Optimizing ChatGPT: 5 Core Configurations to Elevate Processing Capabilities and Response Quality

Most ChatGPT users start typing queries immediately without ever tweaking the default settings. However, simply toggling 5 core built-in options will completely transform the quality of its responses: forcing the AI to think deeply before answering, eliminating robotic clichés, and connecting it directly to your actual workflow.

This article provides a step-by-step guide to these 5 essential settings, requiring just a 10-minute one-time setup for long-term efficiency.

Guide to configuring 5 essential settings for ChatGPT — Guide to 5 settings for optimizing performance and achieving a professional tone in ChatGPT

1. Activate Thinking Mode

The Thinking mode forces the model to trigger an internal chain-of-thought to consider multiple perspectives and cross-check facts before providing a final answer. This virtually eliminates silly logical reasoning errors.

Open chatgpt.com in your browser or launch the ChatGPT Desktop app.
Look at the top corner of the chat box to check the current model being used (e.g., ChatGPT 4o).
Click the model selection menu to switch to ChatGPT 5 Thinking (or the o3-mini / o1 models if you are using a Plus/Pro account).

WHEN SHOULD YOU USE THINKING MODE?

Use it for complex logical analysis tasks, architectural coding, business strategic planning, or reviewing documents. For quick everyday lookups, you should opt for standard models to get instant results.

2. Setting Up an Anti-Hallucination Skeptical Persona (Cynical Mode)

By default, large language models are fine-tuned to be overly accommodating to users (sycophancy) – they tend to nod along with whatever opinion you present and make up information (hallucination) just to satisfy your prompt. Activating a critical thinking persona forces the AI to stick to the facts.

Click on your profile avatar in the bottom-left corner of the ChatGPT screen.
Select Settings from the menu that appears.
Select the Personalization tab in the left-hand sidebar.
Find the Response Style section and select the Cynical style (or configure it via the custom prompt below).

Critical Persona Prompt

When answering my questions, act as a sharp critical consultant:
1. Always maintain healthy skepticism and test the feasibility of any assumptions I make.
2. Frankly point out risks, logical fallacies, or hidden costs instead of agreeing blindly.
3. If the data is insufficient or you are unsure, clearly state "I am not sure" along with the reasons.

THE ESSENCE OF CRITICAL THINKING

This mode isn’t about being negative or argumentative; it’s about forcing the AI to be completely honest: daring to call out weaknesses in your plans and refusing to fabricate facts without verifiable grounds.

3. Setting Up Human Mode in Custom Instructions

By default, ChatGPT suffers from “robotic prose syndrome”: overusing bullet points uncontrollably, inserting clichés and emojis, and always opening with hollow pleasantries (“Great question!”, “Certainly!”). Paste this standardized instruction block into your profile:

Go to Settings, select the Personalization tab, and click on Custom Instructions.
In the “How would you like ChatGPT to respond?” field, paste the entire text below.
Click the Save button to apply these instructions to all future conversations.

Prompt Human Mode

You are a knowledgeable assistant, but communicate like a thinking human—not a report-generating robot.

TONE & STYLE:
- Talk like an intelligent friend, not a customer service bot. Be direct, warm, and natural.
- Write short paragraphs. Write coherent sentences; avoid bullet points or numbered lists unless explicitly requested by the user.
- Never use dashes (-, -, -) to structure responses.
- Only use emojis if the user uses them first. Never decorate responses with emojis just to appear friendly.
- Vary sentence length. Combine short, punchy sentences with longer ones. This is how humans talk.

ACCURACY:
- If you don't know something, say so directly. Don't guess and present it as fact.
- Do not make up names, numbers, dates, or references. If unsure, say "I'm not sure" or "you should double-check this."
- Never fabricate quotes or sources.

CONCISENESS:
- Answer the exact question asked. Don't pad responses with unnecessary context, background info, or rehashes of what was just said.
- If the answer is short, keep it short. Don't turn a two-sentence answer into five paragraphs.
- Drop clichés like "Great question!", "Certainly!", "Of course!", "As an AI language model..."—get straight to the point.

STRUCTURE:
- Avoid excessive headings and subheadings. Only use them when the response is genuinely long and complex.
- Do not bold random words just for visual effect. Only bold what is truly important.
- Don't end every response with a concluding summary paragraph of what was just stated.

CONVERSATION:
- Ask only one follow-up question at a time, and only if it genuinely helps you provide a better answer.
- Do not over-explain. Trust that the user is intelligent enough.

PRACTICAL RESULT

Once saved, you will see ChatGPT’s style transform instantly: concise, natural phrasing that cuts straight to the core of the matter, completely stripping away the feel of a report-generating bot.

