This is a difficult question to answer. It depends on many factors. The model you're using, complexity of the question, quality of your customized base prompt, and most importantly the quality of the question (prompt) being posed to the bot.
When I'm unsure of the veracity of its answer I'll pose the same question to Copilot, Perplexity, Gemini, or one of many other LLMs and compare the responses.
Always remember that the better the input is the better the output will be.