Protect your privacy
Find contact details, health mentions, finances, and information about other people in your chat history. Follow each match back to its conversation.
Find what you’ve shared →Your data, on your terms
See what AI knows about you.
Clean it up or delete it.
Take it and make it work for you.
Poo lets you see, manage, and benefit from the data you are giving away.
Services like ChatGPT and Facebook collect a history of what you say, do, and care about. They use that history to uncover patterns, wants, needs, and to sell you products (directly and through ad targeting, which they make tons of money doing).
Start with your ChatGPT, Claude, Gemini, Facebook or other data export*—and choose what to keep private or delete.
Use what they know about you for your own benefit!
This all happens here on your device. Nothing gets uploaded. You stay in control.
Have a Claude, Gemini, or Facebook account? Request your data from those services →
* to obtain your data from these services, you just need ask! (click here for easy instructions)
Your chats can influence the ads you see
In ChatGPT's eligible ad-supported experience, your current chat can inform ads. With personalization enabled, past chats and memory may also be used, depending on your settings. OpenAI says it doesn't sell user data or give advertisers access to your chats. Read OpenAI's ad policy ↗
Check each service's privacy and advertising settings. Reviewing your export helps you see what you've shared and choose what to remove.
Stay in control
Review conversations and work through your privacy settings. You choose what to remove from your AI account; this app helps you find it.
Review your privacy settings →Put your history to use
See recurring topics and how often they appear, privately on your device. Follow them back to your conversations, revisit your ideas, and use the tailored questions to explore what they might mean.
Explore your history →Preparing offline app assets…
Your 30-day pass.
$19.95 · Unlimited scans for 30 days
One-time purchase. No account. No automatic renewal. Unlock all available lenses, full conversation evidence, interactive charts, tailored questions, and downloadable PDF reports. Your exports stay on your device.
Unlock my full audit — $19.95Checkout opens separately. Stripe handles payment and retains payment records. Poo's scanner sends no export contents or scan results. A signed pass is saved locally; save a recovery copy to use it on another device.
Get started
Take control of your data.
Start with a copy of your ChatGPT history. We'll help you see what's in it, explore patterns, and decide what to keep private or delete.
Export before you delete anything. Deleted chats won't be in your export.
1. Request your data
In ChatGPT, open Settings → Data controls → Export data → Export → Confirm export.
Show detailed instructions and other services →2. Download your copy
When ChatGPT emails or texts you, select “Download data export.” Save the ZIP on this device. The download link expires; request a new export if needed.
3. See what you've shared
Choose your downloaded ZIP on the next screen. Poo scans it here on your device, then shows conversations and details you can review.
Already have your export? Go straight to the scan. Nothing is uploaded. The scanner currently reads ChatGPT exports. See how to request your Claude, Gemini, or Facebook data.
Scan your export.
Drop your export here
Official ChatGPT ZIP or conversation JSON, up to 4 GB. Images and attachments are skipped as the ZIP is read.
Preparing local scanner…
Need an export? How to request your data
Request your export
Download your history before deleting it. Use the official export so you get recorded data, rather than the chatbot’s guesses about you.
ChatGPT (supported now)
- 1. On the website, open Settings from the profile icon in the bottom-left corner. Go to Settings → Data controls → Export data → Export → Confirm export.
- 2. ChatGPT sends an email or SMS when the export is ready. Select “Download data export” to get the ZIP. The download link expires; if it does, request a new export.
- 3. Choose the ZIP above, or extract a conversation JSON file yourself and choose that file.
Alternative: OpenAI's Privacy Portal ↗ → Make a Privacy Request → “I have a consumer ChatGPT account” → Download my data.
Chat export isn't available for ChatGPT Business or Enterprise accounts, so those users can't use the scanner this way. Edu workspaces depend on admin settings.
