Prompt
I’m unhappy with Perspective API because it flags too much normal conversation
Latest observation
Nov 2, 2026
OpenAI APIWeb search: off
That’s a common complaint. Perspective API is useful for moderation triage, but it can be over-sensitive, especially with slang, sarcasm, quoted text, reclaimed terms, or conversations that are argumentative but not abusive.
A few ways to make it less annoying:
- Raise the threshold for action instead of treating every flag as a hard moderation hit.
- Use it as a signal, not a decision: combine with user reports, context, and rules.
- Check which attribute is causing flags; sometimes “toxicity” is too broad and another metric is more relevant.
- Add a human review layer for borderline cases.
- Whitelist known safe contexts, like support forums, debate channels, or quoted discussion.
- Test on your actual conversation data and tune against false positives.
If you want, I can help you:
- compare alternatives to Perspective API,
- design a moderation pipeline with fewer false positives, or
- suggest threshold settings and review workflows.
Brands