What this guide helps you do
Create accurate and responsible history documentaries with AI
Key takeaways
- Build the episode around a historical question, not a spectacle prompt.
- Separate evidence, interpretation, and visual reconstruction.
- Review period detail and disclose realistic synthetic scenes.
Film evidence
Build history around evidence, then choose reconstruction
A responsible Pompeii episode begins with a bounded question and source package, not a request for dramatic Roman imagery. The writer should identify what is documented about the city and eruption, what historians infer, and which experiential details the film will reconstruct for orientation or emotion.
Generated scenes can make scale and sequence legible, but realistic confidence must not outrun provenance. Connect claims to sources, avoid fabricated quotations, keep unsupported appearance details off-screen or clearly illustrative, and place reconstruction disclosure where viewers will encounter the images. A final expert or editorial review remains necessary for sensitive historical subjects.
Pompeii: Vesuvius Erupts · 01:01 · Historical reconstruction. This is a finished first-party Onira production, not customer proof or archive footage.
View the full film and production notesWhat to watch for
- Documented event sequence separated from illustrative staging
- No generated frame presented as archival evidence
- Sources and disclosure remain available with the published film
Section 1
Choose a question the evidence can answer
The strongest history documentaries are specific enough to investigate. Instead of 'the Roman Empire,' ask how water, grain, or road logistics shaped daily life in one city and period. The narrower question produces a clearer source set and more concrete visual direction.
Define the geographic and chronological boundary, key people or groups, and the consequence the episode will explain. Record where historians disagree so the script does not turn interpretation into settled fact.
- +One historical question and one defensible thesis.
- +A bounded time, place, and cast of actors.
- +Known uncertainty and competing interpretation.
Section 2
Build a compact source pack
Use primary sources when they are relevant and interpretable, then pair them with reliable secondary scholarship that supplies context. Record title, author or institution, publication details, URL, access date, and the claims each source supports.
A source list is not the same as claim-level verification. During review, trace names, dates, quotations, causal statements, and disputed interpretations back to supporting material. Remove a claim when the evidence is too weak for the certainty of the narration.
- +Primary evidence for direct historical material.
- +Secondary scholarship for interpretation and context.
- +Claim-to-source notes for the final script.
Section 3
Direct reconstruction as reconstruction
Period-aware prompts should describe architecture, materials, clothing, tools, weather, light, social context, and the physical scale of a scene. Avoid asking a model for vague 'historical cinematic' imagery, which invites visual shorthand and anachronism.
Generated scenes are illustrations. They cannot establish that an undocumented event looked exactly as shown. Do not imitate archive markings or camera artifacts in ways that imply authentic footage, and do not fabricate readable documents, quotations, maps, or inscriptions as evidence.
- +Research material culture before visual generation.
- +Review clothing, architecture, tools, geography, and season.
- +Label realistic scenes as reconstructions when needed.
Section 4
Review with two separate lenses
The factual review asks whether the narration, captions, dates, names, quotations, and implied causality are supportable. The visual review asks whether scenes introduce anachronisms, impossible geography, invented text, misleading likenesses, or apparent evidence that the sources do not establish.
Finally, make a disclosure decision under YouTube's current altered-content guidance and preserve source notes with the publication. Trust is not created by claiming that AI is accurate. It is created by making the creator's review visible and specific.
- +Factual script review.
- +Historical visual review.
- +Disclosure, source notes, and correction process.
Working standard
Publication checklist
- 01Question, period, place, and thesis are bounded.
- 02Primary and secondary sources have complete references.
- 03Major claims map to supporting evidence.
- 04Visual prompts include researched period detail.
- 05Generated scenes do not impersonate authentic evidence.
- 06Factual, visual, and disclosure reviews are complete.
Primary references
Sources and further reading
Policy and model capabilities change. These sources were reviewed on July 11, 2026; open the current official page before making a production or publication decision.
Related production guides
Questions
Can AI history documentaries be accurate?
AI can assist research, writing, and reconstruction, but accuracy depends on the source boundary and human review. Generated prose and imagery should never be treated as evidence by themselves.
Should AI historical reconstructions be disclosed?
Review YouTube's current guidance for realistic altered or synthetic content. If a generated scene could lead viewers to believe it is authentic footage of a real event, disclosure is particularly important.
What history topics fit generated visuals best?
Daily life, architecture, crafts, landscapes, material history, travel, and atmospheric reconstructions often fit better than topics requiring exact maps, readable documents, or proof of a specific undocumented action.