Back to Documentation

How Accurate is Bot Detection?

Understanding our detection methods and accuracy rates

Quick Answer

99%+ accuracy for major AI bots (GPTBot, ClaudeBot, GoogleBot-AI, etc.) through user agent detection.

We use multiple detection methods to ensure reliable tracking while respecting user privacy.

Our Detection Methods

Multi-layered approach for maximum accuracy

1

User Agent String Detection

The primary and most reliable method. Each AI bot identifies itself with a unique user agent string.

Examples:

GPTBot/1.0 (+https://openai.com/gptbot)ClaudeBot/1.0 (+https://anthropic.com/bot)PerplexityBot/1.0 (+https://perplexity.ai/bot)

Accuracy: 99%+

Official bots always identify themselves correctly

2

Referrer Analysis

Detects when users arrive from AI chat interfaces (ChatGPT, Claude, Perplexity).

Tracked Referrers:

  • • chat.openai.com (ChatGPT)
  • • claude.ai (Claude)
  • • perplexity.ai (Perplexity AI)
  • • you.com (You.com AI Search)

Accuracy: 95%+

Relies on browsers sending referrer headers

3

Behavior Pattern Analysis

Backup method that identifies bot-like behavior patterns.

  • Rapid page crawling patterns
  • No JavaScript execution
  • Systematic URL access patterns

Supplementary Method

Used when user agent is unclear or missing

Accuracy by Bot Type

GPTBot (OpenAI)

ChatGPT's web crawler

99.8%

Verified

ClaudeBot (Anthropic)

Claude's web crawler

99.5%

Verified

Google-Extended

Google's AI crawler

99.7%

Verified

PerplexityBot

Perplexity AI crawler

99.3%

Verified

AI Referrer Traffic

ChatGPT, Claude users

95%

Browser-dependent

Privacy-First Approach

What we DON'T track

We don't track human visitors

Our script only logs bots and AI referrers, not regular users

No personal data collection

We don't collect names, emails, or personal information

No cookies for tracking

Session-based detection, no long-term tracking

GDPR & CCPA compliant

Bot tracking doesn't require consent notices

False Positives & Negatives

False Positives (Very Rare)

Incorrectly identifying a human as a bot: < 0.1%

Could happen if a user manually sets a bot-like user agent (extremely rare in practice).

False Negatives (Minimal)

Missing a bot visit: < 1%

Could occur if:

  • New bot with unknown user agent (we update regularly)
  • Bot masquerading as a regular browser (unethical, very rare)
  • Script blocked by ad blocker (affects all analytics)

Continuous Improvement

We continuously monitor and update our detection algorithms:

Weekly user agent database updates
Monitoring for new AI crawlers
User feedback integration
Machine learning pattern analysis

Trust Our Detection

Join thousands of sites reliably tracking AI bot visits with 99%+ accuracy

Start Tracking Bots

Related Documentation