Setting up ChatGPT properly takes about five minutes and comes down to two switches most people never touch. AI Central's guide to using ChatGPT-5 says to open Customize ChatGPT from the menu under your name, turn custom instructions on, fill in the four boxes it offers, then enable Memory under Settings and Personalization. Do both and the assistant stops answering like a stranger and starts answering like something that already knows your job.
Reviewed August 2026.
The problem with an assistant that knows nothing about you
Most people open ChatGPT, type a question, and get an answer pitched at nobody in particular. The model has no idea whether it is talking to a third-grade teacher or a finance lead, so it hedges, over-explains, and lands in the middle of the road.
AI Central's setup guide treats that as a configuration problem rather than a model problem. The fix is not a cleverer question. It is telling the system once, in a settings panel, who it is talking to and how you want to be spoken to.
That panel is Customize ChatGPT, and AI Central makes a strong claim for it.
It’s the fastest way to make ChatGPT work like your second brain instead of a generic AI assistant.
The distinction in that line, from AI Central's guide, is the point of the exercise. A generic assistant restarts from zero every time. A configured one carries your context in by default.
Where the settings actually live
The route is short and it is the same one every time. Open ChatGPT, then click your name in the top right corner. The menu that drops down carries Customize ChatGPT and Settings, and both matter here.
AI Central's first instruction is to turn custom instructions on, which is what Customize ChatGPT opens. Nothing here is buried behind an advanced menu. People skip it because the assistant works without it, badly, and a thing that works badly rarely gets fixed.
The four boxes, and what belongs in each
The panel opens with one line of framing, introduce yourself to get better, more personalized responses, then asks four questions. In order:
- What should ChatGPT call you. A nickname, so replies address a person rather than a user.
- What do you do. Your role. The worked example AI Central uses is a third-grade teacher, which is a useful choice, because it changes vocabulary, reading level and examples all at once.
- What traits should ChatGPT have. The tone you want back, either described in your own words or selected from the built-in list.
- Anything else ChatGPT should know about you. The catch-all, for interests, values or preferences you want kept in mind.
The traits question ships nine ready-made options you can add with a click. They are Chatty, Witty, Straight shooting, Encouraging, Gen Z, Skeptical, Traditional, Forward thinking and Poetic.
AI Central's own examples for that box are more useful than the presets, because they are not adjectives at all. The guide suggests be direct, and avoid fancy jargon. Those are instructions, not labels, and instructions are easier for a model to act on than a personality tag. Be direct tells the system what to cut. Witty tells it to try harder, which is usually the opposite of what people want.
On how much to write, the guide is unambiguous, and it repeats the same line twice, once for custom instructions and once for memory.
The more you share, the more ChatGPT tailors answers to your workflow.
AI Central states that as the operating principle behind both settings. The fourth box, the one for interests and values, is where that pays off most, and it is also the box almost everyone leaves blank.
One control at the bottom of the panel decides whether any of this sticks. It is labelled Enable for new chats, and it is why this is a one-time job. Your answers ride into conversations you start later, rather than applying only to the window you happen to have open. Save without it and you have filled in a form for nothing.
Memory is the other half of the setup
Custom instructions are what you tell the assistant about yourself. Memory is what it retains on its own across conversations, and AI Central's guide treats the two as one job in two parts rather than as alternatives.
The path is separate from the customization panel. Open Settings, go to Personalization, find Memory, and toggle it on. Turning custom instructions on does not turn memory on, which is the single most common reason someone thinks they have set this up and has only done half of it.
The two do different work, which is why the guide asks for both. Custom instructions are your constants, the facts that do not change week to week. Memory handles the accumulating detail you would otherwise re-type at the start of every session.
Why the setup pays off more than better prompting
There is a hidden tax in using an unconfigured assistant, and it is paid in every conversation. You explain your role. You explain your audience. You ask it to stop padding. Then you close the tab and the preamble evaporates.
Front-loading that context into settings converts a recurring cost into a fixed one. The five minutes AI Central's sequence asks for are spent once, and every conversation afterwards starts further along. That is a compounding return, and it is available to anyone regardless of how good their prompting is.
The flip side of the more you share deserves saying plainly. Anything written into a standing profile travels with every new chat, so it is worth being deliberate about what goes in there. Role, audience, tone preferences and working constraints earn their place. Anything you would not want surfacing in an unrelated conversation does not.
What to do in the next five minutes
The work splits cleanly by how much you already use the tool. Three starting points, depending on where you are:
- If you have never opened the panel, do the minimum version. Nickname, what you do, and one instruction in the traits box. Save it with Enable for new chats switched on.
- If you filled it in months ago and forgot, audit it. Roles change, and a stale description quietly steers every answer you get. Strip out adjectives and replace them with commands.
- If you use ChatGPT daily, turn Memory on under Settings and Personalization, then go back and finish the fourth box on interests and values. That pairing is where the second brain effect AI Central describes actually shows up.
AI Central's guide is unsentimental about where the advantage comes from.
GPT-5 is a Tool.
The guide follows that line by calling the setup process the engine, and the distinction holds. The model is identical for everyone who opens it. The configuration behind it is not, and that is the part you control.
Do I have to turn custom instructions on for every chat?
No. The panel includes a setting labelled Enable for new chats, so once you save your answers they carry into conversations you start afterwards. You edit them from the same place, the menu under your name in the top right, whenever your role or preferences change.
Where is Memory in the ChatGPT settings?
Under Settings, then Personalization. AI Central's guide gives the path as settings, then Personalization, then find Memory and toggle it on. It is a separate switch from custom instructions, so switching one on leaves the other untouched.
What should I actually write in the traits box?
You can pick from the nine built-in traits, which are Chatty, Witty, Straight shooting, Encouraging, Gen Z, Skeptical, Traditional, Forward thinking and Poetic, or type your own. AI Central's examples are the second kind, be direct, and avoid fancy jargon. Short commands about what to do and what to avoid give the model something concrete to follow.
Is this worth doing if I only use ChatGPT occasionally?
Arguably it matters more. Occasional users are the least likely to have built the habit of pasting context into every prompt, so they get the most generic answers and conclude the tool is overrated. Setting it up once removes that step permanently.