify on a mobile device though, so the location scripts will likely not be interesting at all.
On the topic of ChatGPT code projects, way I recently made another evening side project[1] using ChatGPT. I found a really nice pattern to use with it is to use one commit per ChatGPT iteration (including commits where it breaks the program, just don't push to main until it's good again). And in each commit, I store the full prompt or reply I said to ChatGPT as a prompt.txt[2]. I'll probably tack it onto the commit description next time for ease of reading. But other friends have found this really useful to be able to see exactly how I+ChatGPT evolved the software with each commit, and I can look back and reference useful prompting patterns I used.
[1] https://github.com/dag10/timelapse [2] https://github.com/dag10/timelapse/commits/e77d11baaaf4e2a5f...
Also it's pretty apparent since Spotify suggests different playlists depending on time of day, weather, etc.
If you listen to a specific genre, it's gonna work great. If you tend to listen to unrelated tracks, your recommendations are gonna be shit. If you listen to a couple of albums that are really not what you listen to usually, it's gonna take you a while to get rid of similar tracks from your recommendations.
So to answer your question, it's more of a snapshot of the mood of the tracks you've interacted with recently, not your mood specifically.