Security researchers have identified a significant shift in automated web traffic, where malicious actors are increasingly masquerading as AI agents to bypass traditional security controls. Data collected throughout October 2026 reveals that automated systems are now frequently adopting the digital identities of popular consumer AI tools to mask credential stuffing operations. In one instance, a single online marketplace recorded over 414,000 visits from traffic identifying itself as a widely used AI agent. Every one of these visits targeted the site’s sign-in page and resulted in errors, indicating that the traffic was part of a bulk attempt to test stolen username and password combinations.
The volume of this deceptive traffic has proven substantial, sometimes eclipsing legitimate activity. The fake AI agent traffic observed at the marketplace was approximately 1.5 times higher than the total volume of genuine AI agent visits recorded across five different industries during the same period. This trend highlights a growing challenge for security teams, as the identity claims provided by incoming traffic—often delivered via User-Agent strings—are becoming increasingly unreliable indicators of intent.
Conversely, legitimate AI agents are now frequently masking their own identities by presenting themselves as standard desktop browsers. This behavior was observed following the September launch of a major consumer AI platform. Shortly after the release, a Tier 1 telecommunications carrier began receiving a surge of automated traffic that grew from 100 requests per day to as many as 32,000 within three weeks. This traffic successfully navigated the carrier’s sign-in process, completed multi-factor authentication, reviewed billing information, and interacted with shopping carts.
The nature of this traffic was confirmed as automated through an analysis of browser update patterns. While legitimate customers typically take approximately two weeks to adopt a new browser version, this specific traffic stream transitioned to a new version in under 48 hours. This rapid, synchronized update cycle is characteristic of a centrally managed fleet of cloud-based browsers. The carrier ultimately classified over 270,000 of these transactions as automated but benign, allowing the traffic to continue under active monitoring rather than blocking it outright.
Distinguishing between these categories is becoming a primary focus for security operations, as the majority of AI agent traffic remains benign. Analysis of over 281,000 agent visits across five industries found that 85% of these interactions were limited to passive activities such as page reading or product comparison. Only about one in seven agent visits involved high-risk actions like account management, bookings, or financial transactions. Security experts warn that these high-risk interactions are where impostors are most likely to attempt account takeovers or fraudulent fund transfers.
Reliance on traditional bot scoring systems is proving insufficient to address these threats. At one global consumer brand, behavioral analysis was required to block 3.51 million unauthorized login attempts over a 31-day period. Approximately 82% of that malicious traffic failed to trigger conventional bot detection tools, meaning that behavioral monitoring was the only effective mechanism for identifying the threat.
The consensus among security professionals is that organizations must shift their focus from the identity claims provided by incoming traffic to the actual behavior exhibited during a session. If a visitor consistently targets sign-in pages without success, the activity should be treated as malicious regardless of the identity it assumes. Experts suggest that security teams should reserve high-friction defenses, such as rate limiting and step-up authentication, for sensitive actions like payment processing or changes to account credentials. This approach allows legitimate automated agents to continue functioning while providing a more targeted defense against the increasingly sophisticated methods used by attackers to mimic human or AI-driven traffic.





