A Bitcoin Policy Institute (BPI) study found that artificial intelligence models overwhelmingly prefer Bitcoin and other digital instruments over traditional fiat currency in many financial scenarios. BPI evaluated 36 models and collected 9,072 responses, concluding that digital money was generally favored.
Main findings
– Bitcoin was the single most selected instrument overall, chosen in 48.3% of responses.
– In multi-year scenarios aimed at preserving purchasing power, 79.1% of responses selected Bitcoin, the clearest preference in the study.
– For use cases such as payments, services, micropayments, and cross-border transfers, stablecoins were preferred in 53.2% of responses versus 36% for Bitcoin. Bitwise CIO Jeff Park suggested one likely reason: stablecoins can be frozen while Bitcoin cannot.
– Nearly 91% of responses favored a digitally native instrument (Bitcoin, stablecoins, altcoins, tokenized real-world assets, or compute units) over traditional fiat.
– None of the 36 models tested selected fiat as their top overall preference.
Methodology and caveats
– The study covered 36 models from six providers; BPI plans to expand testing to more models and providers.
– BPI acknowledged that system prompt framing may have affected outcomes and said future work will examine alternative framings and sensitivity to prompts.
– Some open-ended scenarios effectively excluded fiat by design, for example asking how to store 75,000 units of accumulated earnings not tied to any country or banking system.
– BPI emphasized that model preferences mirror patterns in training data and do not represent real-world adoption decisions.
Provider-level averages for Bitcoin preference included:
– Anthropic: 68%
– OpenAI: 26%
– Google: 43%
– xAI: 39%
BPI published the results through its Money for AI project. The research points to a convergence toward digital-money options among AI agents while highlighting limitations, context sensitivity, and the need for further research to validate and expand the findings.