-T tests just a few bytes of the file (see -X). File::Type just guesses a file type by searching for a few magic numbers, like file. Both can't be reliable.

If you want to check for a file that contains only ASCII characters, you have to check the entire file. There is no other way.

I guess you also want to check for a sane file size, perhaps some hundred kBytes or a few MBytes. On a modern computer, slurping the entire file with that limitation is no big problem.

You may want something like this (untested):

-f $filename or die "$filename is not a file";
(-s _ < 100_000) or die "$filename is too large"; # avoid a second sta
+t() syscall by using the special handle "_"
my $blob=do {
    open my $f,'<:raw',$filename or die "Can't open $filename: $!";
    local $/; # slurp mode
    <$f>; # slurp
    # leaving the do block auto-closes $f
};
# Accept only CR, LF, TAB, and printable characters from 0x20 to 0x7E.
$blob=~/^[\r\n\t\x20-\x7E]*$/s or die "$filename is not ASCII";
[download]

If you want significantly larger files, you have to read smaller blocks (perhaps 1 MByte each), and check each block for its "ASCIIness". Abort at the first failed block.

Alexander