The Podcast Back Catalog Is Now a Searchable Database — and That Changes How Creators Should Think About Research
If you run social accounts for a living, you already know the feeling: you’re scrolling through a podcast episode transcript or half-listening to an interview, and someone says something perfect — a stat, a turn of phrase, an anecdote that would anchor a LinkedIn post, a TikTok script, or a thread. You tell yourself you’ll find it later. You never do. The podcast back catalog is where good content goes to die, not because it isn’t valuable, but because it isn’t searchable.
That’s the gap that matters here. Not “another AI news app,” not “another podcast player.” The real story is that podcast audio — one of the richest, most underutilized reservoirs of quotable material for creators — has finally become queryable at scale. When I saw that Particle — the news aggregation app that summarizes stories from multiple sources — had spun out a podcast search engine called Radar, my first thought wasn’t about news consumption. It was about workflow. Because for anyone who creates content daily, the ability to search 130,000 transcribed podcasts by entity, keyword, or semantic meaning isn’t a nice-to-have. It’s a research department.
This essay isn’t a product review in the traditional sense. It’s an operator’s look at what Radar signals about the creator economy’s next phase: the shift from “creating content” to “mining existing content for insight.” I’ll get into what the product actually does, where it fits against incumbents, what I’d borrow from it even if you never pay for it, and where I think the math breaks. Let’s dig in.
The Problem It Actually Solves: Your Content Research Is Stuck in 2015
Here’s the operational reality for most social media teams: research means one of four things. You’re either scrolling X/Twitter to see what’s trending, reading newsletters, skimming RSS feeds, or — if you’re sophisticated — using a social listening tool like Brandwatch or Mention to track keywords. None of those touch podcast audio. And that’s a massive blind spot, because podcasts have become the long-form brain of the internet.
Think about how the content cycle actually works. A founder goes on a podcast, drops a controversial take about AI replacing junior marketers, and that take becomes a LinkedIn post, a tweet, a TikTok, a newsletter item. The podcast is the source, but it’s also the hardest thing to search. YouTube has transcripts now, sure, but YouTube search is still keyword-based and often surfaces the wrong segment. Apple Podcasts search is famously bad. Spotify has made strides with its AI features, but its podcast search still doesn’t give you the granular, entity-level access that a research workflow demands.
Radar’s approach is different. According to the launch post from co-founder and CEO Sara Beykpour, the platform covers over 130,000 actively transcribed podcasts with 20,000+ episodes added daily. Each episode is fully transcribed, speaker-diarized and labeled, and rich entities — people, companies, etc. — are extracted. That’s not a search box bolted onto an RSS feed. That’s a structured database of spoken content.
For a creator, this changes the research equation. Say you’re a marketing content creator and you want to find every time a specific CMO mentioned “community-led growth” in the last six months. With Radar, you search the entity, filter by recency, and get a list of episodes and timestamps where that person spoke about that topic. You’re not guessing. You’re not hoping the transcript happens to include the right keywords. You’re querying a database that has already done the diarization and entity extraction for you.
In my experience running social accounts, the difference between “good research” and “great research” is usually the difference between a surface-level take and a quote that makes people stop scrolling. Radar gives you a direct line to the latter. And the alerts feature — delivered via Slack, email, or webhook, either in real-time or as daily/weekly digests — means you can set up a monitoring loop for the entities and guests that matter to your niche, then get notified when they appear on a podcast. That’s the kind of thing that used to require a dedicated research assistant.
Why TikTok creators should care more than LinkedIn ones
Let me be specific about who benefits most. LinkedIn creators and newsletter writers who produce text-based commentary will find Radar useful, but their research is often already text-first — they’re reading articles, tweets, and other newsletters. The real unlock is for video creators, especially those on TikTok and Instagram Reels, where the currency is spoken word.
Here’s why: TikTok’s algorithm rewards watch time and completion rate. If you’re making a video that references a podcast quote — a hot take, a controversial prediction, a founder’s candid admission — the audio clip itself becomes your content. You don’t need to paraphrase. You need the exact 30 seconds where the guest said the thing. Radar’s semantic search means you can find that segment without knowing the exact words. The maker team confirmed this in a comment, noting that the current semantic search option matches the terminology used in the search, so “rough ideas should yield some results,” with a “smart search” option coming that will be “closer to an LLM query.” That’s the difference between searching “how to price a SaaS product” and finding the exact moment a founder said “we underpriced ourselves for two years and it almost killed us.”
For video creators, that’s gold. You’re not just researching. You’re sourcing raw material for clips. And the fact that Radar surfaces “relevant news about an entity” alongside podcast results means you can pair a podcast take with a current event — which is exactly the kind of contextual hook that drives engagement on short-form video.
