I think HTML::TreeBuilder only works on bytes (not characters) since it's made to work on files. I had problems getting it to work with characters and ended up decoding everything I extracted from the tree. The way to go might be

my $html_file = encode(
   'ASCII',
   $response->decoded_content(),
   Encode::FB_HTMLCREF
);

my $tree = HTML::TreeBuilder->new();
$tree->parse($html_file);
[download]

and use utf8::decode on everything extracted from the tree if necessary.

It has the advantage of handling multiple byte encoding if HTML::Parser doesn't, and one knows which encoding to use when decoding data extracted from the tree.