If I say the words “Mariupol maternity hospital”, I’m sure you can picture her. A young woman stares blankly at the camera, face bloodied, clutching a tatty quilt, in the middle of a bombsite. Like “Napalm girl” five decades earlier, there are some wartime images that cannot be unseen.

There is one way to banish that gnawing horror from the mind of the viewer, though. What if the scene as it appears really is too awful to be true? What if this woman is not a new mother, an innocent caught in the horrors of an illegal war – but instead an actor, cynically defrauding the sympathies of a right-minded global public?

That was the alternate reality the Kremlin began to construct in the hours after Russian bombs rained down on the hospital. By Moscow’s account, the facility had been cleared of patients and was, at the time of the attack, housing “neo-Nazi” Ukrainian soldiers. Had this been true, the strike would have transformed from potential war crime to legitimate military operation.

At the time, we were nearly two weeks into Russia’s full-blown invasion of Ukraine. Every day, my intrepid Channel 4 News colleagues were sending through hours of footage from the front line, documenting in brutal and often unbroadcastable detail the latest horrors perpetrated by Russian forces on Ukrainian civilians.

On the ground, the Channel 4 News foreign team, headed by our international editor Lindsey Hilsum, has investigated war crimes, while correspondent Paraic O’Brien has produced hard-hitting reports on rape as a weapon of war. All underpinned by our presenters Matt Frei and Krishnan Guru-Murthy, who have led efforts by the programme to rigorously analyse the rapidly unfolding events.

My first instinct was that the scenes from Mariupol were exactly as they appeared, that the young woman really was an imperilled civilian caught in Putin’s deliberate fire. And that instinct, it would transpire, was sadly correct. The Kremlin line would be comprehensively debunked: we would soon know that the woman and others pictured there were not actors, and that some of them did lose their lives that day, including a baby who was stillborn.

But, as a fact-checker, I know that instincts, however compelling, must always yield to due diligence. Yes, Russia had already committed appalling acts in this war. But that didn’t mean every alleged instance should be treated as another on the list without verification. Guessing is not knowing.

What, then, does verification look like? In this case, we were fortunate that journalists from reputable news outlets, including the Associated Press and Sky News, were quickly on the scene. They shared their own images of the devastation Russian artillery had wrought, and managed to track down the woman in the blanket as well as others.

First-hand accounts and images collected by multiple reliable news sources are about as good as it gets in this business. The trickier task is verifying images and videos from social media. This conundrum is not exactly new – so-called “user-generated content” has been a blessing and a curse for traditional news outlets for more than a decade. If genuine, it can provide unparalleled insight and eyewitness footage before news crews have had time to scramble to a scene.

But we also know that, for the most part, social media algorithms are designed to serve up content that is engaging, regardless of its veracity. By definition, if you’re watching a video online and it makes you feel something – thereby passing an early test of newsworthiness – that’s the first sign of potential trouble.

There are a few tricks of the trade that help sift out the fakes. At the most basic end, right-clicking a picture and running a reverse image search can sometimes tell you if it has been taken out of context – for example, if it’s years old or from another part of the world.

At its most sophisticated, image and video verification can mean forensic examination of whatever’s in shot – from a local skyline to a scrap of rubbish – or a deep technical interrogation of the post’s metadata. Bellingcat is often rightly seen as the industry leader in this, but perhaps the most memorable single example was BBC Africa Eye’s Anatomy of a Killing, which used, inter alia, the outline of a distinctive mountain range in a contested video to prove it was what it appeared: footage of a brutal killing by soldiers in Cameroon.

As in Russia’s Mariupol bombing, the veracity of the video had been denied by a state actor – in this case, the Cameroon government. BBC Africa Eye did not shy away from presenting that government’s arguments, just as many outlets – including Channel 4 News and the FactCheck team – gave airtime to the Kremlin’s claims about the Mariupol bombing.

There are no doubt some who would object to this, who argue that to ventilate the propaganda of states that are known to have an uneasy relationship with the truth is to give false equivalence between fact and fiction.