How Radar Differs From What’s Already Out There
Let’s name the incumbents, because that’s where the comparison gets interesting. The closest analog to what Radar is doing is probably Podscribe, which offers podcast advertising analytics and some transcription search. There’s also Listen Notes, which has been the go-to podcast search API for years, and Chartable (now owned by Spotify), which focuses on podcast analytics and attribution. For the creator who just wants to find a quote, there’s also the manual route: open YouTube transcripts, Ctrl+F, and pray.
Here’s where Radar differentiates itself, based on what the launch post describes. First, the scale and freshness: 130,000 actively transcribed podcasts with 20,000+ episodes added daily is a meaningful claim. Listen Notes has a larger index, but Radar’s focus on active transcription — not just metadata — means the content is actually searchable. Second, the entity extraction and speaker diarization. That’s not trivial. Knowing who said something, not just what was said, is the difference between a search engine and a research tool. Third, the API/MCP access. Beykpour explicitly frames this as a tool for “a world increasingly searched by agents” — the idea that AI agents need to be able to query podcast content programmatically. That’s a forward-looking bet, and it’s the right one.
But the most interesting comparison isn’t to another podcast tool. It’s to the news aggregation space where Particle already plays. The launch post mentions that Particle News has been bringing podcast clips into news stories, giving each story “a very human layer of commentary and discussion.” That’s a workflow that Techmeme — which Chris Messina brings up in the comments — has been doing manually for years. The idea that you could automate the “find the best commentary on this news story” loop is powerful. Messina asks whether Particle could “spin up a podcast-focused competitor with Techmeme,” and Beykpour’s response — that they’re building a Trends product on Radar — suggests they’re thinking exactly that way.
Where the math breaks
Let me do the honest math, because there are real limitations here. First, the pricing. Radar is free to try, then $29 per month for Individuals and $399 for Businesses. That’s not cheap. For a solo creator, $29/month is a meaningful line item — roughly the cost of a Canva Pro subscription or a decent Buffer plan. The promo code HUNT gets you 50% off the first month, but that’s a trial discount, not a long-term solution.
The bigger question is whether the search quality justifies the cost. The launch post mentions that Radar’s current search is semantic — matching terminology — with a “smart search” coming that will be more LLM-like. That’s an admission that the current version is still evolving. In my experience testing similar tools, semantic search on transcripts is only as good as the transcription quality and the entity extraction. Podcasts with heavy accents, overlapping speech, or technical jargon are going to produce messy transcripts. The team claims 130,000 podcasts are “actively transcribed,” but transcription quality varies wildly across the industry.
There’s also the question of podcast coverage. 130,000 is a lot, but it’s not everything. Niche podcasts, especially those in emerging fields, may not be in the index. And the freshness claim — 20,000+ episodes added daily — is impressive, but it also means the index is growing faster than the search quality can be validated. I’d want to test this against my own niche before committing to a subscription.
What Creators and Social Media Teams Can Borrow From Radar (Even If You Never Pay)
Here’s the part where I get practical, because even if Radar isn’t the right tool for your budget, the workflow it enables is something every creator should adopt. The core insight is this: podcast audio is a content mine, and you should be mining it systematically.
Start with your niche. Identify the top 10 podcasts that your audience listens to. These are your primary research sources. Then, set up a weekly research block — 30 minutes, same time every week — where you search for mentions of your key topics, competitors, and industry terms. If you have Radar, use it. If you don’t, use Listen Notes or manually check YouTube transcripts. The point is to build a habit of finding quotable moments before you need them, not after.
The second thing to borrow is the entity-based approach. Most creators think in keywords. Radar thinks in entities — people, companies, topics. That’s a more powerful mental model. Instead of searching “AI marketing,” search for the specific founders and CMOs who talk about AI marketing. Follow them, not the topic. That’s how you build a source list that compounds in value.
Third, set up alerts. Even if you’re not using Radar’s alert system, you can replicate it with Google Alerts or Feedly for text-based mentions. The goal is to be notified when your key entities appear in new content, so you can react quickly. Speed matters in social media. The creator who jumps on a podcast take within 24 hours gets the engagement. The one who finds it a week later is late to the party.
The API angle: why this matters for AI-native workflows
There’s a deeper layer here that most creators haven’t fully internalized yet. Beykpour’s comment about “a world increasingly searched by agents” isn’t marketing speak. It’s a signal about where content discovery is heading. As AI agents become more capable — think ChatGPT with browsing, or Perplexity with its answer engine — the ability to query podcast transcripts programmatically becomes a competitive advantage.
For creators, this means two things. First, your content should be structured so that agents can find it. That means transcripts, show notes, and metadata that are clean and searchable. If you’re a podcaster, this is an argument for investing in high-quality transcription services and entity tagging. Second, it means the tools you use should have APIs. Radar’s API/MCP access — bundled into each tier, according to the launch post — is a sign that the product is built for agentic workflows. If you’re building any kind of automated content pipeline, that’s a feature you should care about.
