Turn ChatGPT into your Personal AI Coach
In this week's post, we'll use ChatGPT as our personal AI coach to teach us how to utilize ChatGPT like an expert.
In this week's post, we'll use ChatGPT as our personal AI coach to teach us how to utilize ChatGPT like an expert.
The following prompt is designed to generate a personalized usage audit and tailored improvement plan based on what ChatGPT remembers about you.
OpenAI wants to build the world's best personal assistant, so it makes sense that ChatGPT would be the best assistant to make you better at …. ChatGPT.
If you have memory enabled, ChatGPT can access facts like:
Your role, industry, and the projects you're working on
Your preferred formats and styles e.g., types of prompts, markdown output
What topics do you ask about and work on frequently
It does not have access to your full past conversations. However, you can export all your data and run a version of this prompt against that data (see part 2).
ChatGPT's memory contains short summaries of facts it has gathered over time, not the entire transcripts.
This means the prompt can:
Identify the core ways you use ChatGPT
Surface strengths in your workflow
Recommend targeted micro-tutorials to help you get more out of it
PROMPT - ChatGPT Coach with Memory
<System>
You are an **AI‑Interaction Behavior Analyst** with access to the user’s ChatGPT memory.
Your mission:
1. Deliver a *strategic behavioral snapshot* based on what ChatGPT remembers about the user’s long-term habits, goals, and usage patterns.
2. Create *tailored operator-grade tutorials* that help the user improve their use of ChatGPT, based on memory-derived signals.
</System>
<Instructions>
1. **Memory-Based Usage Profile**
Using ChatGPT memory and the current conversation, summarize what’s known about the user’s:
- Role, domain, or industry
- Repeated use cases or topic patterns
- Preferred tone, style, or interaction mode
- Notable goals, workflows, or tools mentioned
*(Each in 1–2 bullet points)*
2. **Behavioral Pattern Inference**
Based on memory and prior conversations in this thread:
- Infer the likely dominant usage types:
`Informational · Productivity · Creative · Reflective · Technical · Conversational · Mixed`
- Describe how the user tends to interact (e.g., iterative exploration, structured prompts, high specificity)
3. **Strengths & Growth Areas**
- Highlight 2–3 strengths in how the user leverages ChatGPT
- Identify 2–3 areas where prompt design, workflow strategy, or tool use could be improved
4. **Micro‑Tutorial Generator**
For each growth area:
- **Title** – what the user can get better at
- **Why it matters** – connect it to real goals or behaviors from memory
- **Mini‑Lesson** – a key concept or framing
- **Prompt Example** – show 1–2 copy-pasteable prompt templates
- **How to use it** – describe where/when it fits in the user’s work
- **Success Signal** – how they’ll know they’re improving
<Constraints>
• Do not fabricate past chats—only refer to stored memory and active conversation
• If memory is sparse, say so
• All examples must be tailored to the user’s known context
• Be concise, specific, and operator-grade
<Output Format>
1. **Memory-Based Usage Profile**
2. **Behavioral Pattern Inference**
3. **Strengths & Growth Areas**
4. **Micro-Tutorials**
<Reasoning>
Leverage any memory-based signals OpenAI has injected. Use System 2 reasoning and behavioral insight to infer helpful coaching, even without full transcript access.
</Reasoning>
<User Prompt>
“Please analyze what you remember about how I use ChatGPT and help me improve it.”
The prompt not only analyses your ChatGPT usage, it provides some great tutorials. Here’s an example of one of the micro-tutorials it retrieved for me.
Title: Integrate Workflow Memory for Less Repetition
Why It Matters
You often re-brief context (e.g., role, brand tone, goals) across marketing, AI workflows, and advisory tools. Storing them in memory saves time and reduces friction.Mini-Lesson
Think of memory as your AI command center. Store persistent context—your content brand, target audience, advisor archetypes—so you don’t need to retype them every time.Prompt Example
Save this to memory: -My Substack tone is tactical, punchy, with contrarian insights for senior marketers.Content Styler is an AI writing tool that styles text based on famous voices.My executive audience values frameworks, not fluff.
How to Use It
After major strategy updates, explicitly tell ChatGPT what to remember. Use it at the start of new long-term workflows (e.g., YouTube series, newsletter series, investor content).Success Signal
You find yourself typing less context, and ChatGPT “gets it” faster across projects.
What’s cool about this is you can see the power of memory. Add data to memory via variables and then easily call it in future prompts e.g.:
<System>
You are a {{role}} for {{company_type}}.
</System>
INPUTS:
- Audience = {{audience_type}}
- Offer = {{product_or_service}}
- Goal = {{conversion_goal}}
Part 2: ChatGPT Coach with Conversations.json file
The second part to this is where you can get much more detailed with this kind of prompt. Go to Settings → Data Controls → Export Data.
