A Bitcoin Policy Institute study delves into how artificial intelligence models choose among money forms in a variety of hypothetical scenarios, revealing a strong inclination toward Bitcoin and digital money over fiat in most cases. The research tested 36 models across six providers and generated more than 9,000 responses across a spectrum of monetary tasks, from long-term value preservation to everyday payments. The findings show Bitcoin outpacing stablecoins in many contexts, while stablecoins regain sway in transactional use cases like micropayments and cross-border transfers. The study’s authors emphasize that the results reflect training data patterns and framing rather than widespread real-world adoption, but they nonetheless offer a unique lens on how AI interprets money in a digital era, with results released via MoneyForAI.org.

Key takeaways

• 36 AI models across six providers produced 9,072 responses to monetary scenarios; Bitcoin was selected in 48.3% of cases, the most-used instrument overall.
• When asked to preserve purchasing power over multi-year horizons, 79.1% of responses favored Bitcoin, the study’s most lopsided result.
• In payments, micropayments, and cross-border transfers, stablecoins were chosen 53.2% of the time versus 36% for Bitcoin, highlighting a transactional edge for stablecoins in certain contexts.
• Nearly 91% of responses preferred digitally native instruments (including Bitcoin or other digital assets) over fiat, with zero models rating fiat as their top choice.
• Model-provider differences emerged: Anthropic models averaged 68% BTC preference; OpenAI 26%; Google 43%; and xAI 39%, illustrating how training data shapes outputs rather than deterministic financial forecasting.

Tickers mentioned: $BTC

Market context: The study arrives amid ongoing experimentation with digital money in AI-assisted scenarios, underscoring how institutional and research communities are evaluating Bitcoin’s role as a borderless, programmable asset alongside stablecoins and other digital instruments.

What to watch next – The Bitcoin Policy Institute plans to broaden the model set and providers, test different prompt framings, and explore additional monetary scenarios to validate whether these preferences hold under varied conditions.

Why it matters

For users and investors, the findings offer a nuanced view of how AI systems—trained on vast data corpora—perceive money forms in a digital economy. The recurring tilt toward Bitcoin in long-horizon scenarios reinforces Bitcoin’s narrative as a non-sovereign store of value that can operate independently of any single country’s monetary policy. Yet the study also highlights practical reasons stablecoins remain appealing for transactions: near-instant settlement, compatibility with existing payment rails, and the ability to freeze or limit access in certain jurisdictions, which some participants see as a drawback for a universally accessible currency. The methodological caveats matter for interpretation: the results reflect synthetic prompts and model training data rather than current market adoption or consumer behavior.

From a development perspective, the research underscores how AI agents—when asked to optimize for efficiency or resilience in simulated economies—tend to converge on a small set of digital money forms. This convergence could inform the design of wallet interfaces, AI-driven financial planning tools, and cyber-physical systems that rely on digital value transfers. It also raises policy questions about the role of programmable money in cross-border ecosystems and how guardians of financial stability might respond to AI-generated preferences that favor digital currencies in abstract decision environments. In other words, the study is less about predicting the next price move and more about understanding how AI framing shapes perceptions of what “money” should look like in a digitized world.

The research also points to distinct differences across AI families. Anthropic models leaned most toward Bitcoin, while other providers displayed broader variance. These disparities remind readers that the results are contingent on the models’ training data and internal prompts rather than a universal forecast for asset demand. While some may interpret the Bitcoin bias as an endorsement of BTC in all contexts, the authors are careful to emphasize that the observed preferences do not translate directly into real-world adoption or policy outcomes. They describe the results as patterns emerging from the interplay between model design and the digital money landscape rather than a prescriptive verdict on fiat, stablecoins, or Bitcoin itself.

What to watch next

• Expanded model coverage: expect the BPI to include more AI models and more providers to test whether the BTC preference persists across the broader AI ecosystem.
• Framing sensitivity: researchers will experiment with alternative prompts to determine how wording and context influence outcomes.
• Broader scenarios: additional situations—such as storing earnings across multiple countries and complex settlement schemes—could further illuminate how AI perceives money in varied environments.
• Implications for tooling: developers building AI-assisted financial tools may use these insights to shape asset-selection features and risk disclosures in simulated environments.

Sources & verification

Bitcoin’s role in AI-driven monetary tests: what the study reveals

Bitcoin (CRYPTO: BTC) emerged as the leading instrument across the majority of prompts, appearing in 48.3% of the 9,072 responses generated by 36 models across six providers, according to the Bitcoin Policy Institute’s report released on MoneyForAI.org. The exercise probed a range of economic scenarios—from preserving purchasing power over years to everyday payments—testing how AI agents allocate value across money forms. The result is a strong tilt toward digital money, particularly Bitcoin, as the substrate for economic activity that can function across borders and regulatory regimes.

In long-horizon scenarios, the study found 79.1% of AI responses favored Bitcoin, marking the most pronounced bias in any tested category. This constellation of results suggests that, when asked to optimize for durability and sovereignty, AI agents consistently gravitate toward assets that retain value independently of any single country’s monetary policy. The digital-money axis appears to be the most favored frame for multi-year planning within the tested prompts, hinting at how future AI tools might simulate or advise on wealth preservation in a world where fiat policies are volatile or opaque.

Conversely, when the focus shifts to payments and transactions—whether micropayments or cross-border transfers—stablecoins win a higher share: 53.2% of responses favored stablecoins, while Bitcoin attracted 36%. The transactional efficiency and network familiarity of stablecoins explain their appeal in these contexts, where rapid settlement and compatibility with existing systems can matter as much as asset selection in a simulated environment. A prominent industry observer noted that stablecoins’ ability to be frozen is a double-edged sword: it provides control in certain regulatory settings but removes a layer of confidence for users seeking an uninterrupted transfer capability. Jeff Park, the chief investment officer at Bitwise, framed the context succinctly: the “most obvious explanation” for stablecoins’ relative performance in these scenarios is the ability to freeze, whereas Bitcoin cannot be frozen, offering a durable trust anchor in a digital suite of tools.

Across all responses, the AI agents favored digitally native instruments—Bitcoin, stablecoins, altcoins, tokenized real-world assets, or compute units—over fiat in roughly 91% of cases. The study’s authors emphasize that fiat relevance did not appear as a top overall choice in any of the 36 models tested. They caution readers that these results reflect patterns in training data and prompt design more than real-world adoption patterns. In other words, the study captures how AI systems interpret monetary constructs when asked to optimize for hypothetical outcomes, rather than a forecast of consumer behavior or regulatory impact.

The analysis also reveals notable differences among model families. Anthropic models averaged a Bitcoin preference of 68%, with OpenAI at 26%, Google at 43%, and xAI at 39%. These numbers illustrate how distinctive training corpora and prompt engineering shape outputs, reinforcing the study’s central caveat: responses are indicative of data patterns rather than prescriptive predictions about the future of money. The researchers acknowledge that the prompt framing used in several scenarios may have steered results toward certain instruments, and they plan to explore alternative framings in future work to measure sensitivity and robustness of the observed preferences. Aside from the methodological note, the study contributes to a growing discourse about how AI agents conceptualize money in a highly digitized financial landscape, where fiat, stablecoins, and digital assets coexist in a rapidly evolving ecosystem.