The idea aims to prevent users from causing harm to other users, for example negatively impacting someone's reputation, by comparing data that is published on a profile page. A successful detection would require an impersonator to trip over a series of red flags that are initiating an automated investigation of a suspected impersonation. Google mentions more than a dozen red flags that, for example, include a replication of data such as fields, photos, and metadata as well as certain vocabulary that indicates defamation.
If enough signals are detected, the technology would conclude that an impersonation takes place and take appropriate steps such as blocking a profile to prevent a user from causing harm to another user. A key limitation of the approach is, of course, that it almost entirely relies on several matches when user profiles are compared, which requires a victim to be already present on a social network. If there is no profile for the system to compare the potentially fake page to, the detection of an impersonation will fail.