Attackers compromise web servers in order to host fraudulent content, such as malware and phishing websites. While the techniques used to compromise websites are widely discussed and categorized, analysis of the methods used by attackers to identify targets has remained anecdotal. In this paper, we study the use of search engines to locate potentially vulnerable hosts. We present empirical evidence from the logs of websites used for phishing to demonstrate attackers’ widespread use of search terms which seek out susceptible web servers. We establish that at least 18% of website compromises are triggered by these searches. Many websites are repeatedly compromised whenever the root cause of the vulnerability is not addressed. We find that 19% of phishing websites are recompromised within six months, and the rate of recompromise is much higher if they have been identified through web search. By contrast, other public sources of information about phishing websites are not currently rai...