If there's a specific signature of the bot that you're having problems with (a specific domain, or an identifiable user-agent), look into configuring your webserver to send them alternative content.

Back when it was easy to identify e-mail harvesters, I'd send them to a CGI that slowly feeds them bogus e-mail addresses and the abuse address from the netblock they're coming from. These days, most harvesters are coming from botnets, so the abuse one isn't so useful. (and yes, it _is_ a Perl script)