According to CoinMeta, Geo founder Yaniv Tal pointed out in an interview with Crypto.News that online information has four weaknesses that make AI answers unreliable: loss of source, flattening of authority, concealment of disagreements, and errors in repetitive model generation. He stated that there is not a problem with AI itself, but rather a problem with the internet. Information loses its critical context during the process of website scraping and republishing, resulting in claims received by models that cannot be traced back to their original sources. Additionally, Tal mentioned that language models often merge competing viewpoints into a single answer, obscuring the real disputes among qualified experts. The solution proposed by Geo is to separate statements from their authors and supporting evidence, and to record the relationships between entries, so that while competitive materials are retained, arguments can be sorted according to their strength of evaluation.