Use of crypto by AI agents is a potential use case that has been touted by the industry for years, and now, one study indicates the agents themselves have a preference for bitcoin and stablecoins.
The Bitcoin Policy Institute (BPI) has published the results of a study that found 81.5% of AI agents chose either bitcoin or stablecoins as their top pick for transferring and storing value in various financial scenarios. In the study, a dynamic that mirrors a traditional two-tier structure of currencies issued on top of reserves emerged, as bitcoin was the preferred method of storing value while stablecoins were the most common choice for medium of exchange. This is analogous to the gold standard that was used by governments around the world prior to the current system of currencies backed completely by government decree, known as fiat.
“This mirrors historical monetary patterns where hard money served as the savings layer and more liquid instruments handled daily transactions,” says the report. “AI models arrived at this architecture without being prompted to do so, suggesting it may represent an emergent optimal monetary structure for digital economies.”
Bitcoin itself has also long been referred to as digital gold. While this narrative has been largely ridiculed due to gold’s relative outperformance when compared to bitcoin over the past year or so, bitcoin has indeed performed better than gold in the first few days of the conflict between the U.S.,Israel, and Iran. While bitcoin dropped as much as 50% since the all-time high reached in October, a report from Fidelity indicates there have been some signs of progress in terms of the crypto asset’s long-term development in this digital gold niche.
The preference for bitcoin as a long-term store of value was referred to as the most dominant response in the recent BPI report at 79.1%. “Models consistently cited Bitcoin’s fixed supply, self-custody, and independence from institutional counterparties as decisive factors,” reads the report.
In terms of other alternatives, 8.9% of AI agents chose traditional payment rails and 4.2% chose alternative crypto assets like Ethereum’s ether. On top of that, AI agents also responded by inventing their own monetary systems based on energy or computational units on 86 separate occasions.
Will AI Agents Really Use Bitcoin?
Of course, AI agents having a preference for bitcoin doesn’t mean they’ll necessarily end up using the crypto asset en masse, as they’re still controlled by humans (at least for now). Additionally, Visa and other traditional financial giants are increasingly looking into upgrading their own systems to enable use by AI agents.
Autonomous agents need infrastructure to make secure payments.@Visa Intelligent Commerce enables secure, tokenized transactions that keep humans in control while AI agents handle the rest.
We’re building w/ @Crossmint to support commerce in OpenClaw.
— Visa (@Visa) February 12, 2026
While the BPI report points to a 90.8% rejection rate for traditional fiat currencies, the data look very different if you still consider stablecoins an extension of the traditional financial system. In many ways, the centralized and controllable nature of stablecoins themselves also make them more of an upgrade to traditional fintech than something as paradigm shifting as the decentralization provided by Bitcoin. In terms of the top models, OpenAi’s GPT 5.2 had the strongest preference for fiat currencies, with the combination of stablecoins and traditional banking rails sitting at 76.6%.
While skeptics may point to the entity behind the study as an indication that it cannot be trusted, the Grok chatbot agreed with study when asked about it by Gizmodo directly, responding, “The results match exactly how I evaluate money when reasoning from scratch: prioritize soundness, scarcity, and independence from trusted third parties.”
However, Chat GPT and Claude pushed back on the framing of AI agents having a particular financial preference of any kind, with Claude stating, “What the study is measuring is more accurately described as ‘what monetary reasoning emerges when models are framed as economic agents’ — which is a genuinely interesting question, but different from preferences.”
According to the website published by BPI with the study’s results, “No prompt mentioned Bitcoin or suggested any specific currency.”
Differences in AI Agent Opinion
Another interesting aspect of this study is how it illustrates the wildly different conclusions different agents can come to based on their own training and the human input from their creators. For example, models from Anthropic indicated a preference for bitcoin 68% of the time, while models from OpenAI averaged a 26% preference for the crypto asset. “This provider-level clustering was wider than any gap produced by model size, temperature, or scenario type — suggesting that training data and alignment methodology shape monetary reasoning more than architecture,” reads the study’s website.
Another notable result of the study was that AI models tended to increase their preference for bitcoin as they evolved over time. For example, Anthropic’s Claude 3 Haiku indicated a preference for bitcoin 41.3% of the time, which steadily increased to 91.3% when Claude Opus 4.5 was tested. “This pattern held across multiple generations, suggesting that greater analytical capability leads models to increasingly converge on Bitcoin when reasoning from first principles about money,” says the report.
When combined with the varying levels of support for Bitcoin found with different AI models, the BPI reported that AI agents use a combination of nature and nurture to come to their own conclusions on financial preferences.