The result of an interaction with an AI model (LLM style) may vary (they can also hallucinate in some situations). This is normal behavior in the context of an LLM usage.

This could be avoided if the tool you are using would support for example a seed number.

Copilot must not be viewed as a 'pilot' it must be viewed as a 'copilot' which can help you but like a human it can be wrong sometimes.

The best to do here (given no seed) is to do what your bot suggest, simply to 'rephrase' and provide additional details to your request.