A Practical Checklist for Reviewing Suspicious Videos
Use a 9-item AI video verification checklist to check captions, visible details, context, edits, source claims, model disagreement, and human-review needs.
Who this is for
Journalists, fact-checkers, content creators, communications teams — Anyone who needs a fast, scannable checklist to run before sharing, publishing, or reacting to a video — without reading a full explainer first
The problem
Before you publish or share a suspicious video, check what the clip actually shows and what the caption claims it shows. Those are two different checks, and most people run neither one before hitting share.
A full video-verification workflow has a lot of steps worth understanding in depth. Sometimes what you need first is not the explanation — it's the checklist itself, in a form you can run through in under two minutes before you decide whether to share, hold, or escalate.
How ConvergePanel helps
This is the condensed version: nine checks, in order, covering what the caption claims, what the footage actually shows, and whether the vision models reviewing it agree. Each item links to the fuller explanation if you need it — but the checklist itself is the point.
How they compare
| Checklist Item | What You're Checking | If the Answer Is No or Unclear |
|---|---|---|
| Caption claim | What the video is being said to prove, stated as one sentence | You can't check a claim you haven't pinned down precisely — write it out before continuing |
| Actual content | What the footage shows, independent of the caption | If content and caption already look mismatched, treat this as a caption problem before an authenticity one |
| Date | Whether a timestamp or dateable detail exists | Treat the date as unconfirmed and search for the earliest known posting |
| Location | Whether signage, landmarks, or geography confirm the claimed place | Treat the location as unconfirmed; don't repeat it as fact |
| Original source | Whether you can trace this to its earliest known appearance | Reverse video search before trusting the version you have |
| Visible inconsistencies | Anything that looks visually unusual on a plain watch-through | Note it specifically and check it against the AI review's findings |
| Editing signs | Cuts, jumps, or missing frames that could change meaning | Assume the clip may be incomplete and look for the fuller version |
| Vision model agreement | Whether multiple models reviewing the clip reach the same assessment | Treat a split as a signal to slow down, not as noise to average away |
| Human verification need | Whether the stakes justify escalation beyond an AI pass | When in doubt on a high-stakes clip, escalate rather than publish on the checklist alone |
How it works
- 1What does the caption claim? Write down the specific assertion before you look at anything else
- 2What does the video actually show? Watch it once for content alone, separate from the claim
- 3Is the date known? Look for a verifiable timestamp or dateable visual detail
- 4Is the location known? Check for signage, landmarks, or other place-identifying detail
- 5Is there original source context? Find the earliest known post, not the version in front of you
- 6Are there visible inconsistencies? Note anything that looks visually off before running an AI review
- 7Are there signs of editing? Check for cuts, jumps, or missing frames that could change meaning
- 8Do vision models disagree? Run the clip through multiple models and check where they split
- 9Does the clip require human verification? Decide based on the above whether this needs more than an AI pass
Use cases
- Running a fast pre-share check on a video landing in a group chat or social feed
- Using a consistent checklist across a newsroom or content team before publication
- Training new fact-checkers or moderators on the minimum checks for suspicious video
- Documenting that a structured checklist was followed before a publish-or-hold decision
Why a checklist and a full workflow are both useful
The checklist above is deliberately compressed — it's built for the moment right before you decide whether to share or publish, not for understanding why each check matters. For the reasoning behind each item, the fuller explainer on reviewing a suspicious video with AI walks through the same nine checks with context and examples.
Authenticity and caption accuracy are different questions, and this checklist covers both because most video misinformation involves real footage with a false claim attached, not a synthetic clip. Checking only for manipulation and skipping the caption check is the single most common shortcut worth avoiding.
What ConvergePanel Adds to This Checklist
- Multi-model visual review: three vision models — GPT-4o, Claude, and Gemini — independently review extracted frames
- Per-model observations: see what each model specifically flagged, not just a single verdict
- Agreement and disagreement surfaced explicitly, so a split result is visible rather than averaged away
- A documented review record you can attach to an editorial file or escalation note
Frequently asked questions
Is this checklist a substitute for the full video verification workflow?
No — it's the compressed version for a fast pre-share decision. For the reasoning behind each check and worked examples, see the full guide on reviewing a suspicious video with AI.
Which item on this checklist matters most?
Separating the caption claim from what the video actually shows. Most viral video misinformation involves authentic footage misattributed to the wrong date, place, or event — not a synthetic fake — so the caption check catches more real cases than the authenticity check alone.
What should I do if vision models disagree on a clip?
Treat the split as a reason to slow down, not a tie to break. A disagreement means the clip presents ambiguous signals that deserve human review before you act on it either way.
Can I run this checklist without an AI verification tool?
Yes for most of it — caption claim, content, date, location, and source can all be checked manually. The vision-model agreement check specifically requires running the clip through multiple AI models, which is what ConvergePanel's Video Verification mode automates.
Does completing this checklist prove a video is authentic?
No. It structures a fast review; it doesn't provide forensic proof of authenticity or certify that the caption is accurate. High-stakes clips — legal, public-safety, or election-related — warrant escalation beyond this checklist.
Explore related pages
ConvergePanel provides AI-assisted verification for informational purposes only. Not forensic analysis. Not legal evidence.
More in Video Verification
Video Authenticity Review for Fact-Checkers
Review video authenticity, source context, reposting, visual claims, and manipulation risk before publishing a fact-check.
Video Authenticity Review for Researchers
Review visual evidence, video context, source provenance, and uncertainty before using video in research or analysis.
AI Video Review for Media Teams Before Publishing
Use multiple vision models to sanity-check viral clips, visual claims, source context, and uncertainty before publishing. Not forensic proof — a structured review layer.