Short answerOver-refusal usually comes from vague limits, a cautious model or a scope that excludes normal questions. List what the bot may answer, give examples and test borderline questions.
Common causes
- Rules like "only answer about X" without defining X.
- Guardrail wording that is too broad ("never discuss money").
- A different model with different default caution.
- Knowledge files missing the answer, so the bot says it cannot help.
Fix
- Reproduce with a specific question.
- Add an "Allowed topics" list with examples.
- Narrow the guardrail: forbid specific things, not whole areas (templates).
- Add a file or FAQ that covers the question: knowledge problems.
- Retest, including the things it should still refuse.
Balance this against safety: made-up answers and public bot protection.
Check your GPT before you move it
Paste your instructions, file names and Actions schema into the free checker. It runs in your browser and builds a portable bundle. Nothing is uploaded.
Open the free checker