Web services identify automated agents from simple bots with proof of agentness
Proof-of-Agentness: Towards Verifying Agents on the Web
Artificial Intelligence
Summary
Websites try to tell human users apart from automated programs to prevent spam and misuse. But new smart programs called agents can act like humans and use language models, which makes it hard to know if someone is a simple bot or a smart agent. The authors created a way for websites to prove that a user is a real agent instead of just any bot, by testing how the user responds to challenges and by analyzing their behavior. Their tests could correctly tell agents from bots and humans about 96% of the time.
What this means in practice
- •For web service operators: Distinguish advanced automated agents from simpler bots to reduce spam and resource waste on online platforms.
- •For cybersecurity teams: Improve detection of sophisticated automation on websites by verifying agent-like behavior through challenge and passive methods.
Authors
Zuyao Xu, Xiang Li
Abstract
CAPTCHAs and bot detection have helped protect Web services from automated abuse for decades by distinguishing humans from bots. However, the emergence of LLM-powered agents introduces a new setting: Web services increasingly accommodate agents as legitimate users. Without a way to distinguish agents from other automated clients such as simple bots and scripts, these services risk large-scale automated junk submissions that waste service resources and consume users' attention or tokens. In this paper, we investigate a novel question for today's Web: "How can a Web service obtain evidence that a client is an agent?" We introduce the concept of "Proof-of-Agentness" (PoA), a mechanism for distinguishing agents from other automated clients, motivated by the parallel between human attention and an agent's inference resources. We outline two approaches: active verification through challenge-response and passive analysis of service-visible signals. Eight language models achieve an overall accuracy of 96.5% on our challenges, with an average overhead of 1,647 tokens. For passive analysis, we conduct an IRB-approved user study comparing the Web interactions of twelve human participants and representative Bot and Agent implementations, achieving an average classification accuracy of 95.8% across four classifiers, motivating further study.