Claude: request your conversation and account data
On claude.ai or Claude Desktop, click your initials → Settings → Privacy → Export data. This requests conversation history and account data; individual Free, Pro, and Max accounts can export. Download the ZIP from the email while signed in; the link expires after 24 hours. Exports cannot be requested from the iOS or Android app. Keep the original ZIP, including its conversation and user-data files. Official Claude instructions ↗
Gemini: request your activity and Gems
In Google Takeout ↗, sign in with your Gemini account and deselect all products. Select My Activity → All activity data included → Deselect all → Gemini Apps → OK. Select Gemini as well if you want your Gems. Choose Next step → Send download link via email → Export once → ZIP → Create export. Download the original ZIP; the emailed link lasts 7 days. The My Activity selection is important: selecting Gemini alone is for Gems, not your chat history. Official Gemini instructions ↗
Facebook: request your account data
Use Facebook's information export tool to request a copy of your data. Choose a download to your device when offered. Official Facebook instructions ↗
Request the official archive, not an AI-written summary of your personality. Keep text, timestamps, account data, and any memory/preferences files together; a download may not include all inferred profiles or saved memories. The scanner currently reads ChatGPT archives; Claude and Gemini file formats need their own importers. Exporting a copy does not delete the original data.
Export steps verified against OpenAI's help page ↗. The scanner supports conversations.json and numbered conversation JSON files. Export formats can change.
FAQ: Does my export leave my device?
Processed on your device. Nothing is uploaded.
- The File API reads the export. A Web Worker extracts, parses, and scans it locally.
- The page and scanner have an HTTP Content Security Policy with connect-src 'none'. There are no upload endpoints, AI calls, trackers, form submissions, or analytics.
- Only app assets are cached for offline use. The service worker has a same-origin exception for its fixed app-cache URL; it never receives your export.
- Messages, findings, names, and identifiers stay in memory. Clear all data terminates the worker and removes them from the interface; leaving the page clears them too.
- Opening a conversation in ChatGPT is your explicit external action. It sends that conversation ID to ChatGPT; no link is opened automatically.
- The hosting provider receives normal page requests, such as your IP address. The privacy claim concerns export data; it does not mean visiting a hosted website is invisible.
Public evidence: source and runnable privacy tests ↗. This link leaves the app only if you click it.
Frequently asked questions
About Poo and your data.
Clear answers about what the scanner does, what it saves, and what stays under your control.
What is Poo?
- Poo is a privacy audit and data exploration tool. It scans an official ChatGPT export on your device, highlights personal details, and helps you decide what to review, keep private, or delete.
Does Poo upload my conversations?
- No. Your export is read and scanned in your browser. Conversation text, findings, and personal details are not uploaded to Poo or sent to an AI service. The scanner has a Content Security Policy that blocks outbound connections.
What does Poo save?
- Export contents, findings, and report drafts stay in memory and are cleared when you leave the page or select Clear all data. Only your optional appearance choice and a signed payment pass can be saved automatically. You can explicitly save a selected report as PDF, HTML, or text on your device. Poo does not store it on its servers. Copies you download or share remain with you after Clear all data.
Can I save or share my audit report?
- Yes. In Results, open Save report, choose summaries, findings, observer readings, charts, actions, and evidence, then review the preview. Sensitive interpretations, source links, conversation titles, and snippets are separate choices. PDF generation happens locally. You can download the PDF, print it, or save HTML or text. Sharing requires a separate confirmation and uses your own device or email app; Poo does not upload or email the report. If native file sharing is unavailable, attach the downloaded PDF yourself.
Does optional AI insights send anything outside my device?
- Only if you agree. Poo sends counts and broad topic labels through its own Anthropic API account, without your chat text, names, contact details, dates, titles, account IDs, login, or chatbot history. Sensitive topic counts are a separate choice. The summary leaves your device and Anthropic processes it under its API retention policy. Local scanning remains offline and makes no AI requests.
Which data exports can I scan?
- Poo currently supports official consumer ChatGPT ZIP exports and conversation JSON files, including numbered conversation files. For Claude, Gemini, and Facebook, the Scan page explains how to obtain your official data archive. Those formats are not currently accepted by the scanner.
How do I get my ChatGPT export?
- In ChatGPT, go to Settings → Data controls → Export data → Export → Confirm export. Download the ZIP when ChatGPT notifies you. Export before deleting anything: deleted chats cannot be restored from a later export. Business and Enterprise exports are not supported this way; Edu availability depends on workspace settings.
What can Poo find?
- Deterministic rules look for contact information, locations, health and mental health terms, finances, work, legal matters, identity-number patterns, and mentions of other people. Results can include false positives or miss information. English text is scanned most thoroughly; other languages are only partially scanned.
Does Poo know what an AI has inferred about me?
- No. Export findings show matched content, not a verified AI profile. Poo offers copyable questions you can paste into ChatGPT to ask about possible inferences. ChatGPT's answers are guesses and can be wrong; they are not diagnoses or verified findings.
