that only sort of sometimes works. chrome uses some sort of AI to look at the text displayed in a way that could visually be construed as a label and use that

that only sort of sometimes works. chrome uses some sort of AI to look at the text displayed in a way that could visually be construed as a label and use that to identify the field for autocomplete. the only foolproof method i've found is to use contenteditable spans instead of input tags, which has it whole own set of problems.