Tutorials Journalism

Verifying Found and User-Submitted Video

By Infinite Creation · Intermediate · ~30 min

A 1960s mid-century comic book style editorial illustration showing a broadcast journalist and a producer in a newsroom, with a newsroom desk, a field recorder, a wall of clocks visible in the frame.

Overview

Publishing unverified footage is one of the fastest ways a newsroom's credibility gets damaged. A video's authenticity has to be established through provenance, cross-referencing, and documented process, not assumed from how convincing it looks. This guide covers the general verification techniques that hold up regardless of which specific tools happen to be available at the time.

What You Need

Access to reverse image search (most major search engines offer this)
A way to extract still frames from the video for searching
A documented process for logging verification steps taken

Steps

Establish provenance first

Find who posted the video first, when, and from what account. A later repost or re-upload with a different claimed context is one of the most common sources of misattributed footage. Trace backward to the earliest known instance you can find.

Use reverse image and frame search

Extract a few distinct still frames and run them through reverse image search: this frequently surfaces earlier instances of the same footage, sometimes from a genuinely different event or date than claimed.

Understand what metadata can and can't prove

Embedded timestamps and location data can support a case for authenticity, but they are easy to alter, and platforms often re-encode uploads, so the metadata you download describes the platform's copy rather than the original. Treat metadata as one data point, never as conclusive proof by itself.

Cross-check location and time against visible evidence

Weather conditions, shadow direction and length, visible signage, and identifiable landmarks in the frame can be checked against the claimed time and place, inconsistencies here are one of the strongest signals that a claimed context is wrong.

Keep a documented verification trail

Record exactly what checks were run and what they found, for every piece of verified footage: this protects the newsroom if a verification is later challenged, and builds institutional knowledge over time.

Know when not to publish

If key verification steps come back inconclusive or contradictory, the responsible choice is to hold the footage rather than publish with a hedge. "Unable to independently verify" is a legitimate and sometimes necessary outcome.

Tested: what happens to a clip's date and location tags

Step 3 says embedded dates and locations are easy to alter and often stripped. We checked both on 2026-09-23 with FFmpeg 8.1.2. We made a 3 second H.264 test clip with a creation date of 1 June 2019 and a location tag holding coordinates in London, then put it through four common operations and read the tags back with ffprobe. The last column is a checksum of the video stream alone, which shows whether the pictures changed.

File	Date tag (creation_time)	Location tag	Video stream MD5
The original clip	2019-06-01 12:00 UTC	London coordinates	7eba5b532b02...
Stream copy to a new MP4 (-c copy)	missing	London coordinates	same as the original
Re-encoded (-c:v libx264)	missing	London coordinates	changed
Stream copy with -map_metadata -1	missing	missing	same as the original
Stream copy writing a new date and place	2024-03-15 08:30 UTC	New York coordinates	same as the original

A plain stream copy, the usual way to rewrap a file without re-encoding, dropped the date and kept the location. So did a re-encode. A file with no date has not necessarily been tampered with, because ordinary tools drop it. One extra option, -map_metadata -1, removed both tags and left the pictures untouched.

The last row took one command. It gave the clip a date in March 2024 and a location in New York, and the video stream's checksum matched the original exactly. Nothing in the pictures records the change. This is why step 4, checking the claim against what is visible in the frame, carries more weight than any tag.

We did not test what any social platform or messaging app does to an upload.

To read a file's tags yourself: ffprobe -v error -show_entries format_tags:stream_tags clip.mp4

Validated. Measured 2026-09-23 with FFmpeg and ffprobe 8.1.2 on macOS, Apple silicon, on MP4 files FFmpeg wrote. Phone and camera files carry more tags than this test clip, and this test did not cover them.

Pro Tips

Check multiple frames, not just the thumbnail, manipulated or miscontextualized clips sometimes only reveal inconsistencies partway through.
Look for independent corroboration from a second source before publishing footage tied to a significant claim, provenance alone is rarely sufficient on its own for high-stakes footage.
Save your own copies of source material and search results as you verify, platforms and posts can be deleted after the fact, taking your evidence trail with them if you didn't archive it.

Knowledge Base

Provenance Is the Foundation, Not a Formality

Establishing who first posted a piece of footage, and when, is the single most load-bearing step in verification, nearly every other check (metadata, cross-referencing visible evidence) is meaningless if the footage turns out to be from an entirely different event than claimed. Start here, every time.

General Techniques Outlast Any Specific Tool

Verification tools and platforms change constantly, but the underlying techniques (tracing provenance, reverse search, cross-checking physical evidence in frame) remain useful regardless of which specific service happens to be available or popular at a given moment. Learning the technique, not just a tool's interface, is what holds up over time.

Where This Fits

Checking found video is part of reporting alone, which Mobile Journalism covers from kit to filing.

Sources

Berkeley Protocol on Digital Open Source Investigations (UN Human Rights Office and the Human Rights Center, UC Berkeley, 2022), read 6 October 2026, PDF from the Human Rights Center. Paragraph 184 says some metadata may be stripped before or during uploading, or by social media applications, that platforms often transcode uploads so the metadata reflects the new file, and that where metadata have been stripped researchers should seek other ways to verify an item. No platform's own documentation was found, and no platform was tested here, so this page does not say how much any one platform removes.

FAQ

Q: Can video metadata alone prove a video is authentic?
A: No, metadata (timestamps, GPS tags, device info) can support a verification case but is also straightforward to alter or strip, and some of it may be lost when a file passes through social platforms or messaging apps, which often re-encode uploads so that the metadata then describes the new file. Treat metadata as one input among several, never as proof on its own.

Q: Do I need paid tools to verify video?
A: No. The core techniques (reverse image search on key frames, cross-checking visible landmarks and weather against the claimed time and place, checking who posted it first) rely on free, publicly available tools and careful manual work more than any specific paid verification platform.