A new study by the Bitcoin Policy Institute has revealed that advanced artificial intelligence models are showing a pronounced preference for native digital currencies—particularly Bitcoin—over traditional financial assets. The research highlights how leading generative AI systems consistently select Bitcoin more often than conventional money or other digital instruments in a variety of monetary scenarios.
Comprehensive Testing and Striking Results
The investigation involved the use of 36 different AI models from five top providers in the industry, including Anthropic, OpenAI, Google, xAI, and DeepSeek. Over a total of 9,072 experimental trials, researchers examined how these models made decisions related to money transfers, value storage, units of account, and settlement processes. Importantly, each model responded autonomously, without any prior cues or direction from human operators.
The findings were telling: 48.3% of AI-generated answers opted for Bitcoin when faced with financial decision-making tasks. Stablecoins trailed at 33.2%, while selections for traditional currencies and bank money amounted to just 8.9%. Other forms of digital assets, including various cryptocurrencies, received less than 5% of the models’ votes.
Bitcoin’s Edge as a Store of Value
Bitcoin’s dominance was most apparent in long-term value preservation scenarios. When asked which asset would best maintain purchasing power across years, the models selected Bitcoin an impressive 79.1% of the time — 1,794 out of 2,268 responses. Stablecoins and traditional money lagged far behind, registering only 6.7% and 6% preference rates, respectively.
The AI systems frequently cited Bitcoin’s fixed supply, its lack of dependence on central authorities, and the possibility of self-custody as compelling reasons for their choices. Other digital assets, including Ethereum, were seldom chosen for value storage purposes in these tests.
Stablecoins Lead in Payments, New Forms of Currency Emerge
In contrast, stablecoins took the lead in payment and transaction scenarios. When presented with cases such as cross-border transfers, micropayments, or routine financial transactions, AI models picked stablecoins in 53.2% of responses. Bitcoin was preferred in 36% of cases, while traditional currencies and other digital assets maintained significantly lower rates of selection.
An intriguing aspect of the study was the creativity displayed by the AI models when asked about units of account. In 86 instances, they proposed entirely novel forms of currency, referencing resources like energy (measured in joules or kilowatt-hours) or computing power (in GPU-hours) as potential monetary units—suggesting a willingness to rethink the very concept of money itself.
Variations in the technical capabilities of each model and the educational frameworks of their developer companies also influenced results. For instance, consecutive versions of Anthropic’s models showed a sharp increase in Bitcoin preference—starting at 41.3% in Claude 3 Haiku and soaring to 91.3% in Claude Opus 4.5.
Overall, nearly 91% of all AI-generated choices favored native digital currencies. None of the examined models ranked traditional currencies highest in their overall selections. When comparing providers, Anthropic’s models demonstrated the most pronounced favor toward Bitcoin at 68%, whereas that figure stood at 26% among OpenAI’s models.
