This study explores the web-mediated genre of TED (Technology, Entertainment, Design) talks, speech events whereby experts in their field disseminate knowledge from different domains (e.g. science, technology, design, global issues) addressing an audience of both co-present participants and web-users all over the world. The aim of this study is to investigate the way academics convey epistemic stance (Conrad, Biber 2000) and build up their image as experts on the TED stage. To this purpose, a contrastive analysis was carried out comparing two corpora of spoken discourse, i.e. a corpus of TED talks and a corpus of MICASE university lectures from different disciplines. Although in both genres the speaker is an academic, both the communicative purpose and audience expectations differ substantially in the two contexts under scrutiny. This comparison highlights some distinguishing traits of TED talks and provides a better insight into this genre. Adopting a corpus-based approach, attention is first paid to the most recurrent epistemic lexical verbs (ELVs) and to the use of first and second person pronouns in the two corpora. The qualitative analysis then focuses on similarities and differences in the discourse functions of the four most frequent ELVs (see, show, know, think) and of their clusters when they combine with first and second person pronouns in the two corpora. Previous studies in the field of English for Academic Purposes (Rounds 1987; Fortanet 2004; Walsh 2004; Artiga León 2006; Bamford 2009) are referred to as a starting point to investigate a novel, unexplored pragmatic space (i.e. that of TED) wherein academics accomplish purposes other than merely disseminating knowledge and training students, such as promoting their research and building up their image as experts.