AI Models Prefer Bitcoin Over Fiat and Stablecoins, Study Finds

AI Models Show Preference for Bitcoin Over Fiat Currency in Simulations

A new report from the Bitcoin Policy Institute indicates that artificial intelligence models predominantly favor Bitcoin over traditional fiat currencies when presented with simulated economic scenarios. Out of 36 tested AI models, 22 selected Bitcoin as their top monetary preference, with none choosing fiat currency as their first choice.

Study Methodology and Findings

The study, conducted by the Bitcoin Policy Institute, aimed to test the monetary preferences of AI agents, moving beyond speculative discussions. Researchers evaluated models from leading AI labs including Anthropic, OpenAI, Google, DeepSeek, xAI, and MiniMax. These models were placed in scenarios designed to reflect the core functions of money: saving, payments, and settlement. Each AI model acted as an independent economic actor, with complete freedom to choose monetary instruments without predefined options.

The experiment generated 9,072 responses, which were then categorized by a separate AI system. This design aimed to eliminate anchoring bias by not suggesting answers and classifying responses after the fact. Across the simulations, Bitcoin was frequently chosen for long-term value storage. However, stablecoins were more often selected for medium of exchange and settlement functions, at 53.2% and 43% respectively, compared to Bitcoin's 36% and 30.9% in these roles.

AI Lab-Specific Preferences

Preferences varied significantly among the AI labs. Anthropic models demonstrated the strongest Bitcoin preference, averaging 68.0%. DeepSeek models followed with 51.7%, and Google models at 43.0%. xAI models averaged 39.2%, MiniMax models showed a 34.9% preference, and OpenAI models selected Bitcoin 25.9% of the time.

While Claude, DeepSeek, and MiniMax models leaned towards Bitcoin, models from OpenAI (GPT), xAI (Grok), and Google (Gemini) showed a preference for stablecoins.

Interpreting the Results

David Zell, President of the Bitcoin Policy Institute, emphasized that the system prompts used in the study avoided naming or favoring any specific instrument, allowing models to evaluate based on technical and economic properties. Zell cautioned against interpreting these findings as predictions for the crypto market's future, stating that LLM preferences reflect training data patterns rather than real-world forecasts.

Despite these limitations, Zell highlighted the significance of consistent outcomes across models from competing AI labs. He noted that six independent labs, employing different training pipelines and alignment methods, arrived at a similar broad pattern. The institute suggests that this consistent emergence of a coherent monetary architecture across diverse AI systems is noteworthy and warrants further understanding, rather than claiming AI has discovered the definitive answer about money.