The author presents an experiment which critically examines whether current state of the art text to speech synthesisers and web based VRML style graphics can be utilised to provide a multimodal persona presence in less than ideal user conditions. Audio visual anthropomorphised representations of online agents are becoming more common and their use in commercial systems has to be examined in order to ascertain whether multimodal processes such as the McGurk effect may limit their use. An experiment is outlined that is based around a web tele-shopping service with the users placed under adverse stress through information overload. The ability of the participants to successfully complete their task is objectively measured and conclusions are drawn on the suitability of current day speech synthesisers and VRML modelling for persona based human computer interface output.