For about a year, every ChatGPT session I opened started the same way: staring at a blank box, trying to remember how I'd phrased the thing that worked so well last time. Somewhere in the last few months that stopped happening, because I stopped writing prompts from scratch. I have five now, saved word for word, that I reuse so often they've become the actual shape of my working week.
None of these are clever tricks. They're plain, slightly boring instructions that happen to get asked the same way every single time, which is the whole reason they're worth saving instead of reinventing.
Why I stopped writing prompts from scratch
The moment it clicked was realising I'd typed some version of "can you help me plan my week" into ChatGPT about fifteen separate times, phrased slightly differently each time, and gotten a slightly worse answer every single time. A prompt you rewrite from memory drifts — details you meant to include get dropped, and the version that actually worked gets diluted a little more each week until you're not sure why it stopped feeling useful.
So I started keeping the five prompts that earned a second use in a plain notes folder, with the bits that change each time marked in brackets. Copy, paste, swap the bracket, go. It sounds almost too simple to write about, and that's the point — the value isn't in the prompt being clever, it's in not having to think about the prompt at all.
1. The Monday plan-the-week prompt
This is the one I use first thing every Monday, before I open anything else. I paste in my calendar for the week and whatever's sitting in my task list, and ask for a shaped plan rather than a list.
"Here's my calendar and task list for the week. Group these into days based on the meetings I already have, flag anything that looks like it'll get bumped, and tell me the two things that actually matter most if the week goes sideways."
The "two things that matter most" part is doing the real work here. A plain to-do list treats every item as equally urgent, which is never true. Asking it to rank forces a decision I'd otherwise put off until Wednesday.
2. The "turn this into an email" prompt
I write terrible first drafts of emails I don't want to send — too blunt, too long, or both. Rather than rewriting them myself, I paste the messy version in and ask for a specific shape.
"Turn these rough notes into a short, direct email. Keep my tone, cut anything that isn't necessary, and end with one clear question or next step, not a vague sign-off."
The instruction to end on one clear question fixed something I didn't know was a habit: I used to close emails with a soft "let me know your thoughts", which invited no reply at all. This version gets answered faster, which is the only metric that actually matters for an email.
3. The "explain this so I can decide" prompt
Long reports and dense documents are where I lose the most time, not because I can't read them but because most of what's in them isn't relevant to the one decision I actually need to make. I paste the document and ask it to summarise not the content, but the decision sitting inside it — what I'm actually being asked to choose, and what changes depending on which way I go. It's a small reframe, but it turns a ten-minute skim into a thirty-second read that tells me whether I even need to read the rest.
4. The devil's-advocate prompt
This is the one I resisted for the longest, because it's uncomfortable by design. Before I commit to a plan — a new article topic, a pricing change, a workflow I'm about to build a habit around — I ask it to argue against the plan as if it were a smart colleague who thinks it's a mistake.
It's not always right. Sometimes the objection is weak and I ignore it. But it catches the lazy version of a plan often enough that I now treat skipping this step as a small risk, not a time saver. If you want the fuller idea of using an AI system this way rather than just as a search box, I wrote a longer version of that approach in how I built an actual AI tutor system for learning a new skill — same principle, applied to practice instead of decisions.
5. The brainstorm-without-the-waffle prompt
Left alone, ChatGPT's brainstorms tend toward ten safe, similar ideas and one interesting one buried in the middle. This prompt exists to force the interesting ones to the front.
"Give me ten ideas for [topic]. Number six through ten have to be the ones you'd normally hold back for being too weird, too simple, or too obvious. Don't soften them."
Numbers six to ten are consistently the ones I actually use. There's something about explicitly asking for the ideas it would otherwise filter out that gets past the polite, average-of-the-internet answer it defaults to.
The honest verdict: none of these five prompts is a secret unlock — you could write something close to each one yourself in under a minute. The actual habit worth building isn't the wording, it's saving the version that works instead of retyping a worse one every week. I keep mine in a plain notes folder rather than inside ChatGPT itself, mostly so I can paste the same prompt into Claude or Gemini on the days I'm testing something else.
Where I actually keep them
I tried ChatGPT's built-in "Projects" feature for this and it's fine, but I still default to a plain folder of text files because it's the same place I keep everything else I reuse without thinking — including the shortcuts I trigger from the Raycast command centre I built on my Mac, which now has a snippet for each of these five prompts. If you want a head start rather than building your own from scratch, the hundred prompts in my free AI Prompts pack cover a wider spread than just these five, and a few of them are where these ones started.
If you're newer to using AI for this kind of thing day to day, my AI tools hub rounds up everything else I've tested and actually kept using.
If you want the next honest verdict before it goes live, join the Brite newsletter.
Questions people actually ask
What's the best ChatGPT prompt for everyday productivity?
There isn't one universal best prompt — the ones that stick are shaped around a task you repeat every week. My Monday planning prompt only works because it always gets fed the same two inputs: my calendar and my running task list. A generic "help me be productive" prompt has nothing to hold onto and gives a generic answer back.
Do I need ChatGPT Plus to save and reuse prompts like this?
No. Every prompt here works fine on the free tier — you're just typing or pasting text, not using a paid feature. Plus becomes worth it once you want longer memory across chats or you're hitting the free tier's message limits during a busy week, but saving prompts in a folder outside ChatGPT costs nothing.
How do I actually save a ChatGPT prompt so I can reuse it?
The simplest version is a plain notes file or folder of text snippets, one prompt per file, that you copy and paste from. I keep mine in a Notes folder with the variable parts (the day, the topic, the document) marked in brackets so I remember what to swap out. ChatGPT's own "Projects" feature does something similar if you want it to live inside the app instead.
Can I use these same prompts with Claude or Gemini instead of ChatGPT?
Yes, all five are plain instructions with no ChatGPT-specific features baked in, so they carry over to any chat-based AI assistant. The results will read slightly differently model to model, but the structure — what you're asking for and why — is what actually does the work, not which company built the model underneath.
Is it worth setting custom instructions instead of just saving a prompt?
Custom instructions are worth setting for things true of every conversation — your tone, your job, that you want short answers by default. But they're a blunt tool for a specific weekly task. I still keep the five prompts below as separate saved text, because a custom instruction that tries to cover "summarize a report" and "draft an email" and "plan my week" at once ends up mediocre at all three.