All input files have to be in CSV format.
The most important thing to note is that the structure of the import file should not be altered once the mapping to be used for the import has been created. That is if you set up an import mapping and then add some more columns into the CSV file, the fields will shift and almost all imports will either fail or cause unintended results.
The import schema and mapping fields can accept free text input. The import process will try to match up the text with pre-configured schemas and mappings in the system where possible, for example "Legacy" import schemas.