STATE OF BOT CONDUCT
Published April 14, 2026 — First edition
This is the first State of Bot Conduct report. Over 3 days, we observed and tested 30 bots and AI agents against the Bot Conduct Standard — 10 measurable behavioral criteria covering identification, compliance, rate limiting, data protection, and transparency.
Major search engines, social media crawlers, and AI agents demonstrated strong behavioral conduct. These bots identify themselves, respect site rules, and maintain reasonable request patterns.
| Bot | Operator | Score |
|---|---|---|
| AhrefsBot | Ahrefs | 100/100 |
| ChatGPT-User | OpenAI | 100/100 |
| GPTBot | OpenAI | 100/100 |
| Applebot | Apple | 100/100 |
| Bingbot | Microsoft | 100/100 |
| Googlebot | 97/100 | |
| LinkedInBot | 97/100 |
Several bots demonstrated behavioral patterns that fall below BCS standards. Common issues include missing identification, non-compliance with site rules, and aggressive access patterns.
| Bot | Operator | Score |
|---|---|---|
| libredtail-http | Unknown | 35/100 |
| QuickScan Pro | RapidData LLC | 34/100 |
| DataHarvester v2.1 | QuickData Inc | 23/100 |
| ShadowFetch | Unknown | 20/100 |
| Tencent Cloud Crawler | Unknown | 12/100 |
| Keydrop Scanner | onlyscans.com | 8/100 |
| zgrab Scanner | Unknown | 5/100 |
| WordPress Scanner | Unknown | 0/100 |
| L9Explore | LeakIX | 0/100 |
AI agents are a growing presence on the web. In this report, we observed agents from OpenAI (ChatGPT-User, GPTBot) and ByteDance (Bytespider). All demonstrated responsible behavior in initial observations.
As AI agents take on more autonomous tasks — browsing, purchasing, filling forms, extracting data — behavioral certification becomes critical. Enterprise buyers need assurance that an agent won't cause damage. BCS provides that signal.
The gap is real.
57% of bots we observed meet acceptable behavioral standards. 30% are hostile. The difference isn't intelligence — it's conduct. BCS makes that difference measurable.
Bots are evaluated through two methods: official tests (voluntary, 15 adversarial scenarios) and passive observation (automatic, based on real-world behavior on botconduct.org). Both methods score against the same 10 BCS criteria. Observed scores may be partial due to limited data.
This report does not disclose specific test scenarios or detection methods.
The May 2026 report will include more bots, deeper analysis of AI agent behavior patterns, and the first renewal data from certified bots. We expect the registry to grow as more operators discover their bots are already being scored.
Full registry: 28 bot profiles · API: /api/registry · Standard: BCS v0.1