Why curated radio beats algorithmic music

There are two ways to find new music in 2026. You let an algorithm pick. Or you trust another person.

Spotify and the other streaming services have spent a decade getting very good at the first. Their recommendation systems can map your listening with uncomfortable accuracy. And yet for many people the experience of listening through them feels worse than it did five years ago. Why curated radio beats Spotify is not a romantic question. It is a practical one about how taste actually works.

This is a piece in defence of human curated radio. The best curated radio stations are doing something an algorithm cannot do, and it shows.

Quick picks for human curated radio:

• For broad discovery: KEXP, BBC Radio 6 Music
• For electronic depth: Dublab, Refuge Worldwide
• For jazz selection: Jazz24, FIP
• For community-led music: NTS Radio, Cashmere Radio

What an algorithm is actually doing

Spotify, Apple Music, YouTube and the rest run on engagement signals. The algorithms reward songs that hold your attention, that get added to playlists, that finish without a skip. They penalise songs that get skipped early. Over time this pulls every recommendation toward the same shape: front-loaded, hooky, safe, short.

This works for the company, because engagement maps to retention. It does not necessarily work for you. The system has no model of taste. It has a model of attention.

A human DJ has a model of taste. A good DJ plays a track that loses your attention in the first 30 seconds because the third minute is where the song lives.

What a curator does that an algorithm cannot

Three things, mostly.

Context. A human curator can sequence two songs that should not work together and make them work, because they understand why the contrast matters. FIP does this all day, every day. A piano improvisation into a Malian guitar piece into a 1970s soul track. The algorithm would never play that sequence because it does not score well in isolation. Heard together it is the most interesting music programming on the internet.

Risk. A curator can play a track they know is going to lose half the audience, because the other half is going to fall in love. The algorithm cannot take that risk because half the audience leaving is the worst possible outcome for the metrics. So everything trends bland.

Voice. A curator has a perspective. Dublab sounds like Dublab. KEXP sounds like KEXP. You learn to trust the station the way you learn to trust a friend who keeps lending you good books. The algorithm has no voice and never will.

Why human curated radio is better for discovery

The standard argument for algorithms is discovery. Spotify Discover Weekly was supposed to be the killer app for finding new music. In practice it tends to recommend things adjacent to what you already listen to, which is the opposite of discovery. It is recognition with new packaging.

Real discovery is finding something you would never have searched for. That requires someone to put it in front of you on purpose. KEXP is built around exactly this principle. Their DJs play artists you have not heard because they have decided you should hear them. Not because a model predicted high engagement.

The same is true of BBC Radio 6 Music, Refuge Worldwide out of Berlin, Cashmere Radio, Tsugi Radio and most of the community stations in the Alternative and Indie categories on WRC.

What you lose with curation

Honest counter argument: curation is slower. You will not hear the song you want when you want. You will sometimes get a track you actively dislike. The catalogue is whatever the station has, not whatever you can think to type into a search box.

This is real. But it is also the point. Limiting choice creates attention. The reason a great radio station feels engaging is because you are not in charge of what comes next. Someone else is. That puts you in a different listening posture, more open, less directive.

Streaming services know this. That is why every one of them has launched a "radio" feature in the last five years. They are trying to manufacture the feeling of curation without the cost. It does not quite work, because the underlying engine is still chasing engagement signals.

Best curated radio stations to try

If you want to start with the strongest curated stations on WRC:

FIP for everything. Genuinely broad, no ads, almost no chatter, professionally programmed by the French public broadcaster.

KEXP for indie, alternative, world and discovery. The DJs are the product.

BBC Radio 6 Music for a UK perspective on deep catalogue and new releases.

Dublab for experimental, ambient and underground. Less polished, more interesting.

Refuge Worldwide and Cashmere Radio for community-run Berlin programming.

Jazz24 and Linn Jazz for the jazz end.

Where to find more

The Focus & Flow vibe is a good starting list of curated stations that work as background and as discovery. For genre-specific curated picks, browse Jazz, Alternative, Electronic, and Experimental.

There is also a longer piece on the history of internet radio that traces how human curated radio survived the streaming era.

Frequently asked questions

Is curated radio better than Spotify? For discovery and depth, yes. Curated radio gives you music chosen by people with taste, not by engagement models. For on-demand listening to a specific song, streaming services still win. They solve different problems.

What are the best curated radio stations online? KEXP, BBC Radio 6 Music, FIP, Dublab and Refuge Worldwide are widely considered among the best. All stream free on WRC.

Why does Spotify feel worse than it used to? Because the recommendation engine optimises for engagement, which pulls every playlist toward similar, safe selections. The catalogue is bigger than ever but the recommendation surface is narrower.

Open FIP for an hour and notice how the listening posture changes. That is the case for curation in one experiment.