I’d bet we’re going to see more tools follow this pattern: consumer-facing search interfaces paired with API access for power users. The creator economy is becoming an API economy, and the creators who understand that shift will have a structural advantage.
Where I’m Skeptical: The Product Is Young, and the Use Case Is Narrow
Let me be clear about my reservations, because any honest review needs them. First, the product is a search engine, not a content creation tool. It will help you find podcast moments, but it won’t help you turn them into content. You still need to write the captions, edit the clips, and design the thumbnails. Radar is a research layer, not a production layer. If you’re looking for a tool that automates content creation, this isn’t it.
Second, the pricing structure — $29/month for individuals, $399/month for businesses — is a bet on professional use cases. For a solo creator, that’s a significant investment, especially when the free tier exists. The question is whether the free tier is enough for occasional research, or whether the paid tiers unlock features that make the subscription worthwhile. The launch post doesn’t specify what’s included in the free tier, which is a transparency gap.
Third, the competitive landscape is crowded and moving fast. Spotify has been investing heavily in podcast AI features. Apple is improving its podcast search. And YouTube already has transcripts built in. Radar’s differentiation — entity extraction, speaker diarization, alerts, API access — is real, but it’s also a feature list that incumbents could replicate. The moat here is the transcription pipeline and the entity extraction quality, not the search interface.
Finally, there’s the question of who this is not for. If you’re a casual podcast listener who just wants to find a specific episode, Radar is overkill. If you’re a creator who works primarily in text-based content and doesn’t reference podcast audio, the value proposition is thinner. And if you’re a brand doing social listening at scale, you might be better served by a dedicated listening platform that covers podcasts alongside social channels.
The “half-remembered moment” problem
There’s a specific use case that the comments highlight, and it’s worth addressing because it’s so relatable. One commenter, Jean-Noël Escande, describes it perfectly: “I have so many half remembered podcast moments I gave up trying to find again.” That’s the emotional core of this product. We’ve all been there — you know you heard something, but you can’t remember which episode, which podcast, or the exact wording.
Radar’s semantic search is designed for this. You don’t need the exact quote. You need the rough idea. The maker team’s comment confirms this is a deliberate design choice. And the upcoming “smart search” — which will put an LLM in front of the query — suggests they’re doubling down on this use case. For creators, this solves a real pain point: the “I know I heard it somewhere” research problem.
But here’s my honest take: I’d want to test this before trusting it. Semantic search on transcripts is notoriously finicky. The quality of the results depends on the transcription accuracy, the entity extraction, and the underlying search algorithm. If the transcription is bad, the search will be bad, regardless of how good the UI is. The fact that the team is transparent about the current limitations — and the roadmap for improvement — is a good sign, but it’s also an acknowledgment that the product is still maturing.
What I’d Watch / Test Next
If you’re an operator reading this, here’s what I’d do this week, in order of priority:
Try Radar’s free tier for a specific research task. Don’t just search your own name (we all do that). Pick a content pillar — say, “creator economy trends” — and search for recent podcast mentions. See how many results come back, how relevant they are, and whether the timestamps and entity labels actually help you find the right segment. That’ll tell you more than any review.
Set up alerts for your top 3 competitors or industry voices. Use Radar’s alert system (or replicate it with Google Alerts) and see what surfaces over a week. The goal isn’t to read everything — it’s to build a signal-to-noise baseline. If the alerts are mostly irrelevant, that’s useful information too.
Audit your own podcast content for searchability. If you have a podcast, check whether your episodes are being transcribed and indexed. If they’re not, that’s a gap you should fix. The creator economy is becoming searchable, and the creators who aren’t indexed are invisible to agents and researchers.
Watch the Trends product. Beykpour mentioned that a Trends product is coming, and that’s the most interesting thing on the roadmap. If Radar can surface what podcasters are talking about right now, that becomes a real-time content ideation tool. I’d bet that’s where the real value will be — not in search, but in trend detection.
Compare against your existing research stack. Run the same search in Radar, Listen Notes, and YouTube transcripts. Note the differences in result quality and speed. That comparison will tell you whether Radar is a replacement, a complement, or a nice-to-have.
The bigger picture here is that podcast content is becoming a first-class data source for creators. The tools that unlock that data — whether it’s Radar or something that comes after — will change how we research, source, and create. The creators who start building podcast-mining workflows now will have a head start when this becomes table stakes.
I’m not ready to say Radar is the definitive tool for this. The pricing is steep, the product is young, and the competitive landscape is shifting. But the direction is right, and the workflow it enables — searchable, entity-based, alert-driven podcast research — is one every serious creator should adopt. Whether you use Radar or build your own version with existing tools, the future of content research is spoken-word search. Start mining.
