As artificial intelligence continues to proliferate, more people are using AI agents daily to answer questions, search for items and complete tasks. But recently, something fundamental shifted. With Visa Intelligent Commerce, we have equipped your agent to make purchases on your behalf in a safe, secure and personalized way.
The world of web-based shopping was built for humans. Merchants have spent decades fine-tuning risk systems to detect activity that does not look human. Agent transactions often trigger the very patterns that legacy fraud systems are calibrated to block.
With agents making financial transactions, trust is the critical enabler. Consumers and financial institutions need to trust agents, and merchants need to be able to trust and understand them—or agentic commerce will not happen.
Now more than ever, merchants need to understand what is arriving at their digital doorstep. Is the agent a bot? A crawler? A fraudster? Or is it legitimately acting on behalf of a consumer? If it is representing a consumer, what is it here for, and who is it shopping for?
These are not trivial questions. They are fundamental to trust in agentic commerce. That is why we developed the Visa Trusted Agent Protocol—a framework that enables safe, agent-driven checkout by helping merchants differentiate between legitimate agents and malicious bots.
Our protocol defines the data, formats and processes an agent need to share with a merchant to ensure consistent communication and execution—to create trust and transparency. Trusted Agent Protocol answers three core questions to determine whether an agent action is legitimate:
• Is the agent authorized to engage? The protocol enables merchants to understand the identity and origin of an agent, confirming it is valid and operating with a specific shopping mission.
• Who is the consumer behind the agent? Merchants can understand consumer characteristics and provide a logged-in experience through an agent connection point.
• Can the agent complete the transaction? The protocol enables agents to present tokenized payment credentials aligned to a merchant’s preferred checkout flow so transactions can be completed successfully end to end.
Trusted Agent Protocol is intentionally open and designed on top of existing web infrastructure, allowing merchants, acquirers, Content Delivery Networks and others to adopt it with minimal integration work. With more than 175 million Visa accepting merchant locations around the world, we know they cannot practically adopt custom integrations for each emerging AI platform. Standardization is the only viable path.
Other agent protocols, including those from AI and technology companies, are complementary to Trusted Agent Protocol and generally aim for the same goal: giving merchants assurance that they can trust the agent. The key distinction is that most of these alternatives require custom engineering and new application programming interfaces (APIs). Visa’s approach is infrastructure aligned, globally scalable and designed to coexist with any agent ecosystem. With the Trusted Agent Protocol, we are targeting low code to no code implementation options.
In the era of agentic commerce, the role of Visa in the ecosystem does not change regardless of the protocol being used. We provide trust, security and authenticated value transfer. Our Trusted Agent Protocol defines the data elements agents need to share to create trust and transparency in agentic commerce and can utilize both existing web infrastructure and new protocols. In either path, tokenization plays a key role. Visa continues to make progress in embedding authenticated tokens as the primary mode of payments, with more than half of Visa’s global e-commerce volume already tokenized today, and a goal to achieve 100% tokenization.
The trajectory of agentic commerce will be shaped by both innovation and interoperability, and we are excited about what is ahead. But the landscape agents must navigate is complex. In our latest biannual Threats Report, Visa saw a 25 percent¹ increase in malicious bot initiated transactions over the past six months, with the U.S. experiencing a 40 percent¹ increase, a share expected to grow as agentic commerce scales. Automated bots already execute credential stuffing, enumeration and data-harvesting attacks at industrial scale, and AI can make these bots and other bad actors more capable. At the same time, intelligence also strengthens defenses. With the Trusted Agent Protocol, merchants gain a scalable way to distinguish a legitimate customer’s agent from hostile automation.
As the web pivots from human based to agent based commerce, we all need to inject more intelligence and capability into our websites and digital interfaces to receive this traffic effectively. The Trusted Agent Protocol gives the entire ecosystem a common language for doing exactly that.
Agentic commerce will only scale when it is trusted. Standardization will be how that trust is built. Visa’s framework is one step, an important one, in ensuring that when an AI agent shows up at checkout, merchants know exactly who it is working for.