Because it ignores diacritics. Searching for ā will find a, searching for š will highlight s. This may make sense for some languages where diacritics are used sparingly to indicate an aspect of pronunciation, but in my language and many others diacritics are used for entirely different letters. The letter ā is not a. Treating them as equivalent makes as little sense as treating e and o as equivalent would in English.

If I'm trying to find kāzas (wedding) in a page, I will get hits for kazas (goats). If I'm looking for šauš, a letter sequence that words about shooting begin with, I will also get hits in šausmas (horror) or sauss (dry). It's nonsense. Windows 3.1 notepad.exe could find the actual word I entered in a text file (though the input required setup), the dominant browser in 2025 cannot do that and finds entirely unrelated words because an English speaker has decided they're visually similar.