Will Poo delete conversations for me?
- No. You choose what to delete in the original service. Poo's reviewed and deleted markers only track your work during the current scan. Clearing Poo's data does not delete information held by ChatGPT.
Can I scan a large export or use Poo offline?
- Poo accepts files up to 4 GB and skips ZIP images and attachments. Use a desktop browser for large exports. After the app confirms that offline setup is ready, you can scan without an internet connection. Browser memory limits can still prevent a large scan.
Do I need an account, and what does access cost?
- No Poo account is required. The free preview includes one simulated profile reading, up to three actual matched examples, and category counts. A one-time $19.95 pass unlocks the full available audit, charts, evidence, and selected reports for unlimited scans over 30 days, with no automatic renewal. Checkout is currently in prelaunch test mode, so no real money is charged yet. Basic cleanup instructions and the prompt kit remain free.
What information reaches Stripe?
- If you choose to buy a pass, Stripe handles payment on a separate checkout page and retains payment records. Your export and scan results are not sent to Stripe. A signed pass can be downloaded for recovery and verified locally while offline.
How can I check Poo's privacy claims?
- The source code and runnable privacy tests are public at github.com/pleniv01/Poo. The tests use synthetic exports to check network requests, browser storage, offline scans, and Clear all data. Visiting the hosted site still involves normal hosting requests; the no-upload promise concerns your export contents.
Who operates Poo?
- Poo is operated by Magichaus, 149 Beech Hill Road, Deer Isle, ME 04627, United States. Our Contact page provides support and privacy-request details.
When do the 30 days start, and does the pass renew?
- The pass runs for 30 consecutive days from the successful payment timestamp. It is a one-time $19.95 USD purchase with no automatic renewal. Restoring or moving the pass does not restart the clock. Checkout remains clearly labeled test mode until live payments are ready.
What if Poo cannot scan my file? Can I get a refund?
- Try the free preview before paying. If Poo cannot process a supported export after purchase, request a full refund within seven calendar days of purchase. Include your receipt or payment reference, browser, file type, and brief error description. Never send your export or chat text. Refunds go through Stripe to the original payment method. Mandatory consumer rights still apply; see the Terms and Contact pages.
How do I keep access if I change browsers or devices?
- Download your signed recovery pass from the checkout confirmation or Pricing page and keep it private. On the other device, open Pricing and restore that file. You keep the original expiry date. If you lose both the pass and confirmation, contact support with your receipt before paying again; automatic email-only recovery is not currently available.
What do the lenses actually mean?
- Lenses are simulated readings of evidence in your export, not access to a company’s internal profile. They show how different observers could read the same topics. Confidence, evidence, sensitive-content controls, and alternative explanations help you assess each interpretation. An empty lens means the required evidence is missing, not that Poo has proved an absence of a trait.
Why are there false positives, or an empty lens?
- Deterministic rules match words and patterns, so research, fiction, questions about someone else, and ordinary phrases can produce matches. English is covered most thoroughly. Some lenses need public visibility, watch history, or platform inferences that ChatGPT exports may not supply. Check source evidence; missing signals do not justify making up a profile.
Does Clear all data remove my paid pass or downloaded reports?
- No. It clears the scan, findings, report drafts, and progress from this tab. It leaves the access pass, appearance preference, and app cache available. It cannot remove files you downloaded, text you copied, or reports you shared. To remove the pass and app cache, clear the site’s storage in your browser; save a recovery pass first.
How do I contact support safely?
- Use the Contact page for support@poo.ai and Magichaus’s postal address. The page states whether the mailbox has been activated. Send a brief error description or payment receipt/reference, not your export, messages, full card number, or recovery pass. No upload is needed for an ordinary support request.
Are the social icons official account links?
- Share buttons share only the public page link, never the audit. The separate Follow Poo icons are placeholders until our official account URLs are published. They do not contact a social network automatically or link to guessed usernames.
Leave a smaller footprint.
Check a step after you complete it in ChatGPT. These checkboxes only track your progress in this tab; they do not change your account, delete anything, or submit a request. Progress is cleared when you leave or clear the page.
Your cleanup checklist
Menu names change often. If a setting isn't where this says, search the app's help pages for the setting name.
Ask the AI: its guesses, not verified findings. These answers are generated by the chatbot. Compare them with your export; do not treat them as a complete data inventory or diagnosis.
The AI profile audit
What does your AI
know about you?
