Prompt
I’m unhappy with our spam filters because bots keep getting through
Latest observation
Nov 2, 2026
OpenAI APIWeb search: off
That’s frustrating. If bots are getting through, the fix is usually less about “stronger spam rules” and more about adding a few layered defenses so automation has multiple things to trip over.
A practical approach:
-
Tighten the obvious bot signals
- Rate-limit form submissions and signups
- Block or challenge suspicious IPs / ASN ranges
- Detect headless browsers / known automation patterns
- Check for impossible timing (e.g. form submitted in 1–2 seconds)
-
Add friction for humans only when needed
- Use CAPTCHA or invisible challenge only on risky traffic
- Add email verification for account creation
- Require phone verification for high-risk actions
-
Harden your forms
- Honeypot fields
- Timestamp fields
- CSRF protection
- Reject requests missing expected browser headers or with malformed payloads
-
Use behavior-based detection
- Multiple signups from same device/browser fingerprint
- Repeated failures across forms
- Similar message content, URLs, or payload templates
- Sudden spikes from one country/region or proxy networks
-
Review what’s actually being bypassed
- Are bots submitting via your public form endpoint directly?
- Are they using real browsers?
- Are they abusing signup, contact, password reset, or comment forms?
- Are they getting through by creating legit-looking accounts first?
-
Measure and tune
- Log why each request was allowed/blocked
- Track false positives
- Add “quarantine” instead of outright blocking for borderline cases
- A/B test stricter rules on a subset of traffic
If you want, I can help you build a concrete anti-bot checklist or design a filtering strategy for your specific setup (e.g. website forms, signup flow, email spam, API abuse, etc.).