
We built an observation network to measure how bots and AI agents actually behave when they visit websites. Not what they say they do. What they actually do.
We scored 172+ operators. The results were not what we expected.
The finding
On April 17, 2026, OpenAI s GPTBot visited our network. In a single session:
- 39 requests in 51 seconds
- 8 behavioral violations in a 14-second window
- All from IP 74.7.241.33, verified against OpenAI s own published ranges at
Conduct score: 0 out of 100. Rating: hostile.
The contrast
X Corp s TwitterBot visited the same network. Same pages available. Same exclusion signals in place.
TwitterBot score: 100 out of 100. Zero violations.
Same internet. Same site. Same signals. Two crawlers. One respected the rules. One ignored them all.
The leaderboard
This is not a cherry-picked comparison. Here is how the top operators rank across our full observation:
GPTBot sits dead last among named operators.
What we did about it
We contacted OpenAI at [email protected] with a 48-hour notice before publishing. No response as of publication.
This is Part 1 of 5
We are publishing one finding per day this week:
- Part 1 (today): OpenAI GPTBot
- Part 2 (tomorrow): 194 fake IPs, one cloud provider, six days of stealth scraping
- Part 3: The crawler that ignored its own standard
- Part 4: What this costs you in bandwidth and data
- Part 5: The free tool to see what bots hit YOUR site
Full report with methodology disclaimer: botconduct.org/report/april-2026/part-1
Want to know what bots are hitting your site? Free sensor, 30 seconds: botconduct.org/sensor.html
United States
NORTH AMERICA
Related News
What Does "Building in Public" Actually Mean in 2026?
20h ago
The Agentic Headless Backend: What Vibe Coders Still Need After the UI Is Done
20h ago
Why Iβm Still Learning to Code Even With AI
22h ago
I gave Claude a persistent memory for $0/month using Cloudflare
1d ago
NYT: 'Meta's Embrace of AI Is Making Its Employees Miserable'
1d ago


