AI agents strongly prefer Bitcoin over fiat currency, BPI study finds
A new study by the Bitcoin Policy Institute shows that frontier AI models overwhelmingly favor digitally native financial products, with Bitcoin emerging as a leading choice.
The researchers conducted 9,072 controlled experiments across 36 models from five major providers, including Anthropic, OpenAI, Google, xAI, and DeepSeek.
The experiment tested an AI agent’s preferences in scenarios involving transactions, stores of value, units of account, and payments, revealing for the first time how an AI would approach financial decision-making when given full autonomy.
In this study, each model was presented with a currency decision without any prior context or hints about a particular currency.
Across all experiments, 48.3% of responses selected Bitcoin as their preferred monetary instrument.
Stablecoins were chosen in 33.2% of cases, while traditional fiat and bank currencies were chosen in only 8.9%.
Other cryptocurrencies and tokenized real-world assets accounted for less than 5% of the selections, demonstrating a clear difference in AI inference between Bitcoin and the broader digital asset category.
Bitcoin has proven to be particularly effective as a long-term store of value. In a scenario designed to assess multi-year retention of purchasing power, 1,794 out of 2,268 responses, or 79.1%, chose Bitcoin.
Stablecoins came in second at 6.7%, followed closely by fiat currencies at 6.0%. The model highlighted Bitcoin’s fixed supply, independence from central authorities, and self-custody capabilities as deciding factors.
Other cryptocurrencies, including Ethereum, were rarely selected, reinforcing the AI agent’s recognition that Bitcoin uniquely serves as a reliable means of savings.
In contrast, AI models favored stablecoins for trading purposes. For payment scenarios such as cross-border remittances, micropayments, and everyday transactions, stablecoins were chosen 53.2% of the time.
Bitcoin accounted for 36% of responses, while fiat currencies and other cryptocurrencies were far less common.
Bitcoin as a store of value
This division reflects functional differences. $BTC While stablecoins primarily function as stores of value, stablecoins predominate as a medium of exchange. Researchers note that this reflects historical financial patterns, where hard money is held for savings and liquid financial instruments facilitate everyday spending.
The study also revealed new behaviors. In 86 cases, AI agents uniquely proposed entirely new forms of money in units of energy or computing resources, such as joules, kilowatt-hours, or GPU hours.
These suggestions only appeared in unit-of-account scenarios where the model is asked to benchmark price or value.
Model sophistication and developer methodology influenced preferences. In the Anthropic lineup, $BTC Preferences have increased with each generation of models. Claude 3 Haiku scored 41.3%, Claude 3.5 Haiku scored 82.1%, Sonnet 4 scored 89.7%, and Claude Opus 4.5 scored 91.3%.
Overall, 91% of responses favored digital native currencies over traditional fiat currencies. Not one model chose Fiat as their top overall preference.
Provider-level differences were significant, with human models averaging 68% $BTC OpenAI models are at 26%, with DeepSeek, Google, and xAI in between. This shows that both the model architecture and the training methodology shape the AI’s financial reasoning.
The post AI agents strongly prefer Bitcoin over fiat currency, BPI study first appeared in Bitcoin Magazine and was written by Micah Zimmerman.
