A search layer for spoken media
Particle introduced Radar on Aug. 26 as a search engine for podcasts, with indexed transcripts, speaker labels, alerts and sponsorship information. The company says the service covers more than 130,000 podcasts and processes about 20,000 new episodes each day, according to its product announcement.
Users can search the website, connect through an API or give AI agents access through the Model Context Protocol. That turns long-form audio into a source that can be queried alongside documents and databases rather than consumed only from beginning to end.
The basic value proposition is time. A researcher can locate every recent mention of a company, executive or theme without waiting through dozens of episodes. A communications team can monitor how a topic travels across shows and who is shaping the conversation.
The data has commercial value
Radar also extracts advertising and sponsorship signals. That creates a market for questions that podcast apps were not designed to answer: which brands are buying in a category, which hosts discuss a competitor and how a narrative changes across episodes.
TechCrunch reported that hedge funds are among the highest-volume API customers, citing Particle’s founder. The company prices individual access at $29 a month and a business tier at $399 a month for up to 20 seats.
Those customers illustrate why podcast search is more than a discovery feature. Spoken comments can contain early demand signals, management views and specialized expertise. Making them machine-readable increases their speed and economic value.
Searchability changes the medium
Podcast guests have long spoken in a format that felt conversational and ephemeral. Comprehensive indexing makes those comments easier to find, compare and reuse. Hosts and guests may respond by becoming more deliberate about claims, disclosures and offhand remarks.
AI agents add another layer. A system can monitor new episodes, extract relevant passages and feed them into workflows without a human initiating each search. That can improve research, but it also creates risks around transcription errors, context loss and overconfidence in automated summaries.
Radar plans to extend beyond podcasts to other audio sources. If that expansion succeeds, spoken media will increasingly function like a continuously updated database. The challenge is preserving the nuance of a conversation after software has turned it into rows, alerts and agent inputs.
Media companies should prepare for a world in which every spoken appearance is searchable soon after release. That means correcting factual errors quickly, preserving links to original episodes and training teams to verify transcript snippets before acting. Search makes podcasts more useful, but responsible use still requires listening around the result rather than treating a sentence as the whole conversation.