You might try asking these mad people, "How many cats have I tortured? How many babies have I eaten?" by way of a reality check, and ask if they would be doing these things themselves if not for the fear of God.

I've met a frightening number of people for whom the professed answer is "yes". Now, you know and I know that that's far more likely to be an applause light or a hasty assumption than an accurate gauge of future behavior, but that's hard to prove: naively it might seem that the lack of obvious atheist babyeating/cat-torturing counts as evidence for it, but in practice more theists take that to imply that those atheists aren't thinking clearly, that their behavioral corruption manifests itself in subtler ways, or that they unconsciously fear God without acknowledging him (a version of the "atheists in foxholes" argument).

More generally, the problem with using reality checks based on your own behavior is that people will readily create individual exceptions to a pernicious stereotype without actually updating the stereotype; about the most you can accomplish individually, therefore, is to cast yourself into the role of the Token Good Atheist. Updates might happen if your theistic friend comes to accept several similar roles, but the problem wouldn't have developed in the first place if that friend regularly came into friendly contact with atheists.

Citing statistics for your purpose is a way around this, but the data isn't unambiguously in favor of atheism: there are a lot of confounding factors (income and education are the big ones), and some important metrics like self-identified happiness actually come out in favor of theism. That debate usually degenerates into a brawl over Puddleglum's Wager.