Scholars and policy-makers have proposed Geographical Indications (GIs) as the preferred IPR to integrate traditional knowledge-based industries into the global market. In this paper I ask under what conditions GIs can address problems of poor but skilled artisans in developing countries and what are their limitations in doing so. Based on a case study of the weaving industry in the city of Banaras, that was granted a GI in 2009 for the Banarasi Sari, I argue that GIs, in order to be effective, must be designed through a participatory process in which ordinary artisans, who are the producers of knowledge, are closely involved. Further the GI must be sensitive to the dynamic nature of artisanal knowledge and allow room for market-induced changes in production methods. I show that the Banaras GI freezes current production methods in the name of craft preservation while artisans take a dynamic approach to their knowledge even as they demand protection against imitation. I also argue that a GI cannot address problems that arise out of the political economy of artisanal industries. The lessons from this case study may be applicable to other artisanal industries and to livelihoods of millions of artisans in India and elsewhere.