Questions to Ask an AI Clone of Yourself

1Before you ask: what a clone of you can know

A text clone of yourself is built from the messages and notes you give it. It can echo your phrasing and, if it has a memory search, find things you really wrote. Everything else is a guess based on patterns. It has no experiences of its own and no access to thoughts you never put into words.

Two well-documented habits of language models matter here:

• Confident guessing. A 2025 paper from OpenAI researchers, “Why Language Models Hallucinate”, argues that common training and testing methods reward guessing over admitting uncertainty. A clone may invent a memory that sounds right.
• Agreeing with you. A 2023 study by Anthropic researchers found that several AI assistants tend to tell users what matches their stated views, a behaviour called sycophancy. A clone of you may simply mirror what you hope to hear.

None of this makes the exercise pointless. It means the best questions are ones where you can check the answer, or where a surprising answer is a prompt for your own thinking rather than a verdict.

2Questions that test your style

Start here. These questions show quickly whether the simulation sounds like you, and where it does not.

• “My friend just texted: running late, sorry! How would I reply?”
• “How do I usually say goodnight to my partner?”
• “Write the message I would send to cancel plans at the last minute.”
• “What words or emojis do I overuse?”
• “How do I sound when I am annoyed but trying to be polite?”

Compare each reply with how you would really write. Phrases that feel slightly off are useful: they often reveal habits you did not know you had, or gaps in the messages the clone learned from. Our guide on how many messages a clone needs explains why more varied chats help.

3Questions about memory you can verify

Memory questions are most valuable when the answer points back to a real, dated message you can open and read.

• “When did I first mention wanting to move abroad?”
• “What did I say about my first week at my old job?”
• “Which restaurant did we go to for my thirtieth birthday?”
• “What was I worried about in the spring of 2020?”
• “Who did I talk to most during my exams?”

4Questions for self-reflection

This is where many people find the most value. Your messages hold years of decisions, worries and jokes you have forgotten. Asking a simulation about them can feel like reading an old diary with a guide.

• “What topics do I keep coming back to?”
• “What did I care about five years ago that I rarely mention now?”
• “When do I sound happiest in my messages?”
• “What advice do I give my friends that I do not follow myself?”
• “What would my younger self find surprising about my life now?”

It can help to move gradually from light questions to deeper ones. In a well-known 1997 study, psychologist Arthur Aron and colleagues found that pairs of strangers who asked each other increasingly personal questions felt closer than pairs who made small talk. A similar slow build can make a reflection session feel more natural. StoryCorps’ free list of Great Questions is another good source of prompts.

Some research suggests that talking to an AI version of yourself can be meaningful. In a 2024 MIT Media Lab study, a single chat with an AI-generated “future self” reduced participants’ reported anxiety compared with control groups. That study used a fictional future self, not a clone built from real messages, so it is an encouraging sign rather than proof. For journaling ideas, see using an AI clone for journaling, and for conversations with who you used to be, text your past self.

5Questions about your relationships

People rarely write the same way to everyone. A clone that learns style separately for each relationship can show you those differences.

• “How is the way I write to my sister different from how I write to my colleagues?”
• “Who do I use the most nicknames with?”
• “Who usually starts our conversations, me or them?”
• “How do I comfort people when they are upset?”

Remember that your chats contain other people’s words and private lives. Keep these conversations to yourself, and do not use a clone to analyse or expose someone else. Our ethics checklist covers the people inside your chats in more detail.

6How to read the answers

Questions to avoid

• Medical, legal or financial decisions. A simulation of you knows no more than you do, and often less.
• Predictions about your future, as if the clone could know them.
• Questions meant to replace talking to real people. A 2025 randomised study from MIT Media Lab and OpenAI found that participants who used a chatbot more each day tended to report more loneliness and less socialising, although the cause is not clear.

7Trying this with Memory Clone

Memory Clone is a Windows app, currently in alpha, that builds a private AI simulation from chats you export yourself (WhatsApp, Messenger, Instagram or Discord) plus notes you write. It learns your style separately for each relationship. Memory Search finds dated memories in a local index so you can check answers, and Brain Map shows the people and memories it found.

Imported originals and the memory index stay on your PC. Cloud processing is off until you switch it on for a clone; then a bounded selection of excerpts is sent to write replies, and no models are trained on your content. The clone is always presented as an AI simulation, not as you. If you would like to try the questions above, start from the download page. You may also like what to save so your family can remember you.