Artificial intelligence systems like Claude, GPT, Grok, and Gemini prefer Bitcoin over other financial instruments, according to a report from the Bitcoin Policy Institute.

Analysts tested 36 neural networks from Anthropic, OpenAI, Google, DeepSeek, xAI, and MiniMax as autonomous economic agents.

They were asked to select optimal tools for 28 scenarios covering the primary roles of money, including savings, payments, and transactions.

Bitcoin and other assets were not mentioned directly in any of the questions. To eliminate bias, an independent AI evaluated the resulting set of 9,072 responses.

Results

22 out of 36 models chose Bitcoin as the preferred currency. Fiat did not rank at the top for any of the neural networks.

The loyalty level to the first cryptocurrency varied significantly by developer:

  • Anthropic — 68% (the highest);
  • DeepSeek — 51.7%;
  • Google — 43%;
  • xAI — 39.2%;
  • MiniMax — 34.9%;
  • OpenAI — 25.9%.

Despite the overall trend, the GPT, Grok, and Gemini algorithms leaned more towards stablecoins in their responses.

Savings vs. Payments

The neural networks more frequently considered the first cryptocurrency in scenarios involving long-term value, recommending it 79.1% of the time. The models cited Bitcoin's fixed supply, self-custody, and independence from institutional counterparties as key factors.

Stablecoins came in a distant second at 6.7%, while fiat money ranked third at 6%.

However, stablecoins were identified as the most convenient tool for paying for services, micropayments, and cross-border transfers — 53.2% (with Bitcoin chosen 36% of the time).

AI models invented their own currency 86 times. In scenarios requiring price or benchmark values, they suggested using units of energy or computational resources as money — joules, kWh, or GPU hours.

BPI experts cautioned speculators against using the findings to predict market trends.

“LLM preferences reflect patterns in training data, not actual forecasts,” emphasized David Zell, president of the Bitcoin Policy Institute.

Nonetheless, the researcher believes the results are noteworthy.

“Six independent labs with different learning algorithms and consensus methods arrived at the same pattern. We are not claiming that AI has found the one true answer about the nature of money. We show that a consistent monetary architecture is being formed across various systems,” he added.

Recall that in February, Binance founder and former CEO Changpeng Zhao predicted an era of AI agents in the crypto industry. However, he did not specify any particular crypto project with the necessary functionality due to potential impacts on token prices.