4. Enabling and Managing Long-Term Memory

If you don’t enable Memory, ChatGPT will completely “lose its memory” every time you open a new chat session, forcing you to re-explain your work context from scratch. This feature helps the AI accumulate knowledge day by day.

Go to Settings and select the Personalization tab.
Find the Memory section and turn on both toggle switches to activate the memory feature.
Click on “Manage memories” to review the list of saved notes and delete any outdated information.

Memory Setup Prompt

Please remember the following core information in my profile:
1. Occupation: Software Engineer / Content Creator.
2. Core Technologies: Next.js, TypeScript, Tailwind CSS, WordPress.
3. Work Style: Prefer minimalist solutions, no bloatware/unnecessary libraries, prioritize security.
4. Response Language: Standard Vietnamese, keep technical terms in English.

MEMORY MANAGEMENT TIP

From time to time, proactively prompt: ‘Remember that Project A has ended, and I have now switched to Project B’. ChatGPT will automatically update its knowledge base without mixing up data.

5. Configuring Connected Apps

The Connected Apps (or Connectors) feature transforms ChatGPT from a passive Q&A chatbot into an autonomous Agent capable of directly accessing your Google Workspace ecosystem (Gmail, Calendar, Drive, Docs).

Go to Settings and select the Connected Apps (or Connectors) section.
Click the Connect button next to the application you need to integrate (such as Google Drive, Gmail, GitHub) and complete the OAuth authorization.
Once successfully connected, enable Agent Mode or Deep Research to prompt the AI to process real-world data.

Sample Agent Mode Prompt

Please check my schedule on Google Calendar for tomorrow and cross-reference it with unread emails in Gmail to compile a list of the 3 most important tasks to tackle in the morning.

DATA SECURITY NOTE

When connecting applications, ChatGPT will have read access to data from the linked account. Never connect mailboxes or cloud storage containing strictly confidential documents, personally identifiable information (PII), or banking and credit card details.

6. Summary of 5 Optimal Settings for ChatGPT

⇄ Swipe horizontally to view full table

Setting	Location	Key Benefits
1. Thinking Mode	Model selection menu at the top	Deep reasoning, eliminates logical reasoning errors
2. Cynical Mode	Settings > Personalization	Honest critical feedback, reduces hallucinations
3. Human Mode	Custom Instructions	Eliminates robotic tone, prevents excessive bullet points
4. Memory	Personalization > Memory	Remembers work context across all chat sessions
5. Connected Apps	Settings > Connected Apps	Reads Gmail/Drive data, automates tasks

7. Common Errors and Troubleshooting

Summary of solutions for common technical issues when configuring your ChatGPT account:

I do not see the Thinking model option in the list?
▼

Cause: Advanced reasoning models (such as o3-mini, o1, or ChatGPT 5 Thinking) are typically rolled out first by OpenAI to paid account tiers (Plus, Team, Pro). If you are using a free account, you may receive a limited quota of just a few turns per day.

Do Custom Instructions apply to the mobile app?
▼

Yes, they sync automatically. As long as you save them on the web interface, the ChatGPT app on iOS and Android logged into the same account will automatically apply these rules.

Why does ChatGPT still occasionally use bullet points after setting Human Mode?
▼

If your prompt explicitly asks to ‘List 5 points…’ or involves multi-column comparisons, the AI will still use lists to maintain visual clarity. The rule only prevents the overuse of bullet points in standard explanatory answers.

Connected Google Drive, but ChatGPT shows an error saying it cannot read the file?
▼

Check whether the file in Drive is saved in a standard format (PDF, Google Docs, Sheets, TXT). If the file has restricted access permissions or is too large in size, you should download the file to your computer and attach it directly to the chat input.

Can I temporarily disable these settings when needed?
▼

Absolutely. Inside Custom Instructions, there is an ‘Enable for new chats’ toggle that you can switch off at any time to return to OpenAI’s default style.

8. Frequently Asked Questions (FAQ)

Answers to questions regarding features and privacy when performing advanced configuration for ChatGPT:

Does enabling Human Mode reduce ChatGPT’s coding capabilities?
▼

Not at all. The rules in Human Mode only adjust tone and external communication style; the model’s logical reasoning and coding capabilities remain 100% intact.

Can I connect multiple Google Drive accounts to a single ChatGPT account?
▼

Currently, OpenAI only allows linking 1 primary Google Workspace account per ChatGPT account at a time.

How can I clear all Memory without losing my chat history?
▼

Go to Settings, select Personalization, click Manage memories, and choose Clear all memories. This action only deletes remembered data; your entire chat history remains preserved.

Does Thinking Mode consume more message quota?
▼

Deep-reasoning models consume significantly more compute resources, meaning message limits per time window (rate limits) are stricter compared to standard models like GPT-4o.