5 ChatGPT Settings You're Not Using

Most people open ChatGPT, type a question, and leave. That works. But there are five settings sitting in your account right now that change how the model remembers you, how it responds, and what it does with your data. Turning them on — or off — takes about three minutes total. The difference in output quality is not subtle.

1. Custom Instructions: Tell It Who You Are Once, Forever

Custom Instructions is the single highest-leverage setting in ChatGPT. You fill out two fields — what you want the model to know about you, and how you want it to respond — and every conversation starts with that context already loaded. You never re-explain your role, your audience, or your preferred format again.

To find it: click your profile photo in the bottom-left corner, go to Customize ChatGPT, then Custom Instructions. The two fields are 'What would you like ChatGPT to know about you?' and 'How would you like ChatGPT to respond?'

Here is an example set of instructions for a coach or service business owner:

ABOUT ME:
I run a coaching business helping B2B consultants build client pipelines. My audience is small firm owners, 5–25 employees. I sell high-ticket services ($3k–$15k engagements). I use GoHighLevel for CRM and email, and I publish content on LinkedIn.

HOW TO RESPOND:
- Skip filler phrases like "Great question!" or "Certainly!"
- Use short paragraphs and plain language
- Default to bullet points for steps, lists, or comparisons
- When I ask for copy, write it in first person as if I'm speaking
- If something is unclear, ask one clarifying question before responding

The model follows these rules in every chat, on every device. Update them any time your business shifts. This alone cuts the back-and-forth in half.

2. Memory: Let It Learn Instead of Starting Over

Memory is different from Custom Instructions. Where Custom Instructions are rules you write manually, Memory is facts ChatGPT stores automatically as you chat. Tell it your business name once and it remembers. Mention your target client profile and it stores that too. Over time, the model builds a working picture of you.

Find it under Settings → Personalization → Memory. Toggle it on. Then, in any conversation, you can say 'remember that' and it will explicitly save that fact. You can also view and delete individual memories from the same settings panel.

What to store on purpose:

• Your business model and offer (e.g., 'I sell a $5k done-with-you LinkedIn content package')
• Your brand voice keywords (e.g., 'my tone is direct, non-corporate, and slightly skeptical of hype')
• Your ideal client description
• Tools and platforms you use regularly
• Recurring tasks you ask ChatGPT to help with

One caution: if you share a paid account across team members, disable memory or check what is stored. It will mix context from different users if you are not careful.

3. Temporary Chat: Go Off the Record When You Need To

Most people do not know this mode exists. Temporary Chat turns off memory storage and removes the conversation from your history. Nothing is saved. Nothing is used to train future models. When you close the window, it is gone.

Access it from the pen icon next to 'New Chat' in the sidebar, or from the three-dot menu on a conversation. A small icon confirms you are in temporary mode.

When to use it:

• Drafting anything with sensitive client information or personal financials
• Testing a competitor analysis you do not want stored
• Brainstorming ideas you are not ready to commit to
• Using ChatGPT on a shared or work-managed device
• Any time the conversation touches NDA-covered material

This is not paranoia. It is good operating hygiene. Knowing when to use ephemeral mode is as important as knowing when to let the model learn from you.

4. GPT-4o with Canvas: Edit Output Without Copying It Elsewhere

Canvas is a writing and coding workspace that opens beside the chat window. Instead of generating text into the chat thread, ChatGPT puts a draft into a persistent document you can edit directly — you click on a line, rewrite it, and the model can revise specific sections without touching the rest.

To use it, open a new chat, make sure you are on GPT-4o, and either ask ChatGPT to 'open a canvas' or ask it to write something long-form. It will often open automatically for drafts, emails, or structured documents. You can also click the Canvas icon in the message toolbar.

Where it actually saves time:

• Writing and iterating on a sales email without rerunning the whole prompt
• Building a content calendar you want to edit in-place
• Refining a proposal section by section without losing the full document
• Asking the model to adjust reading level or tone on just one paragraph
• Keeping a reference document open while you work in the chat on a different task

Canvas is available on GPT-4o only, not on older models. If you do not see it, check your model selector in the top-left of the chat window.

5. Data Controls: Stop Your Chats From Training OpenAI's Models

By default, your conversations can be used to improve future ChatGPT models. Most users never change this. If you are running client work, handling proprietary strategy, or just want cleaner data hygiene, you can turn it off.

Go to Settings → Data Controls → Improve the model for everyone. Toggle it off. Your conversations will no longer be used for training. This applies going forward — it does not delete past conversations, but you can clear those separately under the same menu.

Steps to lock down your data settings properly:

• Settings → Data Controls → toggle off 'Improve the model for everyone'
• Settings → Data Controls → review 'Shared Links' and revoke any you no longer need
• Settings → Data Controls → Export Data if you want a local backup of your conversation history
• Settings → Data Controls → Delete All Chats if you want a clean slate (irreversible)

If you are on ChatGPT Team or Enterprise, your data is already excluded from training by default. But if you are on Plus or Free, this toggle matters and it is off by default.

The Fastest Way to Apply All Five

Do not try to optimize everything at once. Here is the order that gives you the most return for the least time:

• Start with Custom Instructions — write two paragraphs about who you are and how you want responses. Takes five minutes, pays back on every single chat.
• Turn on Memory — let it build context over the next two weeks of normal use. Review what it stored after a few sessions.
• Turn off model training — one toggle, thirty seconds.
• Learn the Temporary Chat shortcut for sensitive work — make it a habit, not an afterthought.
• Try Canvas the next time you are drafting something longer than three paragraphs — you will not go back to the chat thread for that kind of work.

None of this is advanced. It is just using the tool the way it was designed to be used. Most people skip the settings entirely and wonder why they get generic output. These five changes fix that.