The hard part isn't finding you a podcast. It's giving you a reason to press play.

Spotify is where I listen to podcasts. But when I arrive with a specific need, say I've had a terrible day and I want help letting go, I leave and search YouTube instead. I wanted to understand why. That question turned into a few weekends of reading. A song costs about 20 seconds to test; an episode costs 45 minutes, and the listener knows that before pressing play. A recommendation system where mistakes are cheap becomes a much harder consumer problem when the wrong guess costs three quarters of an hour.

Why should I trust you with an hour of my time? An hour is a pretty big ask from someone I don't know. There are effectively infinite podcasts. My time isn't infinite. So the problem for me isn't really 'can Spotify find something I might like?' It's 'can Spotify give me enough confidence to actually take the recommendation?'

The more expensive a recommendation is to test, the more I want a reason to trust it. Why me? What did you see in my behavior that made you pick this? Why this? What gives me confidence this creator is worth my time? Are you actually on my side? What are you optimizing for when you put this in front of me? And if your reasoning is wrong, let me correct it.

There is more podcast content than any listener could sample in a lifetime, so abundance isn't the constraint. Spotify's prediction is world class: it can work out what a listener is statistically likely to play. What's still open is decision-making. Prediction produces a ranked list; it doesn't give a person what they need to decide whether an unfamiliar hour is worth taking. The scarce resource in podcasting isn't content, it's confidence: for listeners, that this episode is worth their time; for creators, that Spotify can connect their work with the people most likely to value it.

Choosing a stranger means reading titles from people I've never heard of, judging whether they're credible with little to go on, guessing whether the episode is really about what I need, and committing 45 minutes to find out. Rather than absorb that cost, I open a creator I already know. If that holds, the effort Spotify spends surfacing new creators can end up reinforcing the habits a listener already has, because the familiar option wins by default rather than on merit. What's missing isn't more recommendations, it's a reason to trust one. This is what I want to test, not what I claim to have shown.

Every platform lends a listener someone else's judgment. YouTube lends crowd proof ("2.1M views"): instantly legible, but it counts clicks, not fit, and it compounds toward creators who are already large. Word of mouth is social proof, and Spotify has little social graph to draw on. What Spotify could lend is cohort proof: the judgment of people who value the same thing you value in a host you already follow. It works the same for a creator with 400 listeners as for one with four million, because the cohort is defined by the reason, not the size. And Spotify can count something YouTube can't: someone finishing 45 minutes and coming back three weeks later unprompted. None of that is currently lent back to the listener at the moment of deciding.

Replace "Episodes you might like" with a reason: because you finish these, because listeners who return to a host you follow also stayed with this one. Keep the reasons behavioral ("because you finish these" is observable and checkable; "because you value candor" can feel like being analyzed). And let the listener correct it. A reason that proves right makes the next unfamiliar recommendation cheaper to accept, and a wrong one is visible enough to fix. Over time the trust being borrowed shifts from the creator to Spotify itself.

I don't need more choices. I need more confidence about which ones deserve my time. Maybe the best signal of a good podcast recommendation isn't that I pressed play. It's that Spotify introduced me to someone I chose to hear again.

This is strongest for unfamiliar, personality-led, long-form conversational podcasts, where topic tells you least and the time commitment is largest. It's weaker for daily news, sports and serialized narrative, where the subject or the story carries the return, and for shows already followed, which is continuity rather than discovery. So this is a layer on top of ranking, for one decision: whether to risk an unfamiliar hour on a person you don't know yet.

The same signal that helps a listener risk an unfamiliar hour is what an unknown creator needs in order to be found at all. A recommender built on listening volume can't say much about a creator until that creator is already large, which is the loop every emerging podcast is stuck in. But a transcript exists the moment the first episode does. Match its attributes against listeners already known to value them, and a creator with 400 listeners can be pointed at a plausible first thousand. That changes what Spotify can promise a new creator at onboarding: not a tour of publishing and analytics tools, but an answer to the question they actually arrived with, which is whether there's an audience here for them. Who are my people, why would they stay, and what should I do next. And it makes the listener product a creator-growth engine: creator investment is currently bought with advances and guarantees, which are expensive and easy to outbid; intelligence about who stays and why is a by-product of work already done for listeners.

North star: durable discovery. Did an unfamiliar creator become someone the listener voluntarily came back to, measured as unprompted return at thirty days.

Spotify built the Taste Graph: twenty years of learning what you like. Podcasting is where it can learn who you value, why, and who might matter to you next. Where I'd start: a few weeks of listener interviews before anything is built.