Stop Repeating Yourself: How to Use ChatGPT Projects and Memory to Save Hours
Here’s a small, expensive habit almost everyone has: every time you open ChatGPT, you re-explain yourself. Who you are, what you do, the project you’re on, the tone you want, the format you like. Then you finally ask your question — and tomorrow you do the whole ritual again. Over a week that’s a surprising amount of typing spent just getting the AI back up to speed.
The fix isn’t a clever prompt. It’s using three features that are already sitting in the app and that most people never touch: Custom Instructions, Memory, and Projects. Set them up once and ChatGPT stops arriving at every conversation with amnesia.
Custom Instructions: your standing rules
Custom Instructions are the settings you fill in once that get applied to (almost) every chat. There are two boxes: one for who you are and one for how you want ChatGPT to respond. This is where you put the stuff you are tired of repeating.
Be concrete. Instead of “be helpful,” try things the model can actually act on:
- “I run a small e-commerce store; assume that context unless I say otherwise.”
- “Answer first, then explain. Keep it tight — I’ll ask if I want more.”
- “When you give code, no long preamble, just the code and a one-line note.”
- “Don’t hedge with disclaimers unless something is genuinely risky.”
The payoff is immediate: every new chat already knows your baseline. It’s the single highest-return five minutes you can spend in the app, and it pairs well with learning to pick one AI and actually get work done instead of hopping between tools that all need re-briefing.
Memory: what it picks up as you go
Memory is the newer, more automatic layer. When it’s on, ChatGPT quietly notes durable facts about you across conversations — that you’re learning Spanish, that your dog’s name is Biscuit, that you prefer metric units — and brings them back later without being asked. You’ll sometimes see a small “Memory updated” note when it saves something.
The trick with Memory is to treat it like a shared notebook you can edit, not a black box. A few habits keep it useful:
- Steer it out loud. Say “remember that I always want invoices formatted this way” and it will hold onto it deliberately.
- Read it occasionally. You can open the saved memories in settings and see exactly what it thinks it knows. Delete anything stale or wrong — an outdated “fact” quietly skewing answers is the main failure mode.
- Go incognito when you need to. A temporary chat won’t read or write memory, which is what you want for a one-off or anything sensitive.
Projects: a folder with a brain
Projects are where this really clicks for work. A Project is a container that holds a set of chats plus its own instructions and its own uploaded files. Everything you do inside that Project inherits that shared context automatically.
Say you’re writing a book, running a client account, or planning a house renovation. Spin up a Project, give it a one-time briefing (“This is for my client Acme; here’s their brand voice; here are the three PDFs of their guidelines”), and drop the reference files in. Now every chat you start inside that Project already knows all of it. You never re-paste the brand guide again. Start a chat about the newsletter, another about the landing page — both are already fluent in Acme.
This is the difference between one giant, messy conversation where the AI slowly loses the thread and a tidy set of focused chats that all draw from the same well. It also keeps your unrelated lives apart: the renovation Project doesn’t bleed into the client Project.
How to set it up in ten minutes
You don’t need to overhaul anything. Do this once:
- Fill in Custom Instructions with your role and your three or four strongest formatting preferences.
- Turn Memory on and, over the next week, say “remember that…” whenever you catch yourself explaining something you’ll need again.
- Create a Project for each ongoing thing in your life — work, a side hustle, a recurring personal task. Give each a short briefing and its reference files.
If you already build repeatable prompts, this stacks neatly with features like reusable ChatGPT Skills for the workflows you run over and over. Instructions set the tone, Memory carries the facts, Projects group the work, and Skills package the repeatable jobs.
The honest limits
None of this is magic, and a couple of caveats will save you frustration. Memory can be wrong — it sometimes latches onto a throwaway comment as if it were a permanent truth, so the occasional cleanup matters. It’s also tied to your account, so a teammate on their own login won’t share it. And for anything genuinely private, remember that a temporary chat is the clean-slate option that writes nothing down.
There’s also a light-touch discipline worth keeping: the more context the tool carries, the more confidently it answers — which is great until it’s confidently wrong. Keeping a habit of sanity-checking AI answers you’ll actually rely on matters more, not less, once the assistant sounds like it knows you.
The bottom line
The gap between people who find ChatGPT mildly useful and people who lean on it daily usually isn’t prompt wizardry — it’s setup. The heavy users told it who they are once and organized their work into Projects, so every conversation starts halfway to the answer. Spend the ten minutes. The version of the tool that remembers you is a different, and much faster, assistant than the one that meets you fresh every morning.
Related Reading
- ChatGPT Scheduled Tasks: Put Your AI on Autopilot for the Work You Keep Forgetting
- Make AI Actually Remember You: A 2026 Guide to Memory in ChatGPT, Claude, and Gemini
- ChatGPT Skills: Teach the AI a Job Once, Get the Same Result Every Time