Create a claim ledger that lists every factual assertion in your script, verify each against a primary source, and label any remaining uncertainty. Document every revision so that your audience and platform can see what was changed and why.

Key takeaways

  • Document every factual claim in a claim ledger before you start editing.
  • Verify each claim against a primary source, not a summary or AI-generated paraphrase.
  • Label uncertainty explicitly in the video and description when a claim cannot be fully verified.
  • Keep a revision log that tracks what changed, who changed it, and the reason for each change.

What Is a Claim Ledger and Why You Need One

A claim ledger is a simple table or spreadsheet where you list every factual assertion in your script, the source you used to verify it, and the level of confidence you have in that source. You can build it in a shared document, a project management tool, or even a plain text file. The key is to create it before you begin editing, so you are not trying to reconstruct your verification process after the fact.

Your claim ledger should contain at least four columns: the exact claim as it appears in the script, the URL or identifier of the primary source you used, a confidence rating (verified, partially verified, or unverified), and the date of verification. If you are working with a team, add a column for the person who performed the check. This ledger becomes the single source of truth for every factual decision you make during editing.

When you use AI tools to help generate or rewrite portions of your script, you must treat the AI output as a draft, not a verified fact. The claim ledger forces you to separate the creative process from the verification process. It also makes it easy to spot claims that have no source at all, so you can either find one or remove the claim before publishing.

  • Create the ledger before you start editing, not after.
  • Include columns for claim text, source URL, confidence rating, and checker name.
  • Treat all AI-generated statements as unverified until you check them against a primary source.

For a structured approach to managing generative-AI risks in your workflow, refer to the NIST Generative AI Risk Management Profile , which provides a cross-sector profile for identifying and managing generative-AI risks.

Verify Every Claim Against a Primary Source

A primary source is the original document, data set, or official record that contains the information you are citing. For example, if you claim that a certain law was passed on a specific date, your primary source is the official legislative record, not a news article summarizing it. If you claim that a scientific study found a specific result, your primary source is the study itself, not a press release or a blog post.

When you verify a claim, open the primary source and read the relevant passage. Do not rely on a search engine snippet, a Wikipedia summary, or an AI-generated paraphrase. If the source is a PDF, use the find function to locate the exact words you are citing. If the source is a video or audio recording, note the timestamp where the relevant statement appears. Record this detail in your claim ledger so that anyone on your team can re-verify the claim without starting from scratch.

If you cannot find a primary source for a claim, the claim should be labeled as unverified in your ledger. You have three options: remove the claim, find a credible secondary source and label it as partially verified, or keep the claim but clearly state in the video that the information could not be independently confirmed. Never present an unverified claim as fact.

When you publish on YouTube, you must follow the platform's disclosure rules; the YouTube: Disclosing Altered or Synthetic Content policy requires creators to disclose realistic altered or synthetic content when viewers could mistake it for reality.

Label Uncertainty Explicitly in the Video and Description

Viewers deserve to know when a claim is uncertain, contested, or based on limited evidence. Do not bury this information in a footnote at the end of the description. Instead, state the uncertainty in the video itself, either in the narration, a text overlay, or a clearly visible disclaimer. For example, you can say: The following statistic comes from a single study that has not been replicated, or This claim is based on a leaked document that we have not been able to authenticate.

In the video description, include a section called Sources and Uncertainty that lists every claim that could not be fully verified. For each one, explain why it is uncertain and what you did to try to verify it. This transparency builds trust with your audience and protects you if the claim later turns out to be incorrect.

If your video contains synthetic or altered content that could be mistaken for real footage, you must also disclose that fact in a way that is clear to viewers. The goal is to ensure that no one watches your video and leaves with a false impression about what is real and what is generated.

  1. Identify every claim in your ledger that is not marked as fully verified.
  2. For each uncertain claim, write a brief explanation of why it is uncertain.
  3. Add that explanation to the video narration or a text overlay at the moment the claim is made.
  4. List all uncertain claims in a dedicated section of the video description.

Document Revisions with Content Credentials

A revision log is a record of every change made to the video after the initial fact-checking pass. It should include the date of the change, the person who made it, the specific change made, and the reason for the change. If you remove a claim because it could not be verified, note that in the log. If you add a new claim after the initial ledger was created, you must fact-check it before publishing and add it to the ledger.

For videos that are likely to be updated over time, consider using Content Credentials to attach a cryptographically verifiable record of the video's provenance. This allows viewers and platforms to see whether the video has been altered since it was first published. The C2PA specification provides a standardized way to encode this information into the video file itself.

Even if you do not use Content Credentials, keep a plain-text revision log in your project folder. This log is useful if you ever need to explain to a platform, a fact-checker, or your audience why a particular version of the video differs from an earlier one.

To implement tamper-evident provenance in your video files, review the C2PA Content Credentials Specification , which describes how Content Credentials can store cryptographically verifiable provenance information and enable tamper evidence.

Final Checklist Before You Hit Publish

Before you publish, run through this checklist to catch any gaps in your fact-checking process. It takes only a few minutes and can save you from a major correction or takedown later.

Review your claim ledger and confirm that every claim is either verified or explicitly labeled as uncertain. If you find a claim with no source, remove it or find a source before publishing.

Check your video for any synthetic or altered content that could be mistaken for reality. If it exists, add a disclosure in the video and in the description. Then review the platform's disclosure requirements to make sure you comply.

  • Every claim in the final script has a corresponding entry in the claim ledger.
  • Every claim is either verified against a primary source, labeled as uncertain, or removed.
  • Synthetic or altered content that could be mistaken for reality is disclosed in the video and description.
  • The revision log is complete and includes the reason for every change.
  • The video description includes a Sources and Uncertainty section if any claims are uncertain.

Frequently asked questions

How do I handle a claim that I can only verify from a secondary source?

Mark the claim as partially verified in your claim ledger. In the video, state the secondary source and explain that you could not confirm the claim from the original primary source. For example, say: According to a Reuters report published on June 1, 2025, which cites an unnamed official.

Do I need to re-verify a claim if I use a different wording in the final edit?

Yes. If the meaning of the claim changes because of rewording, you must re-verify it. If only the phrasing changes but the factual assertion stays the same, you can keep the original verification in your ledger, but note the wording change in the revision log.

What if I discover an error after publishing?

Immediately update the video and description to correct the error. In the description, add a correction notice that explains what was wrong and what was changed. Update your claim ledger and revision log to reflect the correction. If the error is significant, consider re-uploading the corrected version.

Sources and further reading

Primary documentation used by the Banned.Video editorial team.

  1. NIST Generative AI Risk Management Profile
  2. YouTube: Disclosing Altered or Synthetic Content
  3. C2PA Content Credentials Specification