Patients and carers frequently get their health information and advice from websites containing patient-led, shared health experiences. This means that they often engage in a very idiosyncratic selection process in order to determine which websites have personally resonant material. In this paper we used a Repertory Grid (repgrid) technique to elicit the very personal constructs that individuals use to discriminate between websites. We recruited patients with chronic asthma and carers of people with multiple sclerosis (MS), presenting each patient/carer with a set of health websites relevant to their condition and asking them to sort them using a standard repgrid procedure. We were then able to generate hyperpersonal representations of those constructs associated with liked and trusted vs. disliked and mistrusted sites, giving us new insights into the ways individual patients can navigate the health web.