How the family tree is built

1. Every episode, back to the start

We collect each show's full episode history from its RSS feed. Feeds often only keep the last few hundred episodes, so we fill the gaps from the publishers' own archives (the Guardian's goes back to 2006), from Podcast Index, and from old copies of feeds saved by the Wayback Machine. That's 59,265 episodes across 124 podcasts so far.

2. Who was on each episode

Show notes name the people taking part, but not in any standard way — “Jimbo is joined by…”, “Gary, Alan & Micah discuss…”. An AI model reads each episode's title and notes and lists who actually took part (host, co-host, guest or interviewee), leaving out players and managers who are only being talked about. Nicknames and first names are matched to full names using what we know about each show.

3. Parents and children

A show's founders are the people fronting it in its first year. If a founder had already hosted, or been a regular on (three or more appearances), an earlier show, that earlier show counts as a parent. A one-off guest spot counts as a weaker link — shown dashed in the tree, but not used for lineage. Following parents back gives a show's ancestry.

We can name founders for 115 shows; 78 of those have at least one strong parent. 61 descend from the Guardian's 2006 World Cup Show and 53 from The Times' The Game, which launched a month later — 48 of them from both.

What it can't tell you

• Some feeds start years after the show did, so their earliest regulars may be missing.
• We only see people the show notes name. A regular who's rarely credited will look like an occasional guest.
• Shared people show influence, not that one show deliberately spun out of another.
• Guest data is still being processed for later episodes of many shows, so counts will keep growing.