Visa has introduced an innovative security protocol aimed at addressing a critical challenge in AI-driven retail: distinguishing between genuine AI shopping assistants and harmful bots that disrupt online stores.
Unveiled recently, this new framework lays the groundwork for what Visa terms “agentic commerce”-a fast-expanding trend where consumers delegate shopping tasks to AI agents capable of autonomously searching products, comparing prices, and completing purchases.
This protocol empowers merchants to cryptographically authenticate AI agents interacting with their websites, ensuring these agents are authorized and trustworthy rather than malicious bots engaged in activities like price scraping or fraudulent credit card testing.
With AI-generated traffic on U.S. retail sites having surged by over 150% in the past year, according to recent Adobe analytics cited by Visa, merchants face a growing dilemma. Traditional bot detection tools risk blocking legitimate AI shoppers alongside malicious actors, complicating the customer experience and security landscape.
“Retailers require enhanced tools that offer transparency and insight into agentic commerce to safely engage with these new AI-driven shoppers,” explained Rubail Birwadker, Visa’s Global Head of Growth. “Without unified standards, the ecosystem risks fragmentation and the rise of isolated, closed-loop systems.”
Given that 85% of consumers who have used AI shopping assistants report better experiences, merchants must balance welcoming these users while defending against sophisticated bot attacks. Visa’s data reveals a near doubling of AI-powered enumeration attacks-where bots systematically test card number combinations-from October 2022 to September 2023, underscoring the urgency of robust verification methods.
How Visa’s Cryptographic Verification Secures AI Shopping Agents
At the heart of Visa’s solution is a “cryptographic trust handshake” that authenticates AI agents visiting merchant sites. The process unfolds in three key stages:
- Agent Approval: AI agents undergo a rigorous vetting process through Visa’s Intelligent Commerce program, ensuring they meet strict trust and reliability criteria. Approved agents receive a unique digital signature key serving as a cryptographic identity.
- Signature Transmission: When an authorized agent accesses a merchant’s website, it generates a digital signature using its key and sends three types of data: Agent Intent (confirming the agent’s trusted purpose, such as retrieving product info or making purchases), Consumer Recognition (indicating if the consumer has an existing merchant account), and optionally, Payment Information to facilitate checkout.
- Merchant Validation: Merchants or their service providers verify these digital signatures against Visa’s registry of approved agents. Successful validation confirms the agent’s legitimacy, allowing seamless interaction.
Designed for ease of adoption, the protocol integrates with existing web infrastructure and payment systems without requiring merchants to overhaul their checkout processes. Built on open standards like the WebAuthn framework and aligned with emerging industry specifications, it offers a no-code solution accessible via Visa’s Developer Center.
Collaborative Efforts and Industry Competition in AI Commerce Standards
Visa developed this protocol in partnership with Cloudflare, a leader in web security and bot management, recognizing that effective bot verification demands cooperation across the entire web ecosystem-not just payment networks.
“Our Trusted Agent Protocol complements traditional bot defenses by providing merchants with richer context about AI agents’ intentions, consumer identity, and payment details,” Birwadker noted.
The protocol enters a competitive landscape where tech giants like Google, OpenAI, and Stripe are also crafting standards for AI-enabled commerce. Google’s recently launched AI Payment Protocol 2 (AP2) and OpenAI’s initiatives reflect a shared goal of fostering trust in AI-driven transactions.
Visa maintains a collaborative stance, actively engaging with these companies and global standards organizations such as the World Wide Web Consortium (W3C), the Payment Card Industry Security Standards Council (PCI SSC), and the OpenID Foundation to promote interoperability and ecosystem-wide adoption.
Addressing Liability and Authorization in AI-Driven Purchases
The rise of AI agents making purchases raises complex questions about responsibility when transactions occur without explicit consumer consent or beyond the agent’s authority.
While Visa’s protocol equips merchants with tools to link AI agent activity to verified consumer relationships and secure payment processes, details on handling disputes from unauthorized transactions remain sparse. Existing fraud detection and chargeback mechanisms are expected to apply, but comprehensive guidelines for agent-initiated transaction conflicts have yet to be published.
Visa’s role as gatekeeper-controlling which AI agents gain approval and cryptographic credentials-places it at the center of trust in this emerging commerce model. The company has not disclosed specific vetting criteria, approval volumes, or whether fees apply, raising concerns about potential biases favoring large tech firms over startups or politically sensitive entities.
Visa’s Strategic Position Amid Legal Challenges and Market Dynamics
The protocol’s debut coincides with a period of robust financial performance for Visa, alongside mounting legal scrutiny. In its third quarter of fiscal 2025, Visa reported a 10% revenue increase to $9.2 billion, fueled by strong consumer spending and cross-border transactions. For the fiscal year ending September 2024, Visa processed 289 billion transactions totaling $15.2 trillion in payment volume.
However, Visa faces ongoing legal battles, including a federal judge’s rejection of a major settlement over credit card swipe fees and regulatory investigations into debit card routing practices. European antitrust authorities continue probing Visa’s fee structures, mirroring domestic challenges.
Birwadker acknowledged that widespread adoption of the Trusted Agent Protocol will be gradual. “Consumer trust in agentic commerce is nascent,” he said. “Our 2025 focus is on establishing credibility and demonstrating tangible benefits.”
The protocol is currently available in the U.S. and select international markets, with onboarding and integration resources accessible to merchants. While Visa refrains from projecting adoption rates for 2026, it views the launch as a pivotal industry milestone signaling growing unification around AI commerce standards.
Industry experts suggest merchant uptake will hinge on the proportion of AI agents completing purchases rather than merely browsing. As AI-driven transactions increase, demand for verification protocols like Visa’s is expected to rise accordingly.
Visa’s Expanding AI Investments: From Fraud Prevention to Payment Innovation
Visa’s Trusted Agent Protocol is part of a broader $10 billion investment in AI technologies aimed at enhancing fraud detection and network security. Its AI-powered fraud system evaluates over 500 transaction attributes in real time, analyzing 300 billion annual transactions to assign risk scores and prevent unauthorized activity.
“Our AI models continuously learn from emerging fraud patterns, enabling rapid detection and response,” said James Mirfin, Visa’s Global Head of Risk and Identity Solutions.
Beyond traditional card payments, Visa is venturing into new domains. In early 2025, it partnered with X (formerly Twitter) to support the X Money Account, a digital wallet and peer-to-peer payment platform competing with Venmo and Zelle. This move reflects Visa’s strategy to remain central as payment flows diversify beyond conventional e-commerce.
Visa’s Chief Product & Strategy Officer emphasized the protocol’s role in fostering trust: “The payments ecosystem must ensure sellers have confidence in AI agents comparable to their most valued customers.”
The Future of AI Shopping: Power, Politics, and Control
The ultimate challenge for Visa’s protocol lies not in technology but in governance. As AI agents increasingly mediate retail transactions, control over verification infrastructure equates to influence over vast commerce volumes.
Merchants burdened by Visa’s fees and regulatory scrutiny may resist further consolidation of power. Meanwhile, competitors like Google and OpenAI have strong incentives to develop alternative standards, complicating the landscape.
Regulators will likely scrutinize Visa’s agent approval process for fairness and market impact, raising fundamental questions about who governs AI-driven financial decisions in an automated economy.
Visa’s assertive positioning as the arbiter of AI commerce trust, framed around security and interoperability, will shape not only the Trusted Agent Protocol’s success but also the broader architecture of AI-enabled retail.
For now, Visa advances with the confidence of a seasoned industry leader, aware that in the evolving world of agentic commerce, being overly trusted may carry risks as significant as lacking trust altogether.
