How to fact check a short

On 19 August 2026, a direct HTTP pull of four StoryShort pages found 11 words of body text on each one before JavaScript ran, meaning a viewer who loads that page without scripts reads almost nothing of what the finished video claims. This page gives you a five-step method to check the claims inside your own short before you publish it.

The short answer

Fact checking a short means marking every claim in the script first, then verifying each claim's number and date against a source that is not another video, then fixing the narration itself rather than patching the caption, and finally keeping the check list alongside the video file. A claim is verified when you can name where its number and its date both come from. If either one fails, the claim fails, even if the rest of the sentence is accurate.

The steps

1. Mark every claim in the script before you check any of them

Read your script once and underline every statement that asserts a fact: a date, a quantity, a name attached to an action, a cause attributed to an event. Write each one on a separate line in a plain list. Do not start verifying yet; the point of this pass is completeness, because the claims you skip are usually the ones you feel surest about, and those are the ones that survive longest unchallenged. A documentary script carries more distinct factual claims than you expect once you separate them properly. Counting them changes how you write the next script too, because you start noticing which sentences exist only to sound authoritative.

Output: a numbered list of every factual claim in the script, nothing checked yet.

2. Check the number and the date separately

A correct figure paired with a wrong date is still wrong, so treat the two as independent checks. If your script says a ship sank in a given year carrying a given number of people, verify the casualty figure against one source and the year against another, because encyclopedic summaries often merge numbers from different editions or different reporting waves. Write down, for each claim, the value you asserted, the value the source states, and the source's own measurement or publication date. When the two values differ, do not split the difference; take the source's version or cut the claim. This is also the step where rounding errors surface, since a figure repeated across three videos tends to lose precision each time.

Output: a two-column record per claim, asserted value versus sourced value, with dates.

3. Trace each claim to a source that is not another video

Another YouTube video is not a source; it is a second-hand retelling whose own provenance you cannot inspect. Follow the chain until you reach a document, archive record, official report, or dataset you can open yourself. If a video cites a book, find the book. If it cites a news report, find the report. If the chain ends at another video, treat the claim as unverified and either remove it or rewrite it to say explicitly that the figure is disputed or unconfirmed. This constraint is the reason ViewMade pulls footage from real archives and ships a media credit file naming each clip's origin with every render: the citation trail has to exist outside the video itself for anyone to check it.

Output: per claim, a named primary source you personally opened, or the claim marked unverifiable.

4. Fix the script, not the caption

When verification turns up an error, edit the narration text and re-record the affected lines. A corrected caption over an unchanged voiceover does not fix anything; viewers hear the error, and the subtitle contradicts the audio, which reads as sloppiness rather than diligence. Re-render after the script change so the captions, the voiceover, and any on-screen text all carry the same corrected wording. Keep the old version archived if you want a comparison, but publish only the corrected render. If the correction changes the video's central claim rather than a supporting detail, say so plainly at the top of your description instead of burying it.

Output: a re-rendered file where audio, caption, and on-screen text all state the corrected version.

5. Keep the check list with the video file

Store the claim list, the source names, and the asserted-versus-sourced table in the same folder as the final render, using a filename convention that ties them together. Questions arrive weeks later: a commenter challenges a figure, another creator asks where you got a clip, a platform review requests provenance. If the record lives only in your memory, the answer does not exist. If it lives in a saved file, answering takes minutes rather than a reconstruction effort. This habit also protects future-you during a re-edit, because you can see immediately which claims were already verified and reuse that work instead of redoing it.

Output: an archived folder containing the video, the claim list, and the source table together.

What this will not fix

Verification cannot exceed the reliability of the source you verify against. If the primary document itself contains an error, your careful process will faithfully reproduce that error and attach your credibility to it. A number passed through two intermediaries does not become more true on the second pass; each hop is a chance for transcription mistakes, unit confusion, or deliberate embellishment, and none of these are detectable from the receiving end. The method above reduces the risk by pushing you toward primary documents, but it cannot eliminate it. Where a figure matters enough to anchor the whole short, look for a second independent primary source that agrees. If no such agreement exists, present the figure as contested rather than settled, and let the audience hold both versions.

Where to go next

Once your claims are verified, /guides/how-to-cite-sources-in-a-video shows how to present those sources inside the video and its description so viewers can follow the same trail you did. If you are producing true crime shorts specifically, /guides/how-to-write-a-true-crime-short covers the scripting side, where claim density tends to be highest and the cost of a wrong detail is correspondingly larger.

<!-- faq -->

Frequently asked questions

How long should fact checking take for a single short?

Most of the time goes into tracing sources, not marking them. Scripts dense with historical figures and casualty counts run longer because each number needs its own date check. If a short takes far longer than expected, the usual cause is claims that should have been cut at step one.

Do I need to fact check opinion or analysis segments?

You need to check the facts underneath them. An analysis line like "this response was slower than comparable incidents" rests on response times, so those times get marked and traced like any other number. Pure commentary with no underlying figure needs no source trace, but mark it anyway during step one so you consciously decide rather than overlook it.

What if the only available source is another video?

Treat the claim as unverified. Either drop it, or keep it and label it in the script itself as reported but unconfirmed, citing the original uploader by name. Labeling preserves the information while being honest about its status. Never silently adopt a figure just because several videos repeat it; repetition across retellings is correlation among copies, not confirmation.

Should I show my sources in the video or only in the description?

Both serve different readers. On-screen citations build trust while watching; a description list lets someone actually open the sources afterward. At minimum, put the full list in the description with dates. ViewMade writes a media credit file naming each clip's source into every render package, which handles the footage side automatically, but spoken claims still need your own list.

Does fact checking matter for channels that post daily?

It matters more, not less, because volume multiplies the audience for each error. A channel publishing daily makes a large number of verifiable assertions per month, and a single recurring mistake compounds across uploads. The checklist-and-archive habit from step five is what keeps daily output sustainable, since verified claims become reusable assets for later scripts.