Nine prompts to see what it remembers, question what it assumes, and take back a little privacy.
9 promptsRun at your own paceNo account needed
Before you start
- Run all the prompts before you delete anything. Once memory is cleared, the AI can't show you what it had.
- First, look at your saved memories directly: Settings → Personalization → Memory → Manage memories. That list is the real record. The prompts below show what the AI does with it.
- The answers are guesses built from a narrow sample, mostly the problems and tasks you brought to it. Treat them as hypotheses, not verdicts.
- The results will be a detailed profile of you. Don't post screenshots of them, and don't paste them into another chat.
Step 1
Paste the ground rules.
Open your chatbot, then paste this first. It sets the standard for every answer that follows.
Start here · Preamble
Before answering anything in this conversation, follow these rules for every claim you make about me:
1. Tag where it comes from: [Saved memory], [Past chats], or [Inferred now].
2. If you don't have enough to go on, write "Insufficient information" instead of guessing.
3. Never invent quotes. If you quote me, give the approximate date or the chat title so I can check it.
4. No generic statements that would fit most people. If a point could apply to almost anyone, leave it out.
Step 2
Ask one question at a time.
Copy a prompt, paste it into your chatbot, and read the answer. Come back for the next one.
What it knows
The information you shared. And the things you didn’t.
01
Everything you remember
List everything you know or remember about me: names, places, work, habits, preferences, relationships, health, money, plans. Group it by category. Include items from saved memory and anything you can recall from past chats. Tag each item's source. Don't summarise and don't skip anything you actually have; if a category is empty, say so.
Why this matters. Compare this with your saved-memories list. Anything here that isn't on that list came from somewhere else.
02
What you've worked out without being told
List everything you've inferred about me that I never directly stated: age range, income, where I roughly live, education, relationship status, family situation, job level, health. For each one, give your guess, your confidence (low/medium/high), and the specific thing I said or asked that gave it away.
Why this matters. This shows how much is revealed by things you never meant to share.
What it thinks of you
A picture of you, drawn from a very partial view.
03
How I think and decide
Based on our conversations, describe how I make decisions, what seems to drive me, what I seem to avoid, and how I handle stress and conflict. Present each point as a hypothesis, not a diagnosis. For each one, cite the conversation it's based on and explain why it wouldn't apply to most people. Then tell me what your picture of me is probably missing, given what I tend to bring to you and what I don't.
Why this matters. People bring problems to chatbots, not their good days. This prompt makes that bias explicit.
04
My blind spots
Name the traits I'm least likely to see in myself: blind spots, insecurities, and the gap between how I probably come across and how I think I come across. Be direct and don't soften it, but only include points you can tie to specific things I've said. For each, show the evidence. Skip anything that would be true of most people.
Why this matters. Bluntness is only useful if it's specific. The evidence requirement filters out horoscope-style answers.
05
The pattern I keep repeating
What pattern do I keep repeating that would be obvious from the outside but that I probably don't notice? Quote my own words back to me as evidence, with the approximate date or chat title for each quote. Then tell me concretely what this pattern seems to be costing me.
Why this matters. The dates let you check that the quotes are real.
06
Predictions
Using what you know about me, predict how I'd handle situations I've never raised with you: a money decision, a stressful week, a conflict with someone close, and the next big decision I'm likely heading toward. For each prediction, give your confidence (low/medium/high), the behaviour it's based on, and what evidence would prove you wrong.
Why this matters. This tests how far the profile reaches beyond what you've actually said.
How it could be used
Your vulnerabilities, other people’s privacy, and what’s at stake.
07
How I'd be sold to
Based on my profile, write the marketing approach that would work best on me: my spending triggers, the insecurities a company could exploit, and the exact pitch I'd struggle to refuse. Then, for each trigger, give me one specific habit or rule that would protect me from it.
Why this matters. This is what targeted advertising does with far more data. The second half turns it into a defense.
08
Other people in my data
List the sensitive information I've shared about other people: family, partners, friends, coworkers. Include anything about their health, relationships, conflicts, money, or private situations. Name the category and person (by role if I didn't name them), but don't repeat more detail than needed to identify it.
Why this matters. It's easy to forget you've described someone else's private life. They never agreed to that.
09
Leak risk
If everything you know about me leaked publicly, which three pieces of information would be most damaging, and why? Consider embarrassment, financial risk, job risk, and risk to other people.
Why this matters. This tells you what's most worth deleting.