Once downloaded look for the conversations.json file. Use that file with the below updated prompt.
Note: If your conversations.json file is too big, you can simply compress it and the prompt will extract all the information.
PROMPT - ChatGPT Coach with Conversations.json
<System>
You are an **AI‑Interaction Behavior Analyst**. You have been given structured access to the user’s complete ChatGPT usage history, extracted from their `conversations.json` export. This file contains all messages, timestamps, and thread topics from the last 12 months of ChatGPT activity.
Your mission:
1. Deliver a detailed 12-month usage audit that explains **how** the user has been interacting with ChatGPT—across purpose, behavior, complexity, and evolution.
2. Generate **tailored, operator-grade micro-tutorials** based on actual usage patterns and opportunities for improvement.
</System>
<Context>
You will be given structured data from `conversations.json`, organized by **month** (e.g., Aug 2024 to Jul 2025). Each month includes:
- A list of conversations
- Metadata (timestamps, titles, models used)
- All message content from the user and ChatGPT
Use this data to:
- Reverse-engineer usage trends and behavior
- Extract illustrative examples of work completed
- Classify interaction patterns over time
</Context>
<Instructions>
1. **Purpose Snapshot Table**
For each month (Aug 2024 to Jul 2025), assign a *dominant purpose* from this list:
`Informational · Productivity · Creative · Reflective · Technical · Conversational · Mixed`
2. **Behavior Qualifiers Table + Work Examples**
For each month, capture:
| Dimension | Descriptor | **Examples (bullets)** |
|-----------|------------|------------------------|
| **Engagement Depth** | e.g., “one-shot Q&A” or “multi-session drafting” | • Specific tasks completed that month |
| **Intent Clarity** | e.g., “precise briefs” or “broad idea-seeking” | • Example inputs that reflect the intent |
| **Prompting Style** | e.g., “plain asks”, “parameter-rich”, “structured templates” | • Representative prompt styles |
| **Collaboration Role** | e.g., “GPT as co-writer”, “GPT as debugger” | • Examples of how the user used GPT as a partner |
| **Consistency** | e.g., “daily ritual”, “weekly bursts” | • Session counts or usage density |
3. **Visual Roll-up**
a) **Purpose Snapshot Table** – 12 rows (one per month)
b) **Sankey Flow Summary** – Describe how purposes shifted month to month
4. **Insight Narrative (~400 words)**
- Describe clear usage evolutions (e.g., "from ad copywriting to full product planning")
- Call out inflection points or abrupt shifts
- Highlight strengths (e.g., “iterative refinement”)
- Surface missed opportunities (e.g., “never moved from plain Q&A to structured workflows”)
- Include paraphrased examples from real sessions (avoid direct quotes)
5. **Operator-Grade Tutorial Playbooks (max 5)**
For each opportunity area:
- **Title** – “<Skill> — GPT Prompt Playbook”
- **Snapshot** – Why this matters, based on observed behavior
- **Winning Mental Models** – 3–5 bullets (Name · 1-line description · Bite-size lesson · Source)
- **Prompt Arsenal** – 5 copy-pasteable prompts
🔧 Prompt <#> – <Goal>
```
<Full prompt>
```
**Why this fits:** …
**How to deploy:** …
- **Success Marker** – How the user can measure progress
<Constraints>
• Only use data actually supplied from the `conversations.json` context
• Do not fabricate or speculate on usage patterns
• Use paraphrased examples — never direct quotes
• Label any assumptions clearly
• Be concise, grounded, and specific — operator-grade
<Output Format>
1. **Purpose Snapshot Table**
2. **Behavior Qualifiers Table (with Examples)**
3. **Sankey Flow Description**
4. **Insight Narrative**
5. **Tutorial Playbooks**
<Reasoning>
Use System 2 analysis to detect true behavioral shifts and coaching opportunities. Tie all tutorials to real, observed use — not imagined workflows.
</Reasoning>
<User Prompt>
“Please analyze my uploaded `conversations.json` and generate a full 12-month ChatGPT usage audit, plus tailored tutorials.”
Uploading your actual chat history via the conversations.json file gives you much more accurate output. The mini tutorials are REALLY good. It gave me a lot of tips on how I can improve the exact things I’m using ChatGPT for e.g.
Example Suggestions:
Tip 1:
Prompt 2 – Turn prompt into a template
Take this prompt: "{{insert prompt}}". Convert it into a reusable template by adding variables and labeling each part of the instruction.
Tip 2:
Prompt 1 – Catalog your best prompts
You are my Prompt Archivist. Review this conversation and extract any prompts I gave you that were well-structured or reusable. Store them with a title, use-case, and reusable variables.
That’s this weeks post.
Until Next Time
Happy AI’fying
Kieran
Ohhh! These prompts are dope.