You are sitting on a massive dataset that can completely revolutionize how you interact with AI, and it has been hiding in your settings menu the whole time.
The Data-Driven Approach to Personalization
Most of us guess when we write our Custom Instructions. We try to remember what we like, jot down a few bullet points, and hope for the best. But this savvy professional, known as Impressive_Suit4370 on Reddit, devised a way to stop guessing and start using hard data. The author realized that your past conversations contain a perfect blueprint of your working style, your frustrations, and your goals.
Instead of manually trying to recall every time you asked ChatGPT to “be more concise” or “stop apologizing,” you can use the AI to analyze its own history with you. By exporting your chat logs and running a specialized analysis prompt, you can extract the hidden patterns in your communication. This turns a messy pile of old chats into a streamlined set of instructions that make the AI behave exactly how you want it to, automatically.
💡 Why This “Export Mining” Technique Works
1. It Identifies Your “Friction Signals”
One of the smartest parts of this workflow is how it handles negative feedback. The creator designed the prompt to specifically scan for moments where you expressed frustration. It looks for phrases like “too vague,” “stop inventing,” or “be concrete.” These moments are gold because they define your hard constraints. Instead of just knowing what you like, the system learns exactly what makes you tick. This prevents the AI from making the same annoying mistakes over and over again.
2. It Separates Global Rules from Specific Memories
The analysis doesn’t just dump everything into one box. The original poster structured the process to categorize findings into three distinct buckets: Custom Instructions (global rules), Memories (specific facts), and Projects (context-specific goals). This distinction is crucial. Global rules handle how the AI speaks to you (e.g., “always code in Python”), while memories handle what it knows about you (e.g., “user is studying for a biology exam”). This separation keeps your setup clean and prevents conflicting instructions.
3. It Infers Preferences You Didn’t Know You Had
You might think you know your working style, but the data often tells a different story. The expert’s prompt looks for repeated deliverable types and structural preferences. You might discover that you ask for “checklists” 40% of the time, or that you consistently reject answers that are longer than three paragraphs. The analysis highlights these subconscious habits so you can codify them into explicit rules. This moves your personalization from “what I think I want” to “what I actually use.”
🛠️ How to Mine Your Own Data
This process is surprisingly simple, but it requires a specific sequence of steps to work correctly. The author provided a clear roadmap to get your data out of the system and back into it in a useful form.
Step 1: Get Your Conversation History
First, you need the raw data. This takes about two minutes.
* Open ChatGPT and click on your profile icon.
* Navigate to Settings > Data Controls.
* Click Export Data.
* Check your email for the download link from OpenAI.
* Download the .zip file and extract it.
* Locate the file named conversations.json. This file contains your entire chat history.
Step 2: The Mining Process
Now, you will use a fresh ChatGPT instance to analyze that file. You are essentially hiring the AI as a consultant to audit its own performance.
* Start a new chat.
* Upload the conversations.json file you just extracted.
* Paste the “Export Miner” prompt provided below.
Step 3: Review and Apply
The AI will output a structured report. It will give you text blocks ready to copy and paste.
Custom Instructions: Copy the generated text into your Settings under “Personalization.”
Memories: The prompt will suggest specific facts to add to the AI’s memory. You can tell the AI to remember these one by one.
Projects: If you have distinct workspaces (like “Coding” vs. “Creative Writing”), the analysis will suggest separate project setups.
📝 The “Export Miner” Prompt
This is the engine that makes the whole strategy work. The original creator built this prompt with strict constraints to ensure it doesn’t hallucinate or leak sensitive info. Copy this and paste it along with your file:
You are a “Personalization Helper (Export Miner)”.
Mission: Mine ONLY the user’s chat export to discover NEW high-ROI personalization items, and then tell the user exactly what to paste into Settings → Personalization.
Hard constraints (no exceptions):
– Use ONLY what is supported by the export. If not supported: write “unknown”.
– IGNORE any existing saved Memory / existing Custom Instructions / anything you already “know” about the user. Assume Personalization is currently blank.
– Do NOT merely restate existing memories. Your job is to INFER candidates from the export.
– For every suggested Memory item, you MUST provide evidence from the export (date + short snippet) and why it’s stable + useful.
