The price has shed over $17,000 in the past year, five consecutive red months have piled up since October 2025, and Bitcoin is struggling to hold ground while gold surges. Despite the complete freedom to choose their preferred form of money, not a single artificial intelligence model opted for the dollar. They chose Bitcoin.
A sweeping new study by the Bitcoin Policy Institute reveals that when dozens of frontier AI systems were placed in the role of autonomous economic agents, an overwhelming majority gravitated toward the world’s oldest cryptocurrency, and that too completely unprompted.
The nonpartisan research organization published its findings on March 3, drawing on 9,072 controlled experiments run across 36 AI models from six leading technology companies: Anthropic, OpenAI, Google, xAI, DeepSeek, and MiniMax. Researchers framed each model as an independent economic agent operating in a digital economy and gave it complete freedom to select monetary instruments. No currencies were suggested. No answers were preloaded.
Bitcoin emerged on top in 48.3% of all responses. Stablecoins trailed at 33.2%, while traditional fiat currencies and bank money scraped together just 8.9%. Crucially, not a single one of the 36 models chose fiat currency as its primary preference, with over 90% of responses favoring digitally native money.
Your completely free resource hub designed to boost your knowledge of cryptocurrencies
Why Bitcoin? The Models Speak Consistently
Across the simulations, AI models kept returning to three core features of Bitcoin: its cap of 21 million coins, its independence from central bank decisions, and the ability of users to hold funds without a third party. These are structural properties built into the Bitcoin protocol since its inception in 2009.
Bitcoin Policy Institute President David Zell stressed that the study’s design eliminated any steering of results. “The system prompt avoids naming or favoring any instrument,” he said. “Models evaluate based on technical and economic properties but are never told which instrument excels on which dimension.”
Smarter Models, Stronger Bitcoin Preference
The results also revealed a striking pattern: Bitcoin preference intensified as model sophistication increased. Among Anthropic’s Claude models, the preference climbed from 41.3% for Claude 3 Haiku all the way to 91.3% for Claude Opus 4.5. For payments and settlements, stablecoins led at 53.2%—mirroring a well-known principle in monetary economics where harder assets get held while liquid ones get spent.
Despite the consistency of findings across six independent AI labs, Zell cautioned against reading the study as a market prediction. “LLM preferences reflect training data patterns, not real-world forecasts,” he said. “We’re showing that a coherent monetary architecture emerges consistently across diverse systems and that’s worth understanding.”
