Why am I getting inconsistent AI responses for the same question or document?
Last updated: November 4, 2025
Context
You may notice that when asking the same question multiple times in chat or running the same document through checklist analysis, you receive different responses. This can happen with both chat responses and checklist AI answers, where the core information should remain consistent but you're seeing significant variations in the details provided.
Answer
There are several reasons why you might experience inconsistent AI responses:
Processing Issues
Sometimes AI generation jobs can fail during document processing, leading to incomplete or inconsistent responses. If you notice responses showing "No relevant information found" or significantly different answers, the document may need to be regenerated.
Document Age and Model Updates
Documents processed with older AI models may not perform as well as those processed with current models. If your document was crunched several months ago or longer, consider re-crunching it to take advantage of the latest AI improvements.
Question Phrasing
The way you phrase your questions can significantly impact AI responses. For example:
Using abbreviations like "LDs" instead of "liquidated damages" may cause confusion
Broad or ambiguous questions may yield inconsistent results
Legal terminology should be spelled out in full for best results
Tip: See our chat best practices with our Quick Start Guide
What to Do
If you're experiencing inconsistent responses:
Contact support at support@documentcrunch.com and provide the specific crunch link
Include details about what responses you received and what you expected
Support can regenerate AI responses or escalate to the engineering team if needed
For older documents, consider re-crunching to benefit from the latest AI improvements
Try rephrasing your questions using full terms instead of abbreviations
The engineering team continuously works to improve AI consistency and accuracy. When inconsistencies are reported, they help identify and resolve underlying issues to prevent similar problems in the future.