Important questions about belief

But, to my mind, this objection misses one crucial point: whether we publicise them or not, propagandists, be they state-sponsored or private individuals, will find ways to get their messages out – most often, by seeding them on social media and in fringe publications. So the choice for professional journalists is not between “allowing” the public to find out what the propagandist says or not; it’s between whether or not our viewers and readers believe us when we say we’re giving them the full picture. If a cursory glance on social media reveals that there is apparently another side to the story – a side the bogeyman mainstream media “won’t let you see” – that only serves to erode public confidence in journalists. And worse still, it could leave readers and viewers with the erroneous impression that once trusted outlets have ignored a propagandistic claim not because it’s false, but because it’s true.

That’s not to say that all fictions are worth refuting. We fact-checkers must always ask ourselves: by reporting this story, even to debunk it, am I passing on the falsehood to more people than would have ever seen it otherwise, giving it unearned credibility? There’s no hard-and-fast answer to this, but there are two tests I usually apply. First, is the false claim likely to be seen by a lot of people anyway, even if we don’t cover it? If the answer is yes, we can hardly be accused of giving the fake news undue prominence because it is already well known. Russia’s claims about the Mariupol “actors” and Cameroon’s doubt-casting over the killing video clearly pass this test. Secondly, what, if anything, can we add to this? The UK is blessed with a number of highly respected fact-checking organisations, of which FactCheck is only one. In a world of finite time and resources, FactCheck’s energies might be better deployed on another story if a competitor has already done a comprehensive job.

Examining fake or misleading news about conflicts like the one in Ukraine is just part of what we do. The majority of FactCheck’s coverage deals with domestic politics, interrogating the claims of public figures and trying to explain key concepts that will help readers and viewers understand the policy questions of the day.

Methodologically, this presents different challenges than does video verification or the interrogation of wartime propaganda. That’s partly because when you’re looking at phone footage or gathering evidence about an event, you are dealing with something that has already happened. In those cases, you are pursuing a set of falsifiable claims: is this person pictured truly who they say they are? Was this video really shot in that location? What is the chain of events? These are questions about the past. In theory, enough investigation – however difficult – will yield an answer.

By contrast, some of our most vexed political debates are over competing visions of the future. With Brexit it was “Project Fear” versus “sunlit uplands”. With the first and possible-second referendums on Scottish independence, much rests on whether Scotland would be financially better off outside the UK. During Covid, politicians and scientists clashed over what would happen if we eased or tightened restrictions.

So it is tempting to think that we must wait until these things have happened – say, for Britain to leave the EU, Scotland to become independent, or the government to change Covid rules – to find out which projections were right and which were wrong.

Certainly, there is value in looking backwards, and doing so helps hold politicians of all persuasions to account. But waiting for the future is no use in the present. Our viewers and readers are also voters and many of them look to fact-checkers and other journalists to help to decide the balance of evidence lies.

So, while we can’t yet say which is right and which is wrong, the more salient question is: how can we work out which projections are plausible, and which aren’t?

A first principle is that, whether it’s medicine or economics, good science knows its own limits. If it’s an academic article, you’d expect to see confidence intervals (“we, the researchers, are 95 per cent confident that the following will happen”) and language that caveats the findings appropriately, making clear what the researchers still don’t know or can’t reliably anticipate. Conversely, any projection that touts itself as a crystal ball is unlikely to be such.

Beyond that, I’d be wary of any projection that makes truly outlier claims. While they catch the eye, and therefore make for tempting journalistic fodder, projections that stand out from the crowd are, statistically speaking, unlikely to be accurate. Most often, there’s a reason that none of the other experts in this field have reached this conclusion. Just as it is good practice never to lean too heavily on a single opinion poll, but rather report on the general direction of travel, so we should consider a range of projections and see where they collectively point.

Ultimately, projections about the future are not predictions. We can only use them to say what is likely, not what is inevitable. That is the job, and sometime curse, of the fact-checker: to lay out what we know, and, however frustratingly, what we still don’t know.