Joe Rogan is, if nothing else, prolific. He’s published 2,286 episodes in the 15-ish years he’s been making his podcast, at an average of a little over two and a half hours each. New episodes drop three or four times a week, which means getting the full Joe Rogan Experience would take about 240 days of nonstop listening. It’s an impossible task, unachievable by even the most extremely online among us; by the time someone listened through the entire backlog, Rogan would have released another 120 or so episodes.

After the 2024 election, Kaveh Waddell and Patrick Swanson became a little obsessed with Joe Rogan. The two former journalists had met at Stanford, where they were John S. Knight fellows and started an AI consulting lab called Verso, and Swanson became particularly interested in Rogan’s three-hour interview with Donald Trump, which he said he watched six or seven times.

Rogan’s influence on the election was the subject of much discussion from media and political types alike; his ability to reach low-propensity and undecided male voters is unparalleled, and Trump’s appearance on the podcast gave him both an easy, low-conflict platform — Rogan tends not to fact-check his guests — and access to that audience. The Rogan impact is so great that some liberals even started talking about the need to have their own version of him.

“It seemed like such a cultural moment to me that I really wanted to understand what’s going on,” Swanson, the former head of social media for the newsroom at ORF, Austria’s largest broadcaster, told me. Swanson and Waddell started talking about making an AI tool that could do the listening for them, and a few weeks later Roganbot was born.

Roganbot works on a simple premise: it listens to Rogan for you, creates a searchable transcript linked to timestamped audio, and breaks episodes down into key topics, notable quotes, potential controversies, and suggested fact-checks. It can, in essence, be extremely online for you, and then you can ask it questions about what it found in its internet rabbit hole.

Roganbot is a testbed for a larger concept that Swanson and Waddell call “visibility tools” — AI-powered research tools that can help journalists keep tabs on specific topics without having to invest hours sifting through transcripts or data sets. The use cases for visibility tools depend on context: While Roganbot could be helpful for someone trying to keep up with the manosphere, a local journalist might find more utility in having a similar bot listen to city council meetings instead.

“These visibility tools give people eyes and ears where they don’t have them, and multiply the capacity of a newsroom,” said Waddell, who was formerly an investigative journalist at Consumer Reports and a staff writer at Axios and The Atlantic. Each tool can be personalized for the user and asked to look out for different things, Waddell said, “so you can externalize a part of your brain into the system.”

One of the hardest things about building a tool like Roganbot, Waddell told me, is making sure it doesn’t adopt the point of view of the source material; in other words, you don’t want a tool that analyzes Joe Rogan to sound like Joe Rogan. For now, each instance of Roganbot is siloed, which means the AI only knows what happened in a single episode of the podcast at a time.

The dream, Waddell and Swanson said, is to build a tool that can track not just the entirety of Rogan’s oeuvre but the larger right-wing media ecosystem as well, from influencers and streamers to more traditional outlets like Fox News. After all, Rogan, unlike the bot, doesn’t exist in a vacuum, and the ideas that end up on his show often have origins in other places.

“And then you could trace things like where does an idea start? Where does it go next? Who is spreading this? How does it end up on Rogan?” Swanson said. “This would allow us to have a much better insight into this gigantic, important, far-reaching space of media.”

This concept — tracking an idea — is particularly tantalizing. The internet is filled with subcultures within subcultures, each of which develop their own lexicon, and tracing the lineage of an idea (or, often, a conspiracy theory) can feel impossible: by the time the seed of an idea breaks containment and makes its way onto the national stage, it will have taken deep root in some corner of the internet. The promise of visibility tools like Roganbot is in finding that corner much more easily than ever before, allowing journalists to expose the roots when, say, a presidential candidate accuses immigrants of eating animals.

Verso doesn’t offer Roganbot as a public product yet. Swanson and Waddell have been focusing on their primary mission of helping news organizations with AI strategy and building custom tools for them, but say they’ve been approached by engineers, academics, and misinformation researchers alike who have asked to get involved with the project somehow.

Building Roganbot has also had some unexpected side effects: Both Waddell and Swanson get alerts from the bot whenever it analyzes a new episode of the Joe Rogan Experience, which happens each time an episode drops, and Swanson has found himself listening to more of the show than ever before. Waddell, meanwhile, has started seeing the show in a new light.

“I have been both surprised by the degree of conspiracy that will pop up —  like the amount of time that can be devoted to the moon landing or the Zapruder film or whatever — but that’s not the whole story, either,” Wadell told me. “There’s just a lot of hanging out and talking and a kind of unclear political valence. I certainly had a bit of a caricature in my head of what this podcast was and what it represented in the political sphere, and that’s been disrupted somewhat.”