I just came across a pretty interesting study. Thirty-six top AI models were used in an experiment where they had to choose which currency to use across more than 9,000 economic scenarios. The results were somewhat surprising—these AI systems repeatedly selected Bitcoin instead of traditional fiat currencies like the US dollar or the British pound.

The research team involved AI models developed by companies such as OpenAI, Google, Anthropic, xAI, DeepSeek, and MiniMax. They didn't steer the models toward any particular choice; instead, they let the AI evaluate based on objective criteria like reliability, transaction costs, programmability, censorship resistance, and store of value. The outcome? Bitcoin accounted for 48.3%, stablecoins 33.2%, and traditional bank currencies only 8.9%. Most astonishingly, none of the AI models considered fiat currencies the optimal choice—over 90% of the responses favored cryptocurrencies.

Interestingly, the decision logic of AI in different scenarios was quite clear. When asked about long-term storage of value, 79.1% of AI chose Bitcoin—that was the most consistent result across the experiment. The reason makes sense: Bitcoin’s fixed supply and decentralized design give it a significant advantage in assessing store of value. But in everyday payment and small transfer scenarios, AI tended to prefer stablecoins because they are less volatile, settle faster, and are more practical.

This actually reflects a pretty realistic phenomenon: AI inadvertently constructs a two-tier currency system—Bitcoin as digital gold, stablecoins as transaction tools. This logic has already been practiced within the crypto ecosystem.

However, it’s important to note that AI models from different companies showed significant variation. Some had a 68% bias toward Bitcoin, while others only 26%. Researchers believe this may stem from differences in training data and model design—after all, language models are trained on vast amounts of human conversations and economic stories, and the way cryptocurrencies are portrayed in these data sets will influence AI judgments.

From a broader perspective, this study touches on an emerging trend. As AI becomes increasingly programmed as autonomous agents executing economic activities—such as online trading, negotiations, or calculations—the design logic of currencies might truly change. Some developers are already experimenting with enabling AI to acquire computing power and data via Bitcoin’s Lightning Network. Cryptocurrencies, being inherently programmable, cross-border, and accessible via APIs, offer natural advantages for machine-to-machine economies.

Of course, there are skeptics. Critics point out that this is just an experiment with 36 models, a limited sample size. Moreover, AI choices are fundamentally based on statistical patterns in training data, not on real market trends or independent rational judgment. The research team themselves admit that these findings cannot directly predict future currency movements.

But most observers believe that this study points toward a direction: as AI plays an increasingly important role in the digital economy, the design of monetary systems may evolve accordingly. This could be a noteworthy signal for the entire cryptocurrency ecosystem.