Here is a reasonable, if not great overview of the various studies about Alcoholics Anonymous' effectiveness:
http://slatestarcodex.com/2014/10/26/alcoholics-anonymous-much-more-than-you-wanted-to-know/
The biggest defect this essay has is that it too easily dismisses longitudinal studies, claiming they are completely ineffective because of selection bias. Unfortunately, because of the peculiar nature of AA, a proper randomized controlled trial is not feasible (the people who get sober in the control group and/or non-AA-treatment group will, as likely as not, end up going to AA meetings), so longitudinal studies are a reasonable research mechanism, albeit with the thorny self selection bias issue.
PMC4285560, which was published around the same time that blog entry was made, uses sophisticated statistical methods to minimize self selection bias; once that is done, we see that there is a specific effect from the AA program that keeps people sober.
This blog entry's strength is that it has good summaries of studies which are only available in books or behind paywalls. For example, reading its summary of Brandsma 1980, a long-time chestnut anti-steppers love to bring out, shows that the study does not actually reveal a pattern of increased binge drinking among AA members (a claim which is frequently parroted by anti-steppers, such as in the "Rational" wiki).
