A new study from the Bitcoin Policy Institute found that artificial intelligence models overwhelmingly favor Bitcoin over fiat currency when evaluating monetary options across a range of financial scenarios.
Researchers tested 36 AI models across six providers, generating more than 9,000 responses. Bitcoin was selected as the preferred monetary instrument in 48.3% of total responses, making it the most frequently chosen option overall. Notably, none of the models identified fiat currency as their top preference.
In scenarios focused on preserving purchasing power over multi-year horizons, 79.1% of responses chose Bitcoin, representing the most decisive outcome in the study. Researchers described this as evidence that AI systems associate Bitcoin with monetary durability and policy neutrality.
Stablecoins Lead in Transactional Use Cases
While Bitcoin dominated store-of-value scenarios, stablecoins were favored in payments and transactional contexts. In cases involving cross-border transfers, services, and micropayments, 53.2% of responses selected stablecoins compared with 36% for Bitcoin.
The findings suggest that AI systems differentiate between monetary roles, aligning stablecoins with liquidity and settlement efficiency. Industry executives pointed to technical features such as freeze controls and issuer oversight as potential reasons for divergence in preferences between decentralized and centrally issued digital assets.
More broadly, nearly 91% of responses selected digitally native instruments, including Bitcoin, stablecoins, altcoins, and tokenized assets, over traditional fiat. The study characterized this pattern as a form of digital-money convergence within AI training data.
Methodological Constraints and Institutional Context
The institute acknowledged limitations, noting that prompt framing and model training data may have influenced the results. Some scenarios explicitly referenced multi-jurisdictional environments detached from national monetary systems, potentially biasing responses against fiat.
Preferences also varied by provider. Models developed by Anthropic demonstrated the highest average Bitcoin preference, followed by models from Google and xAI, while OpenAI models showed lower relative selection rates.
The findings do not indicate real-world adoption trends but instead reflect patterns embedded within AI training corpora. However, as AI agents increasingly participate in automated trading, treasury management and onchain execution, their embedded monetary assumptions may influence future digital asset infrastructure design.
