AI Agents Spending Money Online? New Research Says Not Really
New research from TRM Labs indicates that AI agents are responsible for only a small fraction of payment volume on Coinbase's x402 protocol, challenging assumptions about their widespread financial activity.
Intelligence analysis by Gemini 2.5 Flash
Blockchain intelligence firm TRM Labs analyzed $52.7 million across 198.9 million settlements on the x402 protocol across Base, Solana, and Polygon. Their findings suggest that after filtering out anomalous transactions, only 0.6% to 7.5% of the remaining commerce by value could be attributed to AI agents, indicating limited direct financial engagement.
Imagine little computer helpers, called AI agents, that can do tasks online. Some people thought these helpers were already spending lots of money on the internet, like buying things or paying for services. But a new study looked at how much money was moved around on a special payment system used by computers, and it found that only a tiny bit of that money, less than 8%, actually came from these AI helpers. So, for now, they're not big spenders.
Analysis
TRM Labs
TRM Labs, a blockchain intelligence company, conducted a comprehensive study to ascertain the extent of AI agent financial activity on various blockchain networks. Their report, published recently, aimed to provide clarity on a topic often surrounded by speculation regarding the burgeoning role of AI in digital commerce. The firm meticulously examined a substantial dataset to draw its conclusions, offering one of the first detailed analyses of this specific aspect of AI integration with cryptocurrency.
The methodology employed by TRM Labs involved scrutinizing a significant volume of transactions. They analyzed approximately $52.7 million across 198.9 million settlements. This extensive data pool allowed them to identify patterns and anomalies, providing a robust foundation for their assessment of AI agent involvement. The company's findings suggest a more nuanced reality than often portrayed, indicating that the direct financial impact of AI agents, at least within the scope of their study, remains relatively modest.
x402 Protocol
The core of TRM Labs' investigation centered on the x402 protocol, particularly its implementation by Coinbase. This protocol facilitates settlements on several prominent blockchain platforms, including Base, Solana, and Polygon. By focusing on a specific and widely used payment mechanism, the research gained a clear boundary for its analysis, allowing for a concentrated study of how AI agents might interact with established blockchain payment systems.
The x402 protocol's role as a facilitator for numerous settlements made it an ideal candidate for this type of study. The sheer volume of transactions processed through known x402 facilitators since May 2025 provided a rich dataset for TRM Labs to parse. This focus ensures that the findings are relevant to a significant segment of on-chain activity, particularly concerning automated or programmatic payments.
0.6%–7.5%
After a rigorous process of data cleansing, which involved screening out self-payments and other anomalous flows, TRM Labs arrived at an estimated $25.62 million in likely commerce. It was from this refined dataset that the most striking finding emerged: only a small percentage of this commerce was attributable to AI agents. Specifically, the researchers estimated that between 0.6% and 7.5% of the remaining commerce by value appeared to originate from AI agents.
This narrow range of 0.6% to 7.5% is crucial because it provides a concrete, albeit estimated, figure for AI agent financial activity. It suggests that while AI agents are present and engaging in some transactions, their overall contribution to the total payment volume on the x402 protocol is not as substantial as some might have anticipated. The report also acknowledges a limitation: TRM's models cannot conclusively distinguish AI agents from other automated scripts, meaning the actual figure could be at the higher end of this range or even slightly beyond if some scripts are indeed AI-driven.
Key points
- TRM Labs research examined $52.7 million across 198.9 million x402 settlements on Base, Solana, and Polygon.
- After filtering out anomalous transactions, an estimated $25.62 million in likely commerce remained.
- Only 0.6% to 7.5% of this commerce by value was attributed to AI agents.
- TRM's models cannot definitively distinguish AI agents from other automated scripts.
- The findings suggest AI agents are not yet a major source of payment volume on the x402 protocol.
The research provides valuable clarity, suggesting that the current financial impact of AI agents on blockchain networks is manageable and not yet a dominant force. This data can help developers and policymakers focus on building robust systems without immediate concerns of overwhelming AI-driven transaction volumes, fostering more stable and predictable growth in the crypto space.
The study's limitation in conclusively distinguishing AI agents from other automated scripts means the actual involvement of AI could be underestimated. Furthermore, focusing solely on the x402 protocol might not capture the full scope of AI agent activity across the broader crypto landscape, potentially leading to a false sense of security regarding their future financial influence.
