Every letter is saved as one or more bytes, and different encodings use different bytes for the same letter. Read a file with the wrong one and "Café" turns into "Café". Type anything below and see the real bytes six encodings would write.
The first byte starts with 110, saying "this letter takes 2 bytes". Each following byte starts with 10, which never starts a letter, so a reader that lands mid-letter can always find the next one. The 11 remaining slots hold the number itself.
Each row saves the text with one encoding; each column reads those bytes back with another, using this browser's own decoders. The diagonal is the only place it survives. The garbled text has a name: mojibake.
Plain text files rarely say. Programs guess from clues: a byte-order mark at the start, whether the bytes form valid UTF-8, zero bytes every other position. Open a text or CSV file, or try one of the typical cases.
Name,City,Amount Zoë Müller,Zürich,€1 250 José Núñez,Málaga,€980 Anaïs Lefèvre,Besançon,€1 105