Tailored descriptions of online educational resources can support users searching for educational resources on the World Wide Web (WWW) by helping them to assess for themselves the relevance and suitability of each resource. Suitable descriptions can be derived from the online metadata stored with each resource. The descriptions take into account user profiles to indicate which data are most relevant, and so they may be tailored to the user's needs. Descriptions derived from metadata may simply be presented in the form of tables, or else they may be presented as text using techniques from natural language generation. An evaluation of the effects of tailoring and of the different presentation methods shows that simple tailoring of the descriptions to a user model can be effective in helping users to identify resources. The evaluation also found that the text descriptions produced by natural language generation techniques are no more effective than tabular presentations of the metadata, although the concise text descriptions may still be valuable where space is limited.