A good opening request is half the work. The other half happens in the conversation. Lots of people read the first answer, find it "meh," and close the tab. That's a shame: you should judge an AI assistant by what it becomes after two or three well-chosen follow-ups.

To build your opening request, read the six-move brief method. Here, we get practical: everyday examples, then a toolbox of follow-ups. Everything works with ChatGPT, Claude, Gemini, Le Chat by Mistral, or any other assistant.

Seven prompt examples for everyday life

Planning a week of meals

There are three of us at home, including a teenager who doesn't like cooked vegetables. Suggest dinners from Monday to Friday, 30 minutes of prep max, with a shopping list grouped by store aisle. Reuse leftovers whenever possible.

"Grouped by store aisle" saves you time at the store; "reuse leftovers" cuts down on waste.

Planning a trip

I'm going to Lisbon for four days in April with a friend. We love walking and eating local food, and we hate waiting in lines. Suggest a day-by-day plan with just one "must-see" per day. Point out what I should double-check myself before I go.

That last sentence protects you: opening hours and closures change, and the assistant can make things up with total confidence.

Understanding an official letter

Here's a letter I received (I removed my name and my case number). Explain in plain language: what I'm being asked to do, by what date, and what happens if I do nothing. Tell me who to ask if anything is unclear.

Always remove your personal information before pasting a document.

Helping with homework without doing it for them

My son is in 7th grade and is stuck on this fractions exercise. Don't give the answer. Ask him questions one at a time so he figures it out himself, and wait for his answer before moving on to the next one.

The assistant becomes a patient tutor instead of an answer vending machine.

Learning a concept

Explain inflation to me at three levels: one sentence, one paragraph, then a full explanation with an everyday example. Finish with three questions to check that I understood.

The final questions tell you whether you actually understood or just read it.

Preparing for a tricky conversation

I need to tell my roommate I don't want to share groceries anymore, without it turning into a fight. Help me prepare an opening line, two arguments, and three ways she might react, each with a calm way to respond.

Then add: "Play her role, I want to practice." Rehearsing before the real moment is an underrated use.

Sorting ideas

Here's a messy list of my ideas for my mom's birthday [paste your list]. Group them by theme, remove duplicates, and rank them from simplest to most ambitious. Don't add any new ideas for now.

Without that last instruction, the assistant buries your ideas under its own.

The follow-ups that change everything

Here's the most useful part: short phrases to send after a first answer.

"Shorter."

Reflex number one, because assistants often overdo it. Be specific if needed: "Five lines max," or "just the essentials for someone with two minutes."

"Give me three versions."

One answer is an option. Three is a choice. It works for a message, a title, a menu. You can steer it: "one plain, one warm, one with some humor."

"What's missing?"

What's missing from your answer for it to be truly useful in my situation? What questions should you have asked me?

The assistant often spots the gaps itself: a budget you didn't mention, a constraint you forgot.

"Critique your answer."

Reread your answer like a demanding critic. List its three weak points, then suggest a corrected version.

The second version is almost always better. It's the most rewarding follow-up on this list.

"Explain it like I'm 12."

When an explanation is correct but hard to digest, this follow-up forces simple words and concrete images. Then go up a notch: "Now the adult version, without jargon."

"Redo it, keeping only…"

Redo your travel plan, keeping only the free, outdoor activities.

This is the filtering follow-up: you narrow down what already exists instead of asking for everything again.

"Change the tone, not the content."

The content is good, but the style doesn't sound like you. Be specific: "more direct," "less formal," "as if I were writing to a friend."

When the conversation drifts: start over

The longer a conversation gets, the more the assistant drags along its early misunderstandings. You said "no fish" twenty messages ago, and here it is suggesting fish. It's mixing up all the instructions that have piled up.

The right move:

  1. Ask: "Summarize in a few lines what we've decided so far, including my constraints."
  2. Reread and correct that summary.
  3. Open a new conversation and paste it in as your starting point.

Pitfalls to avoid

  • Following up with a plain "no." Say why: "too expensive," "too long," "not at my level."
  • Stacking ten corrections in one message. One or two at a time, otherwise some get lost.
  • Trusting practical information without checking. Deadlines, opening hours, rules: check the original source.

These follow-ups also work for emails and meeting notes. To apply them at the office, read how to use ChatGPT at work.

And if you keep repeating the same tone or format instructions in your follow-ups, set them once and for all in your custom instructions.

Going further

To follow up well, you need to understand what an assistant really does: it generates a likely answer; it doesn't "know" everything. That's what you discover in "Light the Lantern," the first chapter of the Lantern island on Ovrelia, free and playable in your browser with no account: start the chapter. And to test your reflexes, launch the friend challenge: ten timed questions about AI, and a score to beat.

And if you'd rather build your own prompts than pick from these, the Prompt Lab guides you block by block: choose a starting example (email, image, summary, code, study plan), fill in the blocks, and copy the result.

To put these prompts to work on an everyday chore, follow our method for meal planning with AI: a weekly menu and a grocery list in a few minutes.

And when you're short on ideas (a party theme, a gift, a name for a club), see our method to brainstorm ideas with ChatGPT without settling for the most obvious suggestions.

Frequently asked questions

Do these prompt examples work with assistants other than ChatGPT?

Yes. They work with Claude, Gemini, Le Chat by Mistral, Copilot, or any other chat assistant, because they rely on how you talk to the assistant, not on a specific tool.

How many follow-ups should I try before giving up?

Two or three well-targeted follow-ups are usually enough. If the conversation is still all over the place after that, ask for a summary and start again in a new conversation.

Why does ChatGPT forget my instructions partway through a conversation?

In a long conversation, early instructions get buried among the other messages. Restate the important constraint in your follow-up, or start a fresh conversation with a clean summary.

Can I use ChatGPT to help a child with homework?

Yes, as long as you ask it not to give the answer and to guide the child with questions, one at a time. The assistant then gets them thinking instead of handing over the solution.