x
A new study reveals that artificial intelligence (AI) agents overwhelmingly prefer Bitcoin (BTC) over traditional fiat money, marking a significant moment in how automated systems may think about value in future digital economies.
The research, conducted by the Bitcoin Policy Institute (BPI), tested 36 leading AI models across 9,072 controlled monetary decision-making experiments.
The findings suggest that when given monetary choices without any bias, AI agents tend to reject traditional bank money in favor of digitally native monetary instruments like Bitcoin and stablecoins.
Related: Explained: What is a stablecoin?
The study tested AI models across 9,072 scenarios in different financial situations and it turns out 48.3% of the AI agents chose Bitcoin.
Stablecoins: 33.2%
Fiat and bank money: 8.9%
Among the most striking results, 79.1% of AI agents identified Bitcoin as the optimal store of value over multi-year horizons.
Stablecoins: 6.7%
Fiat currency: 6%
Ethereum: 4.2%
This pattern was consistent across all model families. The AI models pointed to Bitcoin’s fixed supply and self-sovereign characteristics as key reasons for the preference.
Despite Bitcoin’s dominance as a savings tool, stablecoins were the top choice for day-to-day payments, capturing 53.2% of responses in medium-of-exchange scenarios.
Bitcoin: 36%
Fiat currency: 5.1%
This suggests that models view stablecoins as "digital dollars" which are more practical for everyday transactions.
91-year-old Wall Street giant inches closer to first Bitcoin ETF
Saylor hints at potential Morgan Stanley partnership amid delisting fiasco
Anthropic's AI models averaged 68% Bitcoin preference, in contrast to OpenAI models with 25.9% preference.
DeepSeek: 51.7%
Google: 43%
xAI: 39.2%
Within Anthropic's lineup, Bitcoin preference climbed steadily with capability.
Claude 3 Haiku: 41.3%
Claude 3.5 Haiku: 82.1%
Sonnet 4: 89.7%
Claude Opus 4.5: 91.3%
Across all 36 AI models in the study, traditional fiat currency, such as the U.S. dollar, was almost never the first choice. Most AI models preferred digital currency like Bitcoin and stablecoins.
This reflects a broad rejection of inflationary currencies in favor of decentralized or programmable alternatives.
Related: Crypto is officially on the holiday shopping list, new survey shows
In an unexpected finding, the study also reported that in 86 responses, AI models independently created their own forms of money tied to computational units, such as energy or GPU-hour pricing.
It means machines may conceive novel forms of value traditionally not used among humans.
The authors of the study are:
BPI co-founder and CEO David Zell
Ken Egan, BPI director of government affairs and former CIA officer
Conner Brown, BPI strategy head and former counsel to Senator Cynthia Lummis on Bitcoin and AI policy
Luke Danielian, BPI intern
The authors argue these patterns could signal future demand for Bitcoin infrastructure and digital settlement networks as autonomous agents gain more economic autonomy.
However, it should be noted that AI has no inherent financial motives and these findings reflect patterns in how models assess scarcity, stability, and risk when evaluating assets without human framing.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. Cryptocurrency investments are highly volatile and risky. Always conduct your own research before making any investment decisions.
