How These Documentaries Are Made

We’d rather state this plainly on a page you can find than bury it in a line of small print. A documentary asks you to trust it. That trust should be inspectable.

Every episode begins with reading, not writing.

The starting point is published academic history: monographs, university press surveys, and translated primary sources where they exist and are accessible. Popular histories are used for narrative shape and structure, never as the sole authority for a factual claim.

The full bibliography for each episode is published on this site. Not a representative sample, not “sources include” — the actual list, with authors, titles, editions, and years, so you can go and check us.

Where historians genuinely disagree, the script says so rather than picking the tidier version. Ottoman origins, the population figures for premodern cities, the causes of institutional decline — these are live arguments, and flattening them into a confident sentence would make for smoother narration and worse history.

Where a figure is contested or an estimate, it is described as one.

Scripts are researched, structured, and edited by a person. AI assists with drafting and restructuring prose, in the way a great many writers now use it.

What is not delegated is the factual layer. Every claim that reaches narration is checked against the bibliography by a human. A language model is a competent stylist and an unreliable historian; it will produce a fluent, plausible, and entirely invented date without hesitation. That failure mode is the reason the fact-checking pass exists and is done by hand.

The narrator’s voice is synthetic, generated using text-to-speech. There is no human narrator and we don’t imply one.

The words are written by us. The voice reading them is not a person, and no real individual’s voice has been cloned or imitated.

Historical documentary runs into a hard constraint: there is no footage of 1453, and there are no photographs of Osman. Every channel in this genre solves that somehow — reenactment, museum photography, animated maps, stock libraries. We use a combination of generative imagery, licensed stock, and public-domain material.

The rule we hold to is this:generative images are illustration, not evidence.

They depict scenes for which no photograph could ever have existed — a camp at dusk, a street, a battle line. They are the visual equivalent of a book’s cover art. They are never presented as artifacts, documents, or photographs.

When you see a real manuscript, painting, coin, or building, it is the real thing, and the episode or this site identifies it.

Maps and timelines are built from historical atlases and scholarly sources, and those sources appear in the episode bibliography alongside the written ones.

• We don’t present generated images as historical artifacts, documents, or photographs.
• We don’t fabricate quotations. If a historical figure is quoted, the quotation is real and its source is in the bibliography.
• We don’t narrate claims that aren’t traceable to a published source.
• We don’t build episodes by summarising other creators’ videos.
• We don’t clone or imitate any real person’s voice.

We will get things wrong. Every history does, and the ones that pretend otherwise are the ones to distrust.

When an error is found — by us or by you — the correction is made and logged publicly at the bottom of the relevant episode page, with the date and what changed. Corrections are not quietly edited in.

If you’ve found something wrong, tell us. Specific and sourced is most useful, but “this doesn’t sound right” is a perfectly good place to start.

[email protected]

Because the alternative is asking you to take a documentary on faith, and because the disclosure costs us nothing that’s worth keeping.

There’s a version of AI-assisted history that is genuinely bad: unsourced, unchecked, generated end to end, indifferent to whether any of it happened. That work exists in volume, and it makes readers reasonably suspicious of anything in the neighbourhood. The only useful answer is to show the work.

The bibliography is the real disclosure. This page just explains how it gets used.