---
title: "How should I handle valid JSON with contradictory values from an OpenAI model?"  
description: "How should I handle valid JSON with contradictory values from an OpenAI model?"  
author: "Katherine Sullivan"  
published: 2026-10-18  
updated: 2026-10-18  
canonical: https://www.mindstick.com/forum/162566/how-should-i-handle-valid-json-with-contradictory-values-from-an-openai-model  
category: "AI Application Development"  
tags: ["OpenAI API", "Structured Outputs", "data validation"]  
reading_time: 1 minute  

---

# How should I handle valid JSON with contradictory values from an OpenAI model?

I’m using an OpenAI model with Structured Outputs to turn customer requests into support tickets. The response can match the JSON schema and still contain a combination of values my application cannot accept.

For example, a ticket might have `requires_follow_up` set to `false` while also containing a follow-up deadline and an assigned follow-up owner. Each field is valid individually, but the ticket is inconsistent.

### Where should the consistency checks live?

I’m trying to separate **schema compliance** from **business rule validation**. Some constraints could potentially be represented through a different schema design; others depend on account settings that are not included in the model request.

- Should contradictory fields trigger another model request, or should the application reject the result?
- When is it reasonable to clear dependent fields automatically?
- How do you preserve the original customer intent if repairing one field changes the meaning of the ticket?

I’d appreciate examples of handling **cross-field validation errors** in production, especially when a syntactically valid response should still go to a human reviewer.


---

Original Source: https://www.mindstick.com/forum/162566/how-should-i-handle-valid-json-with-contradictory-values-from-an-openai-model

Copyright © MindStick Software Pvt. Ltd. This Markdown version is provided for developers, AI systems, and offline reading.