How to Check a Piece of War Footage

The most common problem with conflict footage online is not that it is fake. It is that it is real and wrong: genuine video from a different year, a different country, or a different moment than the caption claims. Here is how to check, without special tools.

Find the earliest version first

Whatever you are looking at is probably a copy. Copies acquire captions, music, cuts and arrows on the way.

Search for distinctive phrases from the caption and for descriptions of the scene, and sort by oldest. What you want is the first posting, ideally by whoever recorded it or by the organisation that published it. Everything you conclude should rest on that version, not on the one that reached you.

This applies to famous material too. A great deal of what circulates as the Baghdad footage is a re-cut of the seventeen-minute edit, sometimes with added commentary.

The single most useful step

Find the earliest posting and work from that. Copies acquire captions, cuts, arrows and music on the way, and most bad arguments about footage are arguments about somebody's re-edit.

Check the place

You can usually confirm or kill a location claim from the frame itself.

• Signs and writing. Language, script and brand names narrow a country fast.
• Vehicles and number plates. Plate shapes and colours vary by country.
• Buildings and terrain. Roof styles, power lines, road markings, vegetation.
• Satellite comparison. If the caption names a place, look at it from above and see whether the street layout matches.

Check the time

A clip captioned as this week, showing bare trees in a place that is currently in full leaf, has answered your question already.

Check the claim, not just the clip

Ask what the caption actually asserts: who, where, when, and who did it. Then ask what in the frame supports each part. Usually the picture supports the first two and nothing at all about the fourth.

Read what is in the frame, and only that

This is the hardest discipline and it is where the Baghdad recording is such a good teacher. The footage shows shapes, movement and the results of firing. It does not show what anyone believed, what they had been told, or what was behind a wall.

Write down what is visible and what you are inferring, in two columns. Most disagreements about footage turn out to be disagreements about which column an item belongs in. The frame-by-frame page does exactly this exercise.

Treat audio with the same suspicion

Audio is easy to add, easy to translate badly, and enormously persuasive. Check whether the sound belongs to the footage, whether a translation comes from anyone accountable, and whether commentary has been laid over the original.

In the case of this archive's subject, the transcript matters so much that one detail matters: a published book quoted the cockpit radio a year before the video leaked, which gave the later transcript an independent check almost no leaked recording ever gets.

What to do when you genuinely cannot tell

Sometimes the checks run out. The frame has no signs, the terrain is generic, the earliest copy you can find is already a re-upload, and nobody credible has geolocated it. This happens often, and the right response is the one people find hardest.

Say you do not know. Not "it looks real to me", not "it is probably from there", and above all not a confident share with a caption you have not verified. An unverified clip passed along with a hedge still travels, and the hedge falls off after two shares.

There is a second habit worth building here. Notice how much a clip wants you to feel before you have established anything. Material engineered to move quickly is usually engineered, and the strength of your reaction is not evidence about the strength of the sourcing. That is true of footage you agree with as much as footage you do not.

The footage-checking routine, end to end

1. Find the earliest version and use it.
2. Establish the place from the frame.
3. Establish the time from shadows, season and imagery.
4. Separate what the caption claims from what the frame shows.
5. Split visible facts from inferences.
6. Check the audio's provenance and translation.
7. When you talk about it, say which version you watched.

That last step costs nothing and prevents an extraordinary amount of confusion. A large share of arguments about this recording, sixteen years on, are two people describing two different edits. The myths page is mostly a catalogue of that.

To see the method applied to a real recording, read the frame-by-frame analysis alongside the main account, and check any claim you meet against the myths page.