(Image credit: VCG / Contributor / Getty Images)
Even if you’ve never tried using local AI, there’s a lot to learn from the people who do. These users are unique in that they don't accept the average chatbot as-is. Instead, they tweak settings, customize system prompts and think carefully about how a model behaves before it ever answers a question.And while that may sound technical, especially if prefer to use mainstream chatbots, there's a lesson we can learn from these users.I came across a recent XDA Developers article about improving local LLMs with a system prompt that tells the model to ask clarifying questions before responding to complex tasks. The advice was aimed at people running AI locally, but the more I thought about it, the more I realized it applies to everyone using ChatGPT, Claude or Gemini.Some people still use ChatGPT like Google. They type in a vague request, wait for an answer and then get frustrated when the response feels generic. But by reframing the prompt and telling ChatGPT to ask up to three clarifying questions, the result will make the chatbot feel more like an actual collaborator.I’m convinced this is one of the easiest ways to improve responses, whether you use local or cloud-based AI tools.Why local AI matters hereIf you are running a local model through a tool like Ollama, you can permanently customize its behavior with a Modelfile. Using the SYSTEM instruction, you can bake in rules for how the AI should behave before the conversation even starts.But trust me, this is not just a local AI trick. It is a broader lesson about how to get better answers from any chatbot.Get instant access to breaking news, the hottest reviews, great deals and helpful tips.Even OpenAI’s own prompt guidance recommends asking for clarification when missing information would materially change the answer. That is the key distinction. You do not want a chatbot that asks unnecessary questions every single time, but you do want it to know exactly when your request is too vague to answer well.Here is the system prompt I tested: "Before answering, check whether my request is missing important context. If the answer would change based on my goal, audience, budget, skill level, timeline or preferences, ask up to three targeted clarifying questions first. If the request is clear enough, answer directly and state any assumptions you are making."To be clear, you do not want to tell the AI to "ask me questions." Trust me, that will get old fast. Instead, instruct it to "ask only when the answer would change." That single constraint turns the AI away from just guessing or people pleasing into something much more useful.Why it works (according to science)
