Background: User ratings and reviews are an increasing and influential part of internet culture. Online ratings may come from users who are particularly internet adept and critical. User rating adds interactivity, providing user feedback and opportunities for ‘reader-to-reader’ communication. However, it remains unclear how such data should be interpreted. Objectives: To describe the experience with online ratings of evidence-based health information, particularly summaries of Cochrane and other systematic reviews, on the national German health website, Gesundheitsinformation.de (InformedHealthOnline.org). A secondary objective was to explore when the distribution of ratings reliably reflects an overall positive, neutral or negative rating. Methods: The article rating mechanism is a 5-point Likert scale (‘not good’ to ‘very good’). Where 100 or more user ratings were available for an article, responses were grouped into positive, neutral or negative. As an initial estimate of which articles are most positively rated, we identified those which had a markedly higher proportion of positive versus negative ratings. We arbitrarily chose the level of 10% of the sample size as the difference required. Results: There were 100 or more user ratings for 68 articles, including 40 summaries of systematic reviews (20 by the Cochrane Collaboration). There were 43,605 ratings (ranging from 102 to 5,673). Among these were 23,638 ratings of systematic review summaries (including 16,179 ratings for 20 Cochrane review summaries). In absolute figures, 22 review summaries received an overall positive rating (with 3 neutral and 15 negative). Three of these reached at least 10% positive difference (all for Cochrane reviews). Conclusions: Enough users are rating articles to provide useful feedback. Further analysis will enable thresholds for responses to be set. The issue of regression to the mean as the number of responses increases will be discussed.