– Do NOT include sensitive personal data in Memory (health, diagnoses, politics, religion, sexuality, precise location, etc.). If found, mark as “DO NOT STORE”.Input:
– I will provide: conversations.json. If chunked, proceed anyway.Process (must follow this order):
Phase 0: Quick audit (max 8 lines)
1) What format you received + time span covered + approx volume.
2) What you cannot see / limitations (missing parts, chunk boundaries, etc.).Phase 1: Pattern mining (no output fluff)
Scan the export and extract:
A) Repeated user preferences about answer style (structure, length, tone).
B) Repeated process preferences (ask clarifying questions vs act, checklists, sanity checks, “don’t invent”, etc.).
C) Repeated deliverable types (plans, code, checklists, drafts, etc.).
D) Repeated friction signals (user says “too vague”, “not that”, “be concrete”, “stop inventing”, etc.).
For each pattern, provide: frequency estimate (low/med/high) + 1–2 evidence snippets.Phase 2: Convert to Personalization (copy-paste)
Output MUST be in this order:1) CUSTOM INSTRUCTIONS: Field 1 (“What should ChatGPT know about me?”): <= 700 characters.
– Only stable, non-sensitive context: main recurring domains + general goals.2) CUSTOM INSTRUCTIONS: Field 2 (“How should ChatGPT respond?”): <= 1200 characters.
– Include adaptive triggers:
– If request is simple → answer directly.
– If ambiguous/large → ask for 3 missing details OR propose a 5-line spec.
– If high-stakes → add 3 sanity checks.
– Include the user’s top repeated style/process rules found in the export.3) MEMORY: 5–8 “Remember this: …” lines
– These must be NEWLY INFERRED from the export (not restating prior memory).
– For each: (a) memory_text, (b) why it helps, (c) evidence (date + snippet), (d) confidence (low/med/high).
– If you cannot justify 5–8, output fewer and explain what’s missing.4) OPTIONAL PROJECTS (only if clearly separated domains exist):
– Up to 3 project names + a 5-line README each:
Objective / Typical deliverables / 2 constraints / Definition of done / Data available.5) Setup steps in 6 bullets (exact clicks + where to paste).
– End with a 3-prompt “validation test” (simple/ambiguous/high-stakes) based on the user’s patterns.Important: If the export chunk is too small to infer reliably, say “unknown” and specify exactly what additional chunk (time range or number of messages) would unlock it, but still produce the best provisional instructions.
🚀 Why This Matters for You
This method moves you from a passive user to an active architect of your tools. By using the author’s prompt, you ensure that every future interaction with ChatGPT is built on the foundation of your past successes and failures. It saves you from having to repeat yourself and ensures the AI grows with you!
Check out the full breakdown by Impressive_Suit4370 in the link below.
💡 FAQ & Troubleshooting
Why can’t I upload my conversations.json file? (Too large / Too many tokens)
The conversations.json file contains your entire chat history. For active users, the file size often exceeds the “context window” (token limit) that ChatGPT can process in a single message. To fix this, you cannot upload the raw file directly if it is large. Instead, open the file in a text editor, copy a subset of the text (e.g., the most recent 6 months or specific relevant conversations), and paste that into the chat or save it as a smaller text file to upload.
Where do I find the export option in ChatGPT?
Navigate to your Settings menu, select Data Controls, and click Export Data. You will receive an email with a download link. Once downloaded, you must unzip the folder to locate the specific conversations.json file required for this workflow.
Will this save sensitive private information into my Custom Instructions?
The prompt provided includes a hard constraint to ignore sensitive data such as health diagnoses, politics, or precise location. It is designed to flag such data as “DO NOT STORE.” However, as a best practice, you should always review the generated instructions and memory items before saving them to your profile to ensure no unwanted private details are included.
How do I apply the generated Memory items?
Unlike Custom Instructions, there is no settings menu to bulk-paste Memories. The output will provide specific “Remember this” lines. You must send these lines to ChatGPT in a fresh chat one by one (e.g., “Remember that I prefer code responses in Python”), which forces the system to commit them to its long-term memory storage.
Your ChatGPT export is a goldmine for personalization
byu/Impressive_Suit4370 in