How to Talk to AI: A Practical Guide to Better Results

Talking to an AI assistant is a skill. The same model can give you a brilliant answer or a useless one depending on how you ask. This guide walks through the habits that consistently produce better results — whether you're using Luma, ChatGPT, Gemini, or any other conversational AI.

1. Give it context, not just a task

AI models don't know who you are, what you've already tried, or what "good" looks like for you. A one-line prompt forces the model to guess. Front-load the context: what you're doing, who it's for, and the constraints.

Weak: "Help me write a cover letter."
Strong: "I'm applying for a junior product designer role at a fintech startup. I have 2 years of freelance experience but no full-time job. Write a 200-word cover letter that leads with a specific project I shipped, not with 'I am writing to apply'."

2. Describe the format you want

"Give me 5 bullet points." "Answer in one paragraph, no headings." "Reply as a table with columns for pros, cons, and cost." Format instructions cost you nothing and eliminate the most common frustration — getting a wall of text when you wanted a list.

3. Iterate — don't restart

The best prompt is rarely the first one. Read the first answer, then refine: "shorter", "less formal", "give me a concrete example", "rewrite this for a 10-year-old". Each turn narrows the model toward what you actually want. Starting a brand-new chat for every tweak throws away that context.

4. Ask it to think before answering

For anything reasoning-heavy — math, planning, comparing options, debugging — tell the model to work through it step by step, or to list assumptions first. This noticeably reduces the "confident but wrong" answers.

5. Anchor time-sensitive questions

AI training data has a cutoff. For anything about current events, prices, releases, or people's roles today, tell the model to search the web first, and check the date it references. A model that says "as of my last update" is warning you the answer may be stale.

6. Make it push back

AI defaults to agreeing with you. If you want a real critique, ask for it explicitly: "Play devil's advocate", "list three things that could go wrong with this plan", "what would a skeptical investor ask?". You'll get sharper thinking than if you just ask "what do you think?".

7. Use voice for messy thinking

Typing forces you to structure thoughts before they exist. Voice lets you ramble, contradict yourself, and let the AI extract the actual question. For brainstorming, processing a decision, or explaining a problem, talking is faster than typing.

Common questions