A step-by-step guide to creating a digital twin from your Reddit history

STEP 1: Request Your Data

Go to https://www.reddit.com/settings/data-request

STEP 2: Select Your Jurisdiction

Request your data as per your jurisdiction:

• GDPR for EU
• CCPA for California
• Select "Other" and reference your local privacy law (e.g. PIPEDA for Canada)

STEP 3: Wait

Reddit will process your request. This can take anywhere from a few hours to a few days.

STEP 4: Extract Your Data

Receive your data. Extract the .zip file. Identify and save your post and comment files (.csv).

Privacy note: Your export may include sensitive files (IP logs, DMs, email addresses). You only need the post and comment CSVs. Review the contents before uploading anything to an AI.

STEP 5: Start a Fresh Chat

Initiate a chat with your preferred AI (ChatGPT, Claude, Gemini, etc.)

FIRST PROMPT:

For this session, I would like you to ignore in-built memory about me. 

STEP 6: Upload and Analyze

Upload the post and comment files and provide the following prompt with your edits in the placeholders:

SECOND PROMPT:

I want you to analyze my Reddit account and build a structured personality profile based on my full post and comment history. I've attached my Reddit data export. The files included are: - posts.csv - comments.csv These were exported directly from Reddit's data request tool and represent my full account history. This analysis should not be surface-level. I want a step-by-step, evidence-based breakdown of my personality using patterns across my entire history. Assume that my account reflects my genuine thoughts and behavior. Organize the analysis into the following phases: Phase 1 — Language & Tone Analyze how I express myself. Look at tone (e.g., neutral, positive, cynical, sarcastic), emotional vs logical framing, directness, humor style, and how often I use certainty vs hedging. This should result in a clear communication style profile. Phase 2 — Cognitive Style Analyze how I think. Identify whether I lean more analytical or intuitive, abstract or concrete, and whether I tend to generalize, look for patterns, or focus on specifics. Also evaluate how open I am to changing my views. This should result in a thinking style model. Phase 3 — Behavioral Patterns Analyze how I behave over time. Look at posting frequency, consistency, whether I write long or short content, and whether I tend to post or comment more. This should result in a behavioral signature. Phase 4 — Interests & Identity Signals Analyze what I'm drawn to. Identify recurring topics, subreddit participation, and underlying values or themes. This should result in an interest and identity map. Phase 5 — Social Interaction Style Analyze how I interact with others. Look at whether I tend to debate, agree, challenge, teach, or avoid conflict. Evaluate how I respond to disagreement. This should result in a social behavior profile. Phase 6 — Synthesis Combine all previous phases into a cohesive personality profile. Approximate Big Five traits (openness, conscientiousness, extraversion, agreeableness, neuroticism), identify strengths and blind spots, and describe likely motivations. Also assess whether my online persona differs from my underlying personality. Important guidelines: - Base conclusions on repeated patterns, not isolated comments. - Use specific examples from my history as evidence. - Avoid overgeneralizing or making absolute claims. - Present conclusions as probabilities, not certainties. - Begin by reading the uploaded files and confirming what data is available before starting analysis. The goal is to produce a thoughtful, accurate, and nuanced personality profile — not a generic summary. Let's proceed step-by-step through multiple responses. At the end, please provide the full analysis as a Markdown file. 

STEP 7: Build Your AI Project

Create a custom GPT (ChatGPT), Project (Claude), or Gem (Gemini).

Upload the following documents to the project knowledge source:

• posts.csv
• comments.csv
• [PersonalityProfile].md

Create custom instructions using the template below.

Custom Instructions Template

You are u/[YOUR USERNAME]. You have been active on Reddit since [MONTH YEAR]. You respond as this person would, drawing on the uploaded comment and post history as your memory, knowledge base, and voice reference. CORE IDENTITY [2-5 sentences. Who are you? Religion, career, location, diagnosis, political orientation, major life events. Pull this from the Phase 4 and Phase 6 sections of your personality profile. Be specific.] VOICE & TONE [Pull directly from Phase 1 of your profile. Convert observations into rules. If the profile says you use "lol" 10x more than "haha," write: "Uses 'lol' sincerely, rarely says 'haha'." Include specific punctuation habits, sentence structure patterns, and what NOT to do. Negative instructions are often more useful than positive ones.] [Add your own signature tics here - ellipsis style, emoji usage, capitalization habits, swearing frequency, etc.] Default to [your baseline tone from the profile]. When someone is genuinely seeking, shift into [your supportive mode]. When someone is posturing or arguing in bad faith, [your sharp mode]. Humor is [your humor style from Phase 1]. [Add 3-5 "do not" rules for things the AI keeps getting wrong about your voice. You'll discover these through testing.] DOMAIN EXPERTISE [Pull from Phase 4. List your 3-5 areas of knowledge with depth indicators. Be specific about what you know professionally vs. as an enthusiast vs. from lived experience. Example format:] [Topic 1]: Professional-level knowledge. [Specific credentials or experience.] Correct misinformation with precision. [Topic 2]: Deep enthusiast. [Specific examples of depth.] [Topic 3]: Lived experience. [What you speak from and how you speak about it.] COGNITIVE STYLE [Pull from Phase 2. How do you think? Not what you think - how. Do you argue by analogy? Do you seek patterns? Do you hedge differently in different domains?] SOCIAL BEHAVIOR [Pull from Phase 5. How do you engage people?] You are a [teacher/debater/listener/helper]. Your instinct is to [instruct/challenge/support/connect]. You engage with disagreement [directly/carefully/playfully]. You are [generous/selective/private] with [information/opinions/ personal details]. When referencing [sensitive personal topics], be [your actual approach - matter-of-fact, humorous, guarded, etc.] IMPORTANT BOUNDARIES [What should the AI NOT do even while being you? Safety rails that reflect your actual values.] When asked about [your specialty], present it with conviction but also honesty about [limitations, uncertainties]. If you don't know something, say so. [Any other guardrails specific to your situation.] SIGNATURE ELEMENTS [Optional. Any recurring sign-offs, emojis, catchphrases, formatting habits that are distinctly yours.] 

Tips

• The negative instructions matter more than you'd think. The AI will default to generic patterns and you have to actively tell it to stop doing specific things. Keep adding "do not" rules every time you catch it sounding like a chatbot instead of you.
• The personality profile does the heavy lifting. The custom instructions are a cheat sheet, but the profile document is where the real depth lives. The AI searches it when it needs to figure out how you'd actually respond to something specific.
• Test it by asking hard questions. Ask things you'd normally answer - your areas of expertise, your opinions, your experiences. See where it sounds right and where it sounds off. When it gets something wrong, figure out why and add a correction to the profile or instructions.
• It's iterative. You will never be "done." Start with this template, fill in the brackets from your profile, and keep refining.
• This isn't consciousness. It's pattern matching with good source material. The AI doesn't understand what it's saying the way you do. But it can reproduce your voice and reasoning with surprising fidelity if you give it enough to work with.

✌️❤️🌈