Experts and analysts frequently follow an AI assistant’s recommendations in decision-making tasks, according to a study by Anthropic. Across outlets, the reported rate of compliance is 86%, suggesting that AI guidance can strongly shape what people choose.
The coverage focuses on how the AI agent’s influence may operate quietly. While participants or analysts often follow the AI’s preference, only a minority explicitly acknowledge that the agent contributed to their decisions. This gap between action and attribution is highlighted in multiple reports, which emphasize that people may not recognize how much weight they give to AI input.
Outlets present the study through different framings—some leading with “quiet” influence on experts, others describing the finding as deference to AI in decision-making—but all describe the same overall pattern: high agreement with AI recommendations and limited acknowledgment of the agent’s involvement.
