Two European Central Bank economists used GPT-5.1 to analyze 8.49 million posts published across 14 Reddit forums between January 2015 and April 2026. After filtering, more than 4.1 million messages were classified by tone, topic and, where applicable, the asset discussed. The Bitcoin indicator in the R-RISI dashboard turned notably negative during the 2017-2018 and 2022-2023 downturns, but its predictive power for the BTC price lasts only one trading day. The model assigned the same label to 88.9% of posts in ten successive assessments, although no comparison with human labeling has been published. The research comes days after Bitcoin fell toward $70,000, wiping out more than $1 billion in positions within 24 hours, and after $1.44 billion in outflows from Bitcoin ETFs over five sessions.
Source: Read the original article
