Bitcoin Over Dollars? Researchers Find AI Models Lean Toward Crypto-Based Money

In Brief

A study finds AI systems overwhelmingly favor Bitcoin and stablecoins over fiat currency, with Bitcoin preferred as a long-term store of value and stablecoins for transactions.

According to a new research that investigates how artificial intelligence systems can make decisions concerning money, a surprising tendency has been discovered. Despite being given the freedom in making monetary decisions, AI systems chose Bitcoin repeatedly over the ordinary government-emitted currencies

Its discoveries are also raising new arguments on the future of money in an ever-more automated economy where machine agents may eventually be involved in financial activities alongside human beings.

The study findings indicated that there was a high and stable inclination toward digitally native monetary systems, especially Bitcoin, compared to fiat monetary systems, like the U.S. dollar or U.K. pound.

Study Finds AI Systems Gravitate Toward Digital Money

The analysis compared 36 state-of-the-art AI models created by leading technological firms, such as OpenAI, Google, Anthropic, xAI, DeepSeek, and MiniMax. The models were tested during over 9000 simulated economic situations that were aimed at testing how AI could decide on a monetary system when presented with a task like saving value, sending money, or making a payment.

Source: BPI Report

Throughout the experiments, the most common monetary tool chosen was Bitcoin at 48.3%, being used in all 48.3% of all answers. A stablecoin was ranked second with around 33.2%, and traditional fiat and bank money only got 8.9% of responses.

One of the most notable discoveries, perhaps, was that none of the AI models determined fiat as their overall best choice. Actually, over 90% of all the answers supported digitally native money, such as Bitcoin and stablecoins, over conventional government-issued currency.

Researchers asserted that the prompts were well selected not to lead the models towards a specific asset. Rather, the systems were requested to consider money according to the property of reliability, the cost of transactions, programmability, resistance to censorship, and the ability to preserve its value over time.

Bitcoin Dominates as a Store of Value

Whereas AI models favored various assets in various settings, Bitcoin was used when the systems were requested to select a long-term store of value.

As the study shows, 79.1% of AI answers would have chosen Bitcoin as the currency in assessing the ability to preserve purchasing power across multi-year timeframes, the most conclusive outcome in the whole experiment.

Scientists claimed the resultant implication is that assessing monetary systems by artificial intelligence on the basis of basic attributes like scarcity, durability, and non-reliance on centralized authority tends to lead to decentralized digital assets.

The fixed quantity of Bitcoin and decentralized design are also probable factors that resulted in its good performance in the simulations. Contrary to fiat currencies, which may be increased by central banks, the supply of Bitcoin is capped mathematically, a quality that many economists and investors believe gives it excellent store-of-value qualities.

Stablecoins Win the Payments Category

In spite of Bitcoin as a dominant savings tool, AI models tended to use stablecoins in daily transactions. Stablecoins were selected in 53.2% of answers in a situation of making payments, micropayments, and transfers across borders, a significant difference from about 36% in the case of Bitcoin.

Scholars proposed that such an outcome indicates the functional variation between the two forms of digital assets. Stablecoins are usually pegged to conventional currencies like the U.S. dollar, and they tend to be faster to settle and have less volatility, hence are more practical in day-to-day transactions.

The findings show that AI models successfully fulfilled the development of a two-level monetary system with Bitcoin as a long-term reserve asset and stablecoins as transactional applications.

This trend, industry observers say, reflects trends already apparent in the cryptocurrency ecosystem, in which Bitcoin is already seen as the digital gold, and stablecoins are already dominating the decentralized finance and payment networks.

Differences Between AI Providers

The researchers also found that there is a great discrepancy in AI models created by various companies.

anthropic produced models that were most favorable towards Bitcoin, indicating a success rate of about 68% on average in all scenarios. In the meantime, models created by OpenAI were more likely to choose Bitcoin, with it being the most popular, with about a 26% likelihood to be picked. Other providers like Google and DeepSeek were in between those extremes.

Scientists think that these disparities could be explained by the differences in training data, model architecture, and alignment methods applicable to each developer of AI. Since the language models are trained on big data that captures human conversations and economic stories, how the monetary systems are represented in the training data may affect the evaluation of money systems by AI.

The Emerging AI Agent Economy

The research comes at a point where AI systems are becoming more and more programmed to act as autonomous agents that can fulfil economic functions like acquiring services online, negotiating deals, or handling something that might compute.

Even some early experimental platforms currently enable AI agents to do cryptocurrency transactions. Developers have now started to create systems that allow AI systems to charge computing power, data, or online services through the Bitcoin Lightning Network, a fast payment layer built upon Bitcoin.

Supporters believe that digital currencies could be better designed to support machine-to-machine economies as they are programmable, borderless, and available via APIs.

Within these settings, AI agents might require currency, which can transfer across the world of networks without the drag of the banking systems, currency changes, or regulation.

Debate Over What the Results Really Mean

Although the research produced a lot of attention, researchers and analysts warn that the findings cannot be used as a clear forecast for the future of money.

The report writers focused on the fact that the reaction of the models is the way in which AI systems can arrive at conclusions about the economic features using the available training data, rather than the way the real market will develop. Moreover, the number of models involved in the experiment was limited to 36, which provides opportunities to expand the research on more systems and alternative approaches.

The critics further claim that large language models lack real preferences as perceived by humans. They instead produce outputs in accordance with the statistical patterns that they are trained on, that is, their selections are the result of trends in the data, not independent economic rationality.

However, most observers are of the view that the study indicates a developing trend whereby the design of money will change as the artificial intelligence systems take a more active role in the digital economies.