How rankings work

Every number on PodCited comes from something a person said on a podcast. Here is how a sentence in an episode becomes a ranking, and what we leave out.

1. Collecting the conversations

We transcribe full episodes of 22 long-form podcasts, listed at the bottom of this page. Short clips, trailers and paid-tier previews are skipped. An AI model reads each transcript and records every book that comes up: the exact words, who said them, and where in the episode.

2. Sorting what was meant

Each mention gets one of four labels. Only two of them count toward a book's standing.

👍 Endorsement	A speaker recommends a book someone else wrote.	Counts
⚡ Critique	A speaker argues against a book or names its limits.	Counts
🎙️ Author plug	The author is on the show, or a host is talking with the author about it.	Shown, never counted
Passing reference	The book is named without an opinion attached.	Shown, never counted

Some checks are applied by rule rather than left to the model:

If the author takes part in the episode, every mention of their book is an author plug, whatever the tone.
On shows with a single host, an episode devoted to a book is that host's recommendation, not a plug.
Advertisements read on air are dropped, even when they name a book.
A line a show repeats in many episodes (an intro, a sponsor bridge) is treated as show copy, not as fresh praise.
A few books are discussed as historical sources rather than recommended (Mein Kampf, for example). Those are never shown as recommended.

3. Ranking

The All-Time Consensus list orders books by three things, in this order:

Different podcasts. How many separate shows endorsed or critiqued the book. Three shows recommending a book once beat one host recommending it four times.
Conversations. How many episodes endorsed or critiqued it. Several mentions in one episode count once.
Mentions. All published mentions, as a final tie-breaker.

A book gets its own page in search results once it has been endorsed or critiqued in at least two different conversations. Everything else is still on the site, just not offered to search engines yet.

4. Checking the moment

Timestamps are never taken from the model. Each quote is found again in the transcript, and playback starts a few seconds early so you hear the lead-in.
A quote plays as video only when we have found it in that video's own captions; otherwise it plays the podcast audio.
Some podcast hosts insert different ads for each listener, which shifts the episode. For those shows the button says “Listen around”, because the moment may be a few minutes later.

5. The written summaries

The short lines on each book page (what the book is, why it was recommended, who it suits) are written by an AI model from the quotes on that page and checked against fixed rules: no marketing language, no claims nobody made. The quotes themselves are the record; the summaries are a guide to them.

Podcasts we follow

Spotted a mistake, or a podcast we should add? Get in touch.