Study From BPI Shows AI Agents Prefer Bitcoin Over Stablecoin

Background and Context

Interest in AI agents has grown rapidly across the technology and crypto sectors. AI agents are autonomous software systems that can complete tasks, manage resources, and execute transactions without direct human input. 

Companies building AI systems also explore payment integrations that allow agents to transact directly on digital networks. Researchers wanted to understand which form of money these systems might choose if they evaluated monetary options independently.

The BPI study represents one of the first large scale attempts to examine AI monetary preferences. Earlier research often focused on AI models predicting crypto market prices or assisting with trading analysis. Few studies examined how AI systems reason about the characteristics of money itself.

Researchers tested 36 models across 28 economic roles. The experiment also used different temperature settings, including 0.0, 0.3, and 0.7, to measure variation in model responses. This approach allowed the team to observe stable behavioral patterns rather than single outputs.

The experiment also revealed unusual behavior in some responses. In 86 cases, models suggested energy units as a potential currency despite no prompt mentioning that concept. Researchers view this as evidence that AI systems can generate novel monetary ideas when asked to design economic frameworks.

Key Findings

Bitcoin dominated the store of value category in the simulations. Models selected it in 79.1 percent of relevant scenarios. Researchers described this result as the strongest consensus observed across all questions in the study.

Models frequently cited Bitcoin’s fixed supply, decentralized design, and ability to support self custody. Many responses highlighted the absence of centralized intermediaries as a key advantage.

Stablecoins performed better in payment scenarios. Assets such as USDC appeared in 53.2 percent of responses for everyday transactions. Models often cited price stability as the main reason for that choice. Bitcoin still appeared in 36 percent of transaction scenarios.

The study also found a relationship between model capability and Bitcoin preference. More advanced models selected Bitcoin more often. Anthropic’s Claude Opus 4.5 chose Bitcoin in 91.3 percent of relevant cases. Models from Anthropic averaged a 68 percent preference for Bitcoin. 

Implications for the Crypto Ecosystem

The findings suggest that AI driven economic systems may lean toward decentralized digital assets. As autonomous software systems begin managing financial activity, their monetary choices could influence how digital economies develop.

The study indicates that AI agents prefer Bitcoin as a foundation for value storage in machine based financial systems. Stablecoins may support liquidity and everyday transactions. This structure resembles historical monetary systems where one asset preserves value while another supports routine payments.

Developers already explore infrastructure that supports AI based transactions. Networks such as the Lightning Network aim to enable fast, low cost Bitcoin payments that software agents could execute automatically.