# CrewAI Guardrails output issue

**URL:** <https://community.crewai.com/t/crewai-guardrails-output-issue/7310>\
**Category:** CrewAI Community Support\
**Tags:** output\_issues, crewai\
**Created:** [January 22, 2026, 10:43am UTC](https://community.crewai.com/t/crewai-guardrails-output-issue/7310 "2026-01-22T10:43:38Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Looza\_Subedy](https://sea1.discourse-cdn.com/flex025/user_avatar/community.crewai.com/looza_subedy/32/4282_2.png) [@Looza\_Subedy](https://community.crewai.com/u/Looza_Subedy)\
**Post date:** [January 22, 2026, 10:43am UTC](https://community.crewai.com/t/crewai-guardrails-output-issue/7310/1 "2026-01-22T10:43:38Z")

</div>

I am getting this error while running the guardrails

```auto
Failed to parse output into response format: 1 validation error for LLMGuardrailResult
  Invalid JSON: trailing characters at line 6 column 1 [type=json_invalid, input_value='{\n "valid":       
true,\n "...any prohibited content.', input_type=str]
    For further information visit https://errors.pydantic.dev/2.12/v/json_invalid     

```

I found that the guardrails is supposed to output in the following way:

```auto
{
    "valid": bool,
    "feedback": str
}

```

I can see in the error output that my “valid” : True, but there is still some error. Is it because the llm I am using is not able to follow the instructions properly and it is not able to output in the correct format as expected by the guardrails?

---

<div class="post-metadata">

**Author:** ![paul](https://sea1.discourse-cdn.com/flex025/user_avatar/community.crewai.com/paul/32/120_2.png) [@paul](https://community.crewai.com/u/paul)\
**Post date:** [January 22, 2026, 5:23pm UTC](https://community.crewai.com/t/crewai-guardrails-output-issue/7310/2 "2026-01-22T17:23:24Z")

</div>

Which LLM are you using? Are you able to try another one, such as GPT-5.2?