Documentary Storytelling

A Source-Checking Workflow for History Documentary Pages: From Claim Log to Correction Policy

A practical fact-checking workflow for short history documentaries and Facebook pages: log every claim, rank your sources, verify images, and handle corrections in public.

By Rajib Roy Published 5 min read

On a history page, a wrong date is not a typo. It is the whole product failing. Your audience can be generous about a rough voiceover or a soft transition, but they will remember the post where you put a photo from one war next to a caption about another, and the comments will remind everyone for you.

I run Facebook pages about military history and the environment for readers in the US, UK and Canada, and the lesson of that work is that speed and accuracy fight each other unless you have a system. This post is the system I think a documentary page should hold itself to. It is deliberately low-tech: a spreadsheet, a few rules and a public corrections policy. You can start today with no software.

Rule 1: write the script, then log every claim in it

Read your finished script and pull out every statement that could be wrong: dates, places, names, ranks, numbers, quotes and cause-and-effect sentences ("this led to..."). Put each one on its own row in a claim log. It feels slow for the first three videos and then it becomes the fastest part of your process, because you stop re-researching the same facts.

Here is a template. Save it as claim-log.csv and open it in any spreadsheet:

claim-log.csv

claim_id,claim,type,source_1,source_2,confidence,status,note
1,"The treaty was signed on <date> in <place>",date+place,"Archive document, catalogue ref","Scholarly book, page","high","verified",""
2,"About <number> soldiers took part",number,"Official report","Secondary study","medium","needs 2nd source","Sources disagree; say 'estimates range from X to Y'"
3,"Photo shows <event>",image,"Archive catalogue entry","Reverse image search result","low","do not use","Caption in the archive says a different location"

The columns do real work. Type tells you how strict to be: a date needs a primary or scholarly source, an image needs provenance. Confidence forces you to admit when you're guessing. Status gives the final call: verified, needs a second source, or do not use. Nothing goes into the video with a status other than verified or an explicitly hedged line.

Rule 2: rank your sources before you open them

Not every source deserves equal trust. I sort them roughly like this:

  1. Primary sources. Official records, contemporary reports, letters, orders and photographs with an archive catalogue entry. They are close to the event, but they carry the bias of whoever wrote them.
  2. Scholarly secondary sources. Books and papers by historians who cite their own sources. Check how they know what they claim.
  3. Reputable reference works and museum pages. Good for orientation and for finding better sources.
  4. Everything else. Blogs, forums, other social pages and video channels, including ones that look polished. Treat these as leads, never as proof. A fact that only appears on other social pages has often been copied from one place.

The practical rule: no fact goes in on a single non-primary source. If two independent sources agree, publish it. If they disagree, write the disagreement into the script, for example "estimates range from X to Y", instead of picking the number you prefer.

Rule 3: numbers, dates and names get extra suspicion

These are the details audiences love and also the ones that get copied wrong. For each one, ask three questions. Is this figure for the same thing across my sources, say the same unit, the same date range, the same definition of "casualties" or "participants"? Is the date in the same calendar and time zone as the event I'm describing? And is the name spelled the way the primary record spells it, with transliterations flagged?

When the honest answer is "the record is unclear", say so on screen. A documentary that admits uncertainty reads as more credible, not less.

Rule 4: treat every image as a claim too

Images are the most commonly misused material on history pages. A photo can be real and still be captioned wrongly, often by years of copying. Before you use one:

  • Find the original catalogue entry. National and museum archives such as the Library of Congress, the U.S. National Archives, Library and Archives Canada and the Imperial War Museums describe what they hold, often with date, place and photographer. If you can't find the archive record, treat the caption as a rumour.
  • Run a reverse image search. Look at the oldest appearance of the picture and see whether other captions disagree with yours.
  • Check what's actually in the frame. Uniforms, equipment, signage and terrain either support your caption or they don't.
  • Check the license. Being online doesn't make a picture free to use. Wikimedia Commons, for instance, shows a license on each file page, and archives publish their own reuse terms. Read them for each item and keep the credit line.

Log the image in the same claim spreadsheet: what you claim it shows, the archive reference, and whether you could verify the caption.

Rule 5: separate what happened from what it meant

A great documentary mixes facts with interpretation, and viewers deserve to know which is which. "The battle began on that date" is a fact you can source. "That decision doomed the campaign" is an argument. Mark interpretation with language such as "historians argue" or "many scholars see", and when you can, name a source or two who hold each view. That small habit protects you from being accused of passing opinion off as record.

Rule 6: have a corrections policy and use it

You will get something wrong eventually. What matters is what happens next. Mine would be simple:

  • Pin a short "Corrections" note on the page explaining how readers can report an error.
  • When a reader reports one, check it against your claim log first. You'll often find the sources and the disagreement are already written down.
  • If you were wrong, say so plainly in a comment or an edit note, fix the video or caption, and thank the person who found it.
  • If you were right, reply with the source, politely, and move on.

Deleting the post and hoping nobody noticed is the one option that reliably makes things worse.

A lightweight weekly routine

  1. Pick one story and write the script.
  2. Log the claims and give each a type.
  3. Source each claim to two places, or mark it as hedged.
  4. Verify each image against an archive record and a license.
  5. Read the script aloud once more with the log open, and cut anything marked "do not use".
  6. Publish with credits and keep the log, because it's your defense and your head start on the next video.

Where tools can help, and where they can't

AI assistants can help you outline a script, suggest search terms or tidy a caption. They can also produce a confident, wrong date. Anything a model tells you goes into the claim log as an unverified lead, not as a source. The same goes for AI-generated or AI-upscaled "historical" images: if it didn't come from an archive record, don't present it as a photograph from the period.

Not legal advice. Copyright and reuse rules differ by country and by item. This post is a research workflow, not a legal opinion. When you're unsure whether you may use something, don't.

Follow along

Want to see the documentary side of what I do?

I share history and environment stories on my Facebook pages, and I build the tools behind them. You can find the links on my main site.

About Rajib Collaborate

What I ran this against

  • This is a process post, not a product review, so there is nothing to benchmark. The CSV template was loaded with Python's csv module to confirm it parses.
  • I have not invented any statistics. Where an example needs a number, it is written as a placeholder such as <number>.

Sources checked

  • Library of Congress, free-to-use and reuse material and catalogue: loc.gov, site reachable 17 November 2026.
  • U.S. National Archives: archives.gov, site reachable 17 November 2026.
  • Library and Archives Canada: library-archives.canada.ca, site reachable 17 November 2026.
  • Imperial War Museums (UK): iwm.org.uk. My automated check was blocked (HTTP 403), so I could not confirm its current page layout.
  • Wikimedia Commons: commons.wikimedia.org. Licenses differ file by file, so read each file's page.
  • I list these as places to look. I have not checked each institution's current reuse terms, so read them before you publish anything.