Podcasts are booming, but no one's really listening.
Podric is a platform for searching, monitoring and analysing what is said in podcasts. Thousands of hours of conversation hold knowledge nobody could reach, and Podric turns them into something you can search, follow and measure.
- podcasts monitored continuously
- 2,500+
- using AI transcription and analysis
- episodes indexed
- 230,000+
- dozens added daily
- active monitoring projects
- 160+
- run by media-monitoring clients on Pulse
- in media value surfaced monthly
- €4.5M
- estimated from listening data and advertising prices
Podcast content is unstructured and unsearchable
53% of Dutch adults listen to podcasts regularly (Markteffect, 2025)source 1Markteffect, Podcast Monitor, June 2025: 53% of Dutch people aged 18 and over listen to podcasts regularlymarkteffect.nl. But user research shows that there is one huge frustration among them: they cannot find relevant podcast content. Search engines index text and a podcast is audio, so podcast apps search titles and show notes only. A topic, a person or a quote inside an episode stays out of reach, and cannot be linked to in the traditional way.
Organisations cannot monitor podcast content either, leaving communication professionals without a way to extract insights from podcasts for PR, journalism, research, reputation management, compliance and market intelligence.
Podric solves both problems with one solution: an AI-driven search engine that indexes podcast content at scale, allowing for search, analysis and monitoring.
The main challenges
-
We needed full coverage
Proper media monitoring requires full coverage: every episode it misses could be crucial to a client. The index had to cover the whole Dutch-language podcast market, including the niche shows where trade conversations happen.
-
Transcription at scale, while keeping it at a reasonable cost, needs specific expertise
Every episode has to become text before anyone can search it, and the cost grows with every hour of audio. Transcribing hundreds of thousands of episodes requires specialised infrastructure. Keeping it affordable means knowing which model fits which step, and not paying for capacity you don't need.
-
Timestamps are a moving target
A hit has to open the episode at the sentence where the words are spoken. But since podcast hosts insert advertisements dynamically, the same episode changes length from one download to the next. The indexed timestamps have to adjust automatically to every version of the audio.
-
A completely new way of interacting with audio
Searching inside speech is a new concept that turned out to be difficult for users to comprehend at first. We needed an interface that new users could work out without instructions.
How we built it
We built Podric along the three steps of our methodology.
01 - Set the targets
Strategic discovery defines what to build, for whom and how success is measured. We limited the scope to Dutch-language podcasts. Media monitoring demands a complete index, and the Dutch-language market is small enough for a small team to cover in full. A global index would have cost too much to transcribe at this stage. Once the approach holds up in the Netherlands and Flanders, it can move to the next language area.
We set three targets before building: index the complete Dutch-language podcast market with every episode transcribed, sign two media-monitoring clients for our B2B product (Pulse), and reach 1,000 downloads of the Podric mobile app.
02 - Own models, existing feeds
Podric runs several large language models in our own environment, each chosen for the stage of the pipeline it handles well. Self-hosting makes Podric independent of vendors and makes AI costs predictable. Our discovery process for new episodes builds on a technology that has existed for decades: RSS feeds. Our engine subscribes to podcast feeds and new episodes are indexed in near real time.
The same AI engine powers Podric Pulse, our media-monitoring product. Pulse provides B2B clients with a portal to manage podcast queries at scale, and an API to integrate our data stream into their existing infrastructure. Saved searches and Pulse queries run continuously against every new episode, and send an alert or push the mention to the API when they match. For each mention Pulse also calculates a media value, based on listening data and advertising prices that our systems pull in from several sources.
03 - Clicking through sound
Adoption by design treats daily use as a requirement from the first week. For Podric that made UX research as large a part of the work as the AI. We had to design how a user searches inside audio and engages with the result, and we found no existing pattern to copy.
Results are grouped per episode. Each hit shows its timestamp and the sentence around it, with the search words marked, so a listener can judge a result before playing it. Playback starts directly at the precise moment of the hit. Every fragment can be shared as a quote, with the link opening at the exact timestamp. This allows actual deep-linking into podcast content.
What Podric delivers today
Podric has been live as a web search service since March 2024 and as an app since March 2025. The index more than doubled after the app launch, from 100,000 episodes to over 230,000 in September 2026. All three targets from the Discovery phase have been met.
-
Index the complete market
2,500+ shows and 230,000+ episodes, transcribed and monitored in real time.
-
1,000 app downloads
More than 1,200 in September 2026, on iOS and Android.
-
Land two B2B clients
Signed. Media-monitoring organisations run 160+ active monitoring projects on Pulse and receive podcast mentions through the API. Together those projects surface an estimated €4.5 million in media value a month.
related expertise
The expertise behind this case
-
Semantic search and retrieval-augmented generation (RAG) over your documents, archives, audio and structured data. Find what your organisation already knows in seconds.
More on knowledge retrieval and search -
Motion Delta builds AI agents and agentic workflows that carry out multi-step tasks inside your existing systems, with clear rules for what the agent decides alone and where a person steps in.
More on AI agents and agentic workflows
let's talk AI