The ChatGPT Setting Most People Never Change That Makes Answers Way Better

I spent three months getting mediocre ChatGPT responses before I accidentally discovered I'd been ignoring the one setting that actually matters. Not temperature. Not the model selector. The custom instructions panel that's been sitting there the whole time, quietly waiting for me to stop being lazy.

Where This Setting Actually Lives

Click your profile icon. Then "Customize ChatGPT." Two text boxes appear. The first asks what you want ChatGPT to know about you. The second asks how you want it to respond. Most people glance at this, think "I'll fill that out later," and never come back.

I know because I did exactly that. For months. Every single conversation started from zero. I'd re-explain that I'm a writer, that I hate corporate speak, that I need practical examples not theory. Over and over. It was exhausting and I didn't even realize it was exhausting until I stopped doing it.

The setting persists across every new conversation. Fill it out once, and ChatGPT remembers. Not in a creepy way — it's just prepending your instructions to every prompt you send. Simple. But the difference in output quality is genuinely startling.

What I Put in Mine (And What Failed)

My first attempt was too vague. I wrote something like "I'm a writer who tests AI tools. Give me practical advice." The responses improved maybe 10%. Still generic. Still that weird ChatGPT voice where everything sounds like a helpful assistant trying too hard.

Then I got specific. Way more specific than felt reasonable. I listed the exact phrases I never want to see — "get into," "use," "Worth mentioning." I specified my actual writing style: short sentences, casual tone, first person. I mentioned I'm testing tools for real projects, not just curiosity.

The difference was immediate. Responses got shorter. Less padding. Fewer bullet points when I didn't need them. It started matching my voice instead of its default helpful-robot mode.

But here's what didn't work: putting too much in there. I initially wrote about 800 words of instructions. ChatGPT started getting confused, sometimes contradicting itself or ignoring parts of what I'd written. There's a sweet spot around 300-400 words where it actually retains everything.

The Kick: The Second Box Is Where the Magic Happens

Everyone focuses on the first box — "what should ChatGPT know about you." But the second box, "how would you like ChatGPT to respond," is where I found the real use.

Here's the specific line that changed everything for me: "Never summarize what you just said. Never end with a list of key takeaways. End mid-thought if you have to."

That one instruction eliminated the single most annoying ChatGPT habit. You know the one. You ask a simple question, get a decent answer, then it adds three paragraphs restating everything it just told you. Like a student padding an essay to hit the word count.

I also added: "If you don't know something, say you don't know. Don't hedge with phrases like 'it depends' unless you then explain what it depends on."

And this one, which sounds harsh but works: "Skip the enthusiasm. Don't tell me something is a 'great question' or that I'm going to 'love' the answer."

The responses got noticeably more direct. Less performative helpfulness. More actual help. It's like the difference between talking to someone who's trying to impress you versus someone who just wants to solve the problem.

The Limitation Nobody Mentions

Custom instructions don't override everything. When I asked ChatGPT to write code, it still defaulted to heavy commenting and verbose explanations even though my instructions said to keep things minimal. Same with certain creative writing prompts — it would slip back into default patterns.

The workaround is putting critical instructions in both places: your custom settings AND your actual prompt. Redundant, yes. But it works. For my most important preferences, I have them saved in a text file I paste into complex prompts. Belt and suspenders.

Also, these settings don't transfer to the API. If you're using ChatGPT through another app or integration, you're starting from scratch. This is only for the direct web and mobile interfaces.

One more thing I learned the hard way: custom instructions can make ChatGPT worse for certain tasks. I had "always be concise" in my settings, and when I needed detailed technical explanations, it kept cutting itself short. I had to temporarily edit my instructions. There's no perfect set-it-and-forget-it configuration.

I've been tweaking my custom instructions for about four months now. They're on version 12 or 13 — I lost count. Each version gets a little better as I notice new annoying patterns. The latest addition was asking it to never use the phrase "here's the thing" because apparently that's my personal pet peeve and I didn't realize it until ChatGPT kept triggering it.

What I still can't figure out: whether there's a way to have different custom instruction profiles for different types of work. Sometimes I want creative and exploratory. Sometimes I want robotic and precise. Right now I have to manually switch, which defeats half the purpose. If anyone's found a workaround for this, I genuinely want to know.

Heads up: Some links in this post may be affiliate links. I only recommend tools I've personally tested. Opinions are entirely my own.