A login page does not reveal whether an attacker used an LLM. A browser session that follows redirects, handles form changes, and submits plausible credentials could belong to a person, a conventional automation script, or an AI-assisted agent. For security owners investigating how to detect LLM driven bot traffic on a login page, the useful objective is to identify abusive authentication behavior and reduce account takeover risk.
Two configuration choices deserve comparison: IP-based rate limiting and CAPTCHA challenges. They address different signals, and neither establishes that a particular language model generated the traffic.
The attack: adaptive credential stuffing
Consider a credential-stuffing attempt using previously stolen username and password pairs. An attacker can use AI-assisted browser automation to interpret the login interface, recover from changed labels, and navigate error messages.…