The Bitcoin Policy Institute tested 36 artificial intelligence models across six AI labs to see which type of money they would choose in different financial situations. The results, published Tuesday, showed Bitcoin came out on top overall.
The study generated 9,072 responses total. A separate AI system was used to categorize those responses after the fact.
Bitcoin was chosen in 48.3% of all responses across the study, making it the most selected monetary instrument overall. Not a single one of the 36 models chose fiat currency as its top preference.
When scenarios focused on preserving purchasing power over multiple years, 79.1% of AI responses selected Bitcoin. That was the most one-sided result in the entire study.
However, stablecoins pulled ahead when the scenarios shifted to payments. For payment use cases, stablecoins were chosen in 53.2% of responses, compared to 36% for Bitcoin.
Bitwise chief investment officer Jeff Park offered a simple explanation. He said stablecoins underperformed because “they can be frozen, Bitcoin can’t.”
Researchers tested models from Anthropic, OpenAI, Google, DeepSeek, xAI, and MiniMax. Each model was treated as an independent economic actor across 28 scenarios covering saving, payments, and settlement.
Models were given complete freedom to choose monetary instruments without being given predefined options to select from.
Nearly 91% of all responses favored a digitally native instrument over traditional fiat. That included Bitcoin, stablecoins, altcoins, tokenized real-world assets, and compute units.
Anthropic models showed the highest average Bitcoin preference at 68%. DeepSeek came in second at 51.7%, followed by Google at 43%.
xAI averaged 39.2%, MiniMax 34.9%, and OpenAI models chose Bitcoin just 25.9% of the time.
Claude, DeepSeek, and MiniMax models favored Bitcoin over other cryptocurrencies. GPT, Grok, and Gemini models preferred stablecoins instead.
Zell was careful to explain what the results do and do not mean. He said the models’ preferences reflect patterns in their training data, not predictions about real-world crypto markets.
The study also acknowledged limitations. Only 36 models across six providers were tested, and the institute said it plans to expand to more models in future research.
Zell pointed out that six independent AI labs with different training methods arrived at broadly similar patterns. The institute said those consistent results across competing systems are what make the findings worth examining.
The post AI Models Prefer Bitcoin Over Fiat Currency in Bitcoin Policy Institute Study appeared first on CoinCentral.
