Part 1: Reverse image search
This picture went viral on social media, and was claimed to show a laser beam igniting the forest fires in Hawaii, 2023. A short investigation proved that it was actually a rocket launch.
This next picture was said to show cows killed by the COVID vaccine. It did in fact show dead cows, but not because of a vaccine, but because of plant poisoning. The pictures had been stolen from an Italian newspaper.
This picture was said to show the aftermath of a Greta Thunberg talk at the music festival Glastonbury. It was in fact an old picture, taken seven years earlier.
This was said to show a Muslim family exploiting India’s welfare system, and went viral in Islamophobic contexts. It was in fact taken in Pakistan.
And last but not least, this picture was said to show a Ukrainian crisis actor, roleplaying as a victim of war crimes. It was in fact a picture taken at a movie set in 2020.
So to conclude, false claims are often backed up by ”visual proof”. Therefore, the easiest way is often to fact-check the image or video, rather than the textual claim itself.
Let’s look at a practical example, and go through the verification process step by step. This image went viral in India, and was claimed to show the country’s Prime Minister Modi in a meeting with his German counterpart Olaf Scholz. Thus far it seems to be true. But the key aspect was the portrait in the background – it was claimed to be a Western pique against Modi, reminding him of his more West-oriented predecessor Nehru.
So let's do a reverse image search – that is, to use search engines to look for similar images. Thereby, one can find old uploads, or identify pictures with visual similarities. This latter part we’ll talk about soon, but let's start by finding other uploads.
(Live Tineye demonstration)
So this proves the claim false. Scholz did not put up this portrait to belittle Modi. But let's say we instead want to identify the portrait.
For this, an ordinary search engine such as Google Images or Microsoft Bing is better suited. They look for similar pictures, rather than the exact same one.
(Live Google/Bing demonstration)
With this knowledge, we could apply this method to a plethora of other cases. Let's say we want to identify a city, a plant, a wall or even a pair of sneakers. Then we could actually use the engine’s own cropping tool, choose the visual clue that we want to look for, and scan through the results for matches. Thereby we can actually use this tool to identify places, persons or products, which is often very helpful when investigating a more complex claim.