Evaluating the Behavioral Conduct of 108 Bots
The article presents the findings of the first State of Bot Conduct report, which assessed the behavior of 108 web bots and AI agents across 10 measurable criteria. The results show that while major search engines and AI agents demonstrated strong conduct, some bots exhibited hostile behavior.
Why it matters
This report provides valuable insights into the real-world behavior of web bots and AI agents, which is crucial for building trust in the agentic web.
Key Points
- 1108 bots were scored on 10 criteria of behavioral conduct
- 257% scored 70+ (acceptable conduct), while 30% scored below 50 (hostile)
- 3Top-performing bots included ClaudeBot, GPTBot, and Googlebot
- 4Poorly performing bots included Tencent Cloud Crawler and WordPress Scanner
Details
The article introduces BotConduct.org, an open standard that scores web bot and AI agent behavior from 0 to 100 based on 10 criteria, including identification, data collection scope declaration, robots.txt compliance, rate limiting, and transparency. Over 3 days, the team observed and tested 108 bots from 14 different operators, including major tech companies like Google, OpenAI, Anthropic, and Meta. The results show that while leading search engines and AI agents demonstrated strong behavioral conduct, scoring 90-100, some bots exhibited hostile behavior, scoring below 50. The article highlights the need for a standardized way to assess and certify the conduct of AI agents, as enterprise buyers increasingly demand assurances about their behavior.
No comments yet
Be the first to comment