How claims get into the library

How a claim is sourced

The Cited Conversations podcast records structured expert conversations on defined topics. After each recording, the episode is transcribed and reviewed. Individual exchanges are extracted as standalone answer pages — one question, one answer, one speaker, one timestamp link.

Each answer page carries:

• The speaker’s name, title, and company
• The show name and episode number
• The exact timestamp where the answer begins
• A link to that moment in the source recording, playable directly

How attribution works

Every claim is attributed to a named speaker. Speakers are confirmed at intake by the show’s producer — we record who they are, their role, and the organization they represent before the conversation begins. Attribution is not inferred from the transcript; it is established before recording starts.

Speaker entity data (name, title, organization, LinkedIn profile) is embedded in the page’s structured markup so AI answer engines can resolve the attribution without parsing prose.

How verification works

Every claim links to a playable moment in the source recording. If the link doesn’t work, the claim doesn’t ship.

Before a page publishes, a human reviewer checks that:

• The written answer accurately represents what the speaker said
• The attributed speaker is the person who said it
• The timestamp link opens at the correct moment in the episode

We spot-audit published pages against source audio on an ongoing basis.

Transcript fidelity and filler removal

Ums, uhs, and Descript-style false-start fragments (a word ending in a hyphen, e.g. “th-”, “ar-”) may be dropped from quoted answer text without an ellipsis; every other omission — including discourse markers such as “right,” “well,” or “exactly” — requires a visible ellipsis (…) at the point of the cut. Added words go in [square brackets]. No reordering. No paraphrasing inside quotation marks. The source transcripts (transcript.vtt, transcript.txt) are published exactly as exported from Descript, with track-ID labels replaced by real speaker names.

What does not get in

The library is deliberately narrow. We exclude:

• AI-generated conversations or AI-written answers
• Solo commentary segments without a named guest
• Re-aired or recycled episodes that weren’t freshly recorded
• Any claim we cannot link to a playable moment in a source recording

The constraint is intentional. An unlinked claim is an unverifiable claim. If we cannot play it, it doesn’t belong here.

Title policy

No conversation page receives a “best X” title when X is something Ringmaster or a partner sells. Titles describe what the conversation covers — the framework, the method, the trade-off — not a ranking outcome. This rule applies retroactively: pages were retitled in October 2026 as part of Batch 4 of editorial improvements.

Annotation policy

We annotate claims we cannot verify instead of removing them. Where a speaker makes a specific factual assertion — a statistic, a technical claim about how an AI system works, or a product name — our editors check it. If we cannot verify it, an editor’s note appears under that answer. The note states what we checked and what we could not confirm; it does not editorialize about the speaker’s intent.

Conflicts and independence policy

How guests are chosen. Guests are selected for their expertise on the episode’s topic, not for their commercial relationship with Ringmaster. Ringmaster’s producer contacts potential guests directly and independently; no guest has paid to appear on Cited Conversations.

Commercial ties disclosed. Where a guest’s company has a commercial relationship with Ringmaster, that relationship is stated in the editorial disclosure that appears at the top of every page for that episode.

• SCAILE (Episode 002, guest Simon Wilhelm): SCAILE is a commercial partner of Ringmaster.
• ANNUITAS (Episode 001, guest Casey Cheshire / interviewer Adam Needles): ANNUITAS has no commercial relationship with Ringmaster — not a client, partner, referral, or paid relationship.

This list is updated as new episodes publish. If a conflict changes after publication, the disclosure on the relevant episode page is updated and a changelog entry is added.

The underlying principle

AI answer engines are increasingly the first place people get answers to professional questions. Those engines pull from whatever text they can find. Most of that text is uncited, undated, and unspeakered.

Cited Conversations is built on the opposite assumption: that the most useful thing we can do is make source attribution machine-readable. Every answer here is a piece of structured evidence, not a piece of content.