Editor’s note: This initiative won the INMA Global Media Awards “Best in Show” award last week in Berlin.

Nobody has time to listen to all of it.

Candidates sit for hours with Joe Rogan, Marc Maron, or Theo Von. Wellness influencers tout antidepressants between dance videos. Conservative hosts do battle with the administration they helped elect. Clips circulate on YouTube, TikTok, X, and Instagram until a single conversation has fanned out into hundreds of segments over days or weeks.

The podcast ecosystem is sprawling, and episodes are long. Politics, wellness, comedy, sports — all of it weaves together. A host might float an idea for weeks before it becomes part of a campaign’s message. A fringe argument can move from a wellness show to a political podcast to a candidate’s stump speech without most newsrooms noticing.

Searching the pods

We built Orca to keep up. This internal newsroom tool uses AI to track podcast feeds, download new episodes, transcribe them, and make them searchable. It summarises, identifies speakers, extracts topics, and alerts reporters to conversations that might matter to them.

A reporter can search across episodes, ask questions about a topic, or set up briefings around the things they care about. Instead of spending the morning scrubbing through audio, they can start with the few segments most likely to be relevant.

The point isn’t to replace judgment. It’s to give reporters a chance to filter an information ecosystem that has become far too large to monitor by hand.

Orca in action

We used Orca to write the definitive account of how Epstein consumed the MAGA movement last summer. The tool found 3,123 episodes across 124 shows mentioned Epstein, with chatter jumping more than eightfold in the three weeks after the DOJ said there was no client list — even after Trump told the movement to drop it.

Reporters tracked the moment Joe Rogan drew “a line in the sand,” when Alex Jones turned on the administration, and how the MAGA coalition frayed. Orca later helped reporters pull key throughlines out of the broader Epstein coverage — including podcasters airing unfounded claims about cannibalism and child sacrifice straight into upbeat ad reads.

Orca anchored reporting on how antidepressants went from stigmatised medication to “#lexaprogirly” lifestyle accessory, finding the podcast appearances where influencers — some paid by telehealth companies to promote SSRIs on social media — admitted to emotional numbness, brain fog, and sexual dysfunction they hadn’t disclosed to followers.

And it surfaced U.S. FBI Director Kash Patel’s podcast appearances — including telling Joe Rogan the job was “effin’ wild” and his 2023 criticism of his predecessor for taking the FBI jet on an Adirondacks vacation.

Mapping the narrative

Orca keeps reporters close to the underlying material. The transcript links back to the audio. If a summary points to something interesting, the reporter can check the quote, listen for context and decide what it means.

This is what the good version of AI in a newsroom looks like: clear objective. Large corpus. Unstructured information. Humans verify. A machine can tell us that a theme is appearing across dozens of shows. It cannot decide whether that theme is meaningful, exaggerated, misleading, or worth putting in front of readers.

Podcasting is looser than television, more intimate than social media, and harder to track than either. Orca helps us listen at the scale the moment requires.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .