Two journalists leaf through vintage magazines and reflect on their legacy in Mag Hags. Plus: five of the best US election podcasts

• Don’t get Hear Here delivered to your inbox? Sign up here

Podcast creators beware. This week Google released NotebookLM, an experimental tool to help with note-taking, which can create an audio “deep dive conversation” about any document you care to upload. Read: an AI-generated podcast.

The results, frankly, are astounding. There are two hosts whose US-accented voices are almost impeccably believable as humans, and whose banter, digressions and personal interjections – as well as a tendency to pull in separate pieces of research to help explain your document – are almost indistinguishable from the work of humans. Assuming, that is, we’re talking about quite competent humans.

Feeding in a recent Hear Here intro about the longevity of podcasts compared with that of TV had NotebookLM riffing about Off Menu due to its mention in the article (“I’ve literally listened to episodes of Off Menu in public and laughed out loud – like, full-on belly laughed!”). It also introduced some astute observations not included in the piece itself (“It’s interesting, isn’t it? I think that maybe traditional listener fatigue doesn’t apply in the same way, because podcast audiences are largely self-selecting – like, we choose what we listen to. We’re invested from the start”). I was less chuffed with the podcast it churned out in response to taking a look at a report about the damp in my home (“That sounds like something out of a horror movie!”).

It may not be perfect, but it’s scary how impressive the results are. How long until a fully AI-generated show takes off? And how easy would it be to send somebody one of these audio files and convince them it was a legitimate human-created podcast?