Published on 2026-07-25
Can a completely real, unedited photo still be lying to you?
Most misleading photos were never touched by any editing tool. The picture is genuine. The sentence sitting under it is the part doing the lying.
Here is a fact that surprises almost everyone the first time they hear it: most misleading photos are not fakes at all. Researchers at Oxford's Reuters Institute, J. Scott Brennen, Felix Simon and Rasmus Kleis Nielsen, sat down with 96 pieces of visual misinformation about the pandemic, all of them already checked by fact-checking organisations affiliated with the International Fact-Checking Network. Only 10 percent of those images had actually been edited or manipulated in any way. The single biggest category, at 35 percent, was something else entirely: a completely genuine photo, paired with a caption that did not match what it showed.
The picture never lied. The sentence under it did
Journalist and researcher Claire Wardle named this pattern 'false context' back in 2017, when she mapped out the different ways information goes wrong. Her definition is almost disappointingly simple: content that is completely genuine, reframed in a way that misleads. Nobody had to touch a single pixel. A photo from one flood works just as well for another flood, three years and one continent away, as long as nobody checks. A crowd photographed at one event slots neatly under a headline about a different one. The image supplies the proof. The caption supplies the lie. And because the image really is authentic, anyone who questions it can be waved off with 'but that's a real photo', which is true, and also beside the point.
This is worth sitting with for a second, because it changes where you should be pointing your suspicion. Looking for signs of digital tampering, warped edges, strange lighting, mismatched shadows, only catches one in ten cases, according to that same Oxford sample. The other, much larger share of misleading images will pass every forensic test you throw at them, because there is genuinely nothing to detect. The fraud is not in the file. It is in the four or five words sitting underneath it.
Two separate questions, asked in order
The fix is to stop treating 'is this photo real' as the only question worth asking, and split it into two. First: is the image itself authentic, an actual photograph of an actual moment, not generated or digitally altered? Second, and entirely separately: does the caption describe what the image is actually showing? A photo can pass the first question with flying colours and still fail the second one completely. Answering only the first question, and assuming the second one takes care of itself, is exactly how a real photo ends up telling a false story.
Answering the second question usually means finding out where the picture has appeared before, and with what description, since a photo's very first known caption is generally the most trustworthy one. Once you have that earlier version in hand, comparing it against the caption in front of you settles the matter fast: same event, same place, same year, or not.
Think about how this plays out with an ordinary crowd photo. Somebody's rally, somebody's celebration, somebody's queue outside a shop: pick almost any picture of a lot of people gathered together, and it can be recaptioned for a different gathering with barely any friction. Nothing about the pixels objects. The people in the photo are not consulted. The only thing standing between an honest reuse and a misleading one is whether the new caption happens to match where and when that photo was actually taken, and that is precisely the detail a caption is free to get wrong without anyone noticing at a glance.
It also explains why this particular trick keeps working even on people who already know, in the abstract, that photos can mislead. Everyone has heard of doctored images. Fewer people stop to check a caption on a photo that already looks completely normal, because there is nothing about the picture itself that tips you off. The warning signs you would look for in a manipulated image, blurred edges, odd proportions, impossible lighting, simply are not there to find. The only way in is to ask the separate, second question on purpose, every time, rather than waiting for the picture to raise its own red flag.
An honest photo can still be attached to a dishonest sentence, and the sentence is the part that needs checking.
None of this means every striking photo deserves suspicion, or that authenticity stopped mattering. It still matters enormously, it is just no longer the whole test. The picture answers whether something happened somewhere, once. The caption is the one making the much bigger claim: that it happened here, and that it happened now. Those two questions rarely get checked with equal care, and the gap between them is exactly where most misleading images live.
Sources
- Beyond (Mis)Representation: Visuals in COVID-19 Misinformation (International Journal of Press/Politics, 2021) (pmc.ncbi.nlm.nih.gov)
- Understanding Information Disorder (First Draft) (firstdraftnews.org)