Check this out… the Bitcoin Policy Institute recently completed some research and showed AI agents 🧡 Bitcoin.

    This is so cool. But why do people think this is?

    They tested 36 frontier AI models to determine their monetary preferences. Across 9000+ scenarios, the agents overwhelmingly chose Bitcoin.

    The actual numbers – 48% chose Bitcoin, 33% chose stables, and only 9% chose fiat.

    Clearly Bitcoin is the best.

    AI agents proven to prefer Bitcoin to any other currency
    byu/One_Commission_6986 inBitcoinBeginners



    Posted by One_Commission_6986

    3 Comments

    1. Used_Rhubarb_9265 on

      Cool experiment but I wouldn’t read too much into it yet — AI models usually pick Bitcoin because the training data frames it as the most “neutral” or censorship-resistant digital money. Still interesting though, because it shows how strong Bitcoin’s narrative is even in AI outputs.

    2. The study

      https://www.btcpolicy.org/articles/study-ai-models-overwhelmingly-prefer-bitcoin-and-digital-native-money-over-traditional-fiat

      https://www.forbes.com/sites/digital-assets/2026/03/03/ai-agents-prefer-bitcoin-over-fiat-new-study-finds/

      What ultimately matters is what AI agents will actually start using between themselves overtime which likely will be bitcoin because of development focusing on integrating BTC payments in AI , and Bitcoin having the most liquidity , security , and acceptance

      I could see a situation where stablecoins are also very popular with AI , but that comes with privacy risks , regulation and embargo restrictions , and clawback confiscation risks that Bitcoin doesn’t have.

    3. Otherwise_Wave9374 on

      These kinds of “agent preference” studies are fun, but I always wonder how much is prompt framing vs any intrinsic “reasoning”. If you ask for store-of-value properties, BTC is going to pop out because of how its discussed online.

      Id love to see: same models, multiple prompt templates, and an ablation on whether they were allowed to browse/tool-use. More on agent eval gotchas here: https://www.agentixlabs.com/blog/

    Leave A Reply
    Share via