Do you use ChatGPT every day?
And when was the last time you opened the settings screen?
Most people probably haven't touched it once since they first started using it.
You might think it's enough to just treat it as a clever chat partner—ask a question on the spot, get an answer on the spot.
But this is where the big divide happens.
Even when using the same ChatGPT, the quality of the answers is completely different between those who have tuned their settings and those who use the default configuration.
Moreover, that difference isn't usually caused by paid plans or prompt engineering skills; it's mostly found right inside the settings screen.
Today, I will walk you through seven settings to organize, starting from the top.
Why 90% are using it at half performance
The reason is simple.
If you don't touch the settings, ChatGPT treats every interaction as a "first meeting."
It doesn't know what your job is, what writing style you prefer, or how much background knowledge it can assume you have. If it remembers nothing, you have to explain everything from scratch every single time.
"I'm a freelancer doing this kind of work..."
"Don't use bullet points, use full sentences..."
"Start with the conclusion..."
Are you re-typing these things for every new conversation?
Even if it only takes a few minutes each time, it adds up significantly over days and weeks. Furthermore, if you skip the explanation, the quality of the response fluctuates from day to day.
This is the true identity of the feeling that "ChatGPT is useful, but it doesn't save as much time as I thought."
With default settings, you're always back at square one. Conversely, once you organize these settings, that back-and-forth disappears entirely.
7 Hidden Settings: Turn them on in order
Here is the main topic. I've listed them starting from the easiest. The effects get larger as you go down, so the best part is at the end.
1. Turn on Memory
First is "Memory."
This is a feature where ChatGPT automatically remembers information it deems important during a conversation. In 2026, a new memory system was introduced, making it smarter and better at keeping things up to date.
By turning this on, it gradually learns your work and preferences, reducing the number of times you have to re-explain yourself. Start here.
2. Review Memory and correct mistakes
This is surprisingly unknown.
You can actually see the list of what ChatGPT currently "remembers" (the memory summary). If there's a mistake, you can fix it on the spot.
You can type "Delete this information" in the input field, or select text to fix it partially. If left alone, it might carry old assumptions or misunderstandings forever, so cleaning it occasionally improves accuracy. Memory only truly works when you don't just let it learn, but also correct its mistakes.
3. Write "About Me" and "Response Preferences" in Custom Instructions
This offers the best ROI.
In the Custom Instructions field, if you write:
- Who you are (position, job, target audience)
- How you want it to answer (tone, length, format)
...it will automatically apply to all conversations. You won't have to type instructions like "Start with the conclusion," "Don't use bullet points," or "Use a polite tone" every time.
I have my own writing style rules written here, so my instructions for each session are now usually just one sentence. While Memory is for things it learns automatically, Custom Instructions are for settings you fix yourself. They serve different roles.
4. Use the Model Picker based on your purpose
ChatGPT can now switch its way of thinking depending on the use case. On June 12, 2026, the previous generation (GPT-5.2 series) was retired, and everything was consolidated into GPT-5.5.
The model picker is organized roughly like this:
- Instant: For light questions or small talk. Extremely fast.
- Medium: When you want it to think normally.
- High: For difficult tasks requiring deep thought.
- Above that, there are frames like Pro for even heavier reasoning.
Use Instant when you want speed, and High when you want deep thinking. Just switching this changes the depth of the answer. Many people leave this fixed on one setting.
5. Turn on "Auto-Smart when needed"
This is a subtle but effective one. Even if you are talking to Instant, there is a setting that automatically switches to Medium (deep thinking mode) if a difficult question comes up.
You can toggle this in the General settings. When on, you don't have to manually switch models; it thinks deeply only when necessary. It's the best of both worlds: speed and intelligence.
6. Use Projects to separate boxes by case or theme
This is a step toward advanced usage. Using the Project feature, you can create "boxes" for each case or theme.
You can bundle Custom Instructions and reference files that apply only within that box. For example, if you separate things like "Company A Project," "X (Twitter) Posts," and "Accounting," ChatGPT enters that specific context the moment you open the box. Explaining the situation from scratch almost disappears here. It's also great for preventing conversations from getting cluttered.
7. Connect Integrations (Apps)
Finally, integration with external services. What used to be called "Connectors" are now consolidated under "Apps."
By connecting your calendar, files, or business tools, ChatGPT can answer while looking at that information. You'll be able to say "Based on my schedule" or "Based on that document" without any copy-pasting. At this level, it changes from a mere chat partner to a partner who understands the context of your work.
How daily use changes after setting up
What changes after you organize all this?
The biggest thing is that your first sentence in every conversation becomes shorter. Before, you had to provide a preamble: "I'm a freelancer, my style is X, start with the conclusion..." Now, you can jump straight to the main topic because the other side already knows the context.
Also, the variance in answers decreases. Since the tone, format, and premises are written in the settings, you'll get roughly the same quality whether you ask on a Monday or a Friday. It's a subtle but powerful effect.
It feels less like "using" ChatGPT and more like it's "operating to your specifications."
One more thing: once settings are organized, results remain consistent even if someone else uses it. When you want to share the same premises with a team, writing them in Custom Instructions or Projects is faster and more accurate than explaining them verbally.
One last tip for those who read this far: don't mix up the roles of Memory and Custom Instructions.
Memory is for information that grows automatically. Custom Instructions are for settings you fix yourself. Therefore, write the premises you absolutely want it to follow (style, position, rules) manually in Custom Instructions. Leave fluid, daily changing information to Memory. By separating them this way, you avoid the "Why is it answering based on a weird assumption?" problem that people often run into when they just turn both on blindly.
Summary: One 10-minute session changes every day
To be honest, you don't need to do all seven today. Start with just two: Memory and Custom Instructions.
Turn them on and write down your job and response preferences. It takes about 10 minutes. That alone will drastically reduce the need to re-explain yourself. Performance differences are often determined more by "whether you configured it" than by the intelligence of the model itself.
Before upgrading to an expensive plan, try organizing the settings you can access for free. The perceived difference might actually be larger that way.
I haven't perfectly mastered everything either; sometimes I neglect cleaning my memory and it carries over weird assumptions. But from the day I set up those first two, my relationship with ChatGPT changed significantly.
