How to find new podcasts when recommendations stop working

There's a moment every few months when your podcast app stops surprising you. The "you might also like" box keeps showing shows you already follow. The recommendations all sound vaguely similar. The catalogue stops feeling alive. That's the algorithm telling on itself — it has learned your taste so well that it can't push you anywhere new.

The fix isn't a better algorithm. It's signal from outside the algorithm.

TL;DR

• Your recommendations got worse because the model finally learned your taste and ran out of safe guesses.
• Search by topic, not by show title. AI search engines beat in-app search for "find me a podcast about X."
• The credits, show notes, and guest lists of podcasts you already love are the highest-hit-rate source of next listens.
• Friends with overlapping but distinct taste out-perform any algorithm on the things that matter.
• A second podcast app — used just for browsing — refreshes your queue without committing to a full switch.

Why your podcast recommendations stopped working

The honest answer: recommendation engines work best when they have room to grow. The first few weeks of any podcast app are interesting because the algorithm has barely any signal, so every recommendation is a guess that might surprise you. After a year of listening, the same engine knows exactly what you finish, what you skip, and what you bail on after thirty seconds.

The recommendations become accurate predictions of what you'll tolerate, which is not the same as what you'll love.

You also hit catalogue ceilings. If your app sources from a directory of millions of shows, but your taste maps to a few hundred, the model converges on a tight loop. Fresh shows get filtered out because they don't have enough signal yet. Old favourites get filtered out because you've already heard them. The window of "things you haven't heard but would like" shrinks every month.

The fix is to introduce signal from outside the model. Manual curation, friend recommendations, topic-first search, anything that isn't your own listening history feeding itself.

Search by topic, not by show title

The default podcast search bar is built around show titles. Type "true crime" and you get a list of shows with "true crime" in the name. Useful when you know the show you're looking for. Useless when you're trying to discover.

Modern podcast apps are slowly fixing this. Smart Topics in Podtastic surfaces topics that recur across the shows you already follow, so "the same idea covered on three different podcasts" becomes a single browse-able view. Snipd does something similar. Pocket Casts has chapter search across episodes. The point is to flip the search from "show name leads to episodes" to "topic leads to episodes regardless of show."

The same flip works on the open web. Ask Google's AI Mode or Perplexity "which podcasts have covered the Anthropic IPO most clearly?" and you get an answer ranked by transcript relevance, not show popularity. The hit rate on specific queries is much higher than the in-app search bar.

Our guide to AI search and podcast discovery covers this shift in detail, including which transcripts get indexed and why some shows are invisible to AI search.

Mine the credits and show notes of the podcasts you already love

This is the most underrated technique in podcast discovery. Every podcast you love was made by a small team of producers, editors, and writers, most of whom have worked on other shows. The credits at the end of an episode often name them. Their other projects are statistically much more likely to land for you than any random algorithm pick.

Show notes are the same story. Hosts cite other shows constantly: "this is one of those episodes that reminded me of [X]" or "if you liked this, the [Y] series did the same idea from a different angle." Read the show notes for the last five episodes of your favourite podcast. You'll find three new shows worth trying.

Substack and host blogs are the next layer. Many podcast hosts maintain newsletters where they recommend shows they're listening to. Subscribe to one or two and you'll have a steady drip of curated picks from people whose taste you already trust.

Follow the people, not just the shows

Shows come and go. Hosts and producers move between shows. If you loved a podcast that ended, the team almost certainly went somewhere else. Reply All wound down in 2022; its hosts and producers now run Hyperfixed, Search Engine, and a handful of other projects. Following the people, not the show, gives you continuity of taste across format changes.

The same applies to network founders and editorial leads. Roman Mars at 99% Invisible has a whole network of related shows. Ira Glass at This American Life has spun out Serial, S-Town, and others. Once you find one host whose taste maps to yours, follow the whole ecosystem they built.

A two-minute search for "[host name] new podcast" on Google or LinkedIn usually surfaces the next thing they're working on before your podcast app's recommendation feed catches up.

Ask a friend — the most underrated discovery channel

If you have three to five friends with overlapping but distinct podcast taste, ask them what they're listening to this month. The hit rate beats any algorithm. The reason is obvious: friends know the context of your life, what you find boring, and which shows you've already heard. The model only knows what you finished.

