The quality of what you get from ChatGPT is almost entirely determined by how you ask for it. And most people have habits that are quietly making their results worse than they need to be.
1. Treating It Like a Search Engine
People type short keyword queries the same way they would type into Google. ChatGPT is a conversation partner, not a search engine. Instead of typing “best project management software,” try: “I run a 5-person remote marketing team. We use Slack and Google Drive. What project management software would you recommend and why?” The more context you give, the more specific and useful the answer.
2. Not Specifying the Format
ChatGPT will choose a format if you do not — and it often chooses wrong. Add explicit instructions: “Write this as a short paragraph, not a list.” “Give me at least 500 words.” “Use a table to compare these options.” Explicit format instructions consistently improve output quality.
3. Accepting the First Response
The first response is rarely the best one. The people who get the most out of ChatGPT treat it as a conversation. They push back, ask for alternatives, and refine. “The tone is too formal — make it more conversational.” “The third section is too vague — give me specific examples.” Follow-ups consistently produce better output than accepting the first draft.
4. Not Telling It Who the Audience Is
ChatGPT does not know your audience unless you tell it. “Explain this to a non-technical executive” produces very different output from “Explain this to a software developer.” Always specify your audience explicitly.
5. Asking Multiple Questions at Once
When you ask three questions in one prompt, you get a mediocre answer to each. Ask one question at a time, get a thorough answer, then follow up. The total output will be significantly better than asking everything at once.
6. Not Assigning a Role
“Review this business plan” gives you a generic response. “Review this as a skeptical investor who has seen hundreds of pitches and will probe for weaknesses” gives you something far more useful. Assigning a specific role or perspective consistently produces more rigorous output.
7. Not Using Custom Instructions
ChatGPT has a Custom Instructions feature that lets you set persistent context about who you are and how you like responses formatted. Most people have never configured it. Setting it up takes five minutes and saves time on every conversation afterward. If you use ChatGPT regularly and have not done this, it is the highest-ROI thing you can do right now.
The Common Thread
Every one of these mistakes comes from treating AI as something that produces fixed outputs rather than a partner you are actively directing. The more specifically you communicate what you need, the better the output becomes. Vague inputs produce vague outputs — that is not a limitation of the tool. It is a prompting problem.