Source: Npr
As geopolitical actors intensify disinformation campaigns, LLM-based chatbots are proving more effective than search engine summaries at identifying and rejecting false claims from Russia, China, and Iran. Conversational AI's reasoning architecture appears better suited than retrieval-based summaries for detecting coordinated falsehoods. This creates an irony: consumers relying on ChatGPT or Claude for news may receive more reliable information than those using Google's AI Overviews, despite chatbots' known problems with hallucinations and opacity. The gap stems from search engines' rush to summarize existing web content without additional fact-checking, leaving them vulnerable to coordinated propaganda in ways that conversational models' training and fine-tuning partially counteract.