The trick is making it a regular ritual, not a one-off. "What are you listening to?" once a quarter, ideally with a quick "why?" attached, builds a flow of recommendations you'd never have found otherwise. Some people run a shared Notes file or a group chat thread for this. The format matters less than the cadence.

If you don't have friends who listen as obsessively as you do, the next-best thing is following recommendations from named writers and curators. Podyssey, Podcast Reviews, and a handful of newsletters all curate weekly picks from real humans rather than algorithms.

Use AI search engines to surface specific episodes

This builds on the topic-first idea. AI search engines are now genuinely useful for podcast discovery, especially when the query is specific.

"Which podcast episodes covered the Anthropic IPO in depth?" returns a ranked list with citations. "Find me an episode where someone explains the SpaceX index inclusion rule clearly" surfaces specific episodes from shows you might not follow. The hit rate is much better than browsing alphabetised lists.

The catch: AI search works on transcripts, so shows without good transcripts are invisible. Most major podcasts now have transcripts. Indie shows on smaller hosting platforms sometimes don't. If you can't find what you're looking for, the show may simply not be indexed yet.

Try a different podcast app's discovery surface

Every podcast app has a different idea of what discovery should look like. Apple Podcasts leans on editorial picks and category charts. Spotify leans on algorithmic recommendations layered with their own original content. Pocket Casts has a hand-curated Discover wall. Overcast surfaces shows your friends recommend. Podtastic surfaces recurring topics across your existing library via Smart Topics.

Even if you don't switch apps, install a second one and use it just for browsing. The friction is low. A five-minute browse in a different app once a month exposes you to shows your main app never would. Our overview of the best podcast apps covers the discovery angle in each.

If a browse session turns into a permanent move, see our guide to switching podcast apps without losing your subscriptions for the OPML transfer process.

Build a steady habit of refreshing

The compound effect of these techniques is what matters. Any single one will surface a few shows. Doing two or three of them regularly turns podcast discovery from an occasional frustration into a quiet background process. The reader version of the same idea: read three book reviewers regularly and you never run out of books to read.

For podcasts, that might look like: AI search when you have a specific topic, friend recs once a quarter, show-note mining whenever an episode ends with credits, and a once-a-month browse in a second app. The flow keeps your queue alive without depending on any single source.

Frequently asked questions

How often should I refresh my podcast recommendations?

A rough heuristic: if you can't remember the last time you started a new show, it's time. For most listeners that's every two to three months. Going much longer than that usually means the algorithm has trapped you in a loop.

What if my podcast taste is genuinely niche?

Niche tastes benefit the most from topic-first search and friend recommendations. Algorithms struggle with small audiences because the signal-to-noise ratio is bad. Manual curation has no such limit, and AI search engines are surprisingly good at long-tail "find me a podcast about [obscure thing]" queries.

Are paid podcast recommendation services worth it?

Depends on how much you value time over money. There are a few human-curated podcast newsletter services that work well. For most listeners, the free options — friend recs, show note mining, AI search, a second app for browsing — get you most of the way there.

What about asking ChatGPT or Claude for podcast picks?

They're decent at "give me 10 podcasts about X" prompts but tend to recommend the obvious popular shows. They're better when you give them constraints: "find me a podcast about X that isn't well-known, where the host has a background in Y." The more specific the prompt, the better the picks.

Listen smarter with Podtastic

Want a player that does the thinking for you? Podtastic is a fully featured podcast player for iOS and Android, built around Pod-telligence (the AI features) and Audio Enhancements (deterministic DSP tuned for spoken-word audio):

• Smart Summaries — AI summaries of every podcast and episode so you know what's coming before you hit play
• Smart Topics — key topics surfaced across your favourite shows so you can jump straight to what matters
• Smart Playback — your queue fills itself based on what you actually listen to
• Smart Jump Ahead — auto-skips commonly-skipped sections of an episode (intros, recaps, asides), powered by AI topic detection plus aggregated listening data; a single tap on any control surface jumps you to the next Smart Topic on demand
• Skip Silence — auto-removes silences from speech so episodes flow without dragging
• Enhance Voices — a gentle EQ and compression preset that keeps voices clear in any room

Join the waitlist at podtastic.app to get early access.