It depends on what range of possibilities you might be facing in the data. The first (and easiest) thing to do would be to check for any non-ASCII content, and you can do that on the file as a whole before XML parsing, e.g. like this:

#!/usr/bin/perl

use strict;

my $non_ascii = 0;
while (<>) {
    $non_ascii++ if ( /[^\x00-\x7f]/ );
}
warn "input contains non-ASCII\n" if ( $non_ascii );
[download]

Beyond that, if there is non-ASCII content, the actual nature of such content (what character encoding, what language) might require some guessing... Encode::Guess could be helpful, depending on what language and character encoding are actually present.

People who are smart enough to use XML with non-ASCII data usually have the clue about using utf8 encoding, and if your data falls into this category, Encode::Guess will work fine to confirm that (byte patterns in utf8 are quite distinctive and unmistakable). But if its one or another single-byte encoding (any of the cp125* or iso-8859-* character sets), you'll need to know what the intended language is in order to help Encode::Guess come up with a right answer.