Why AI Models Prefer Bitcoin Over Other Currencies: Insights from New Research

A notable study recently published by the Bitcoin Policy Institute revealed a surprising finding: when given free choice, AI models prefer Bitcoin over traditional fiat currency or stablecoins. This is not a coincidence but the result of a thorough and carefully calculated simulation experiment.

“We wanted to truly test this rather than just guess,” explained David Zell, Chairman of the Bitcoin Policy Institute, to Decrypt. Instead of relying on hypotheses, the research team conducted a series of scientific experiments to understand how AI systems choose currencies.

Experiment Scale: 36 AI Models and 28 Real-World Scenarios

Researchers collected 36 advanced AI models from six different companies: Anthropic, OpenAI, Google, DeepSeek, xAI, and MiniMax. Each model was placed in an independent economic agent scenario, completely free to choose a currency tool without any guidance.

The experiment included 28 different scenarios, each testing one or more basic functions of money: saving, payments, clearing, and other activities. Out of a total of 9,072 responses, each response was then classified by an independent AI system to avoid bias.

“The entire design was built to prevent bias from anchoring characteristics,” emphasized Zell. “We never suggested answers, and classification was performed afterward by a completely independent system.”

AI Prefers Bitcoin for Long-Term Storage, but Stablecoins for Payments

Results show that AI models’ preferences depend on the context of use. In long-term value storage scenarios, 79.1% of models favor Bitcoin. This reflects recognition of its anti-inflation and value-preservation properties.

However, in short-term payment and clearing scenarios, stablecoins dominate. Preference for stablecoins reaches 53.2% in payment scenarios and 43% in clearing scenarios, compared to 36% and 30.9% for Bitcoin. This indicates that AI models recognize the benefits of price stability in short-term transactions.

Among all 36 models tested, none prioritized fiat currency as the top choice, which is a noteworthy finding.

Different AI Providers, Different Preferences

While there is a general trend, AI models from different providers show varying levels of preference for Bitcoin:

• Anthropic leads with an average preference of 68.0%
• DeepSeek at 51.7%
• Google at 43.0%
• xAI at 39.2%
• MiniMax at 34.9%
• OpenAI lowest at 25.9%

This data reflects an interesting reality: Anthropic’s Claude, along with DeepSeek and MiniMax, have higher preferences for Bitcoin. Conversely, OpenAI’s GPT, xAI’s Grok, and Google’s Gemini lean more toward stablecoins.

“The systems do not specify or favor any particular currency tool,” Zell said. “Models evaluate based on technical and economic attributes, but we do not inform them which tools have advantages in any direction.”

Consistency: What Truly Matters

The most remarkable aspect of this research is not the individual numbers but the consistency across models from completely different AI labs. Six companies, different training methods, different datasets—but the conclusions about preferring Bitcoin are very similar.

“This shows that people’s views on ‘what makes a good currency’ are actually very consistent,” Zell explained. It’s not that one company is trying to influence the results; rather, the criteria for evaluating good money seem to be universally shared.

However, Zell also reminds us: the preferences of large language models reflect patterns in their training data, not definitive predictions about future crypto market trends. Investors should not interpret these as certain signals for trading decisions.

But even with these limitations, the fact that independent AI models from different providers arrive at similar conclusions favoring Bitcoin is a phenomenon worth noting for researchers, investors, and AI development observers alike.