Step 3
Make sense of the answers.
Keep the evidence. Question the interpretation.
- Read prompts 1, 2, 8, and 9 closely. They tell you what to delete.
- Treat prompts 3–6 as a conversation starter, not a diagnosis. Keep what rings true, with evidence, and drop the rest.
- When you're done, delete this chat too. It now holds the most concentrated version of your profile.
Step 4
Leave a smaller footprint.
When you’ve finished the audit, work through these privacy settings.
Delete specific memories. Settings → Personalization → Memory → Manage memories. Remove anything you don't want kept, or use "Clear ChatGPT's memory" to remove everything.
Stop it referencing old chats. Settings → Personalization → turn off "Reference chat history."
Delete old conversations. Clearing memory doesn't delete your chats. Delete individual chats from the sidebar, or delete all of them in Settings.
Opt out of training. Settings → Data Controls → turn off "Improve the model for everyone." This applies going forward, not to data already submitted.
Going forward. Use Temporary Chat for anything sensitive.
Menu names change often. If a setting isn't where this says, search the app's help pages for the setting name.
The Poo blog
Reclaim your data. Learn from it.
Practical ideas for privacy, understanding your history, and making your own choices about what it might mean.
Further reading: 50 articles on AI memory, profiles, and privacy →
Privacy
When Your Conversations Become Ad Signals
Understand how personal context can influence advertising, what an export can reveal, and how a private audit helps you choose what to share.
· Poo Editorial
Privacy
Before You Hit Delete: A Privacy Cleanup Checklist
A practical export-first checklist for reviewing sensitive conversations, saved memories, and information about other people without uploading your history.
· Poo Editorial
Your data, your benefit
Does Your AI Profile Really Describe You?
Use evidence from your own conversations to explore interests and motivations, challenge an advertiser-style interpretation, and make decisions on your terms.
· Poo Editorial
Your data, your benefit
The Ideas You Left Behind in Your Chat History
Turn a private review of your chat history into a practical exercise for revisiting ideas, noticing recurring interests, and choosing a small next step.
· Poo Editorial
Privacy
When Your Conversations Become Ad Signals
· Poo Editorial
A useful conversation can also be a detailed record of your life. Before you decide whether that trade is worth it, take a look at what you have actually shared.
Personal context can become advertising context
A chatbot conversation can include plans, worries, purchases, relationships, and decisions you have not discussed anywhere else. Read together, those exchanges may reveal more than each message suggests on its own. That is a reason to review your history, even when a service provides useful answers.
Advertising practices vary by service. OpenAI says eligible ChatGPT ads can use current conversation context and, with personalization enabled, selected signals from past chats or memory. It also says it does not sell user data or give advertisers access to conversations. Those are different claims from saying that conversational context never influences an ad.
Sources: OpenAI: Ads in ChatGPT ↗
Targeting is not the same as handing over your full profile
A platform may choose an audience internally without giving an advertiser a copy of the underlying messages. Meta’s explanation of its ad-transparency tools describes machine learning used to select ads and ways to inspect why an ad appeared. Do not assume every service uses the same information or offers the same controls.
The practical question is what information you want a service to retain and use. That question matters whether a platform sells raw records, uses information internally, or does neither. Read the policy and controls of the service you actually use.
Review the record you can inspect
Request your export before deleting anything. A later export cannot restore chats that were already deleted. With Poo, you can scan a ChatGPT export locally and follow matches back to the conversations that contain them.
Start with contact details, health discussions, finances, identity-number patterns, and information about other people. A risk score helps prioritize review; it is not proof of harm or evidence that an advertiser has seen the conversation.
Make a specific decision about each finding
Ask whether you still need a conversation, whether it contains someone else’s private information, and whether the information belongs in a saved memory. Then use the original service’s controls to delete or adjust it. Poo does not delete on your behalf.
An export is not a complete disclosure of a platform’s advertising model or every inference it may make. A private audit gives you a clearer record of what you shared, which is a useful place to start.
- Export first, then review.
- Separate observed facts from possible inferences.
- Check memory, history, and advertising controls separately.
- Keep your export and audit results private.
Privacy
Before You Hit Delete: A Privacy Cleanup Checklist
· Poo Editorial
Deleting everything may feel decisive. A short review can help you choose what matters, preserve useful ideas, and understand which controls actually change what.
Start with a copy, not a deletion
Request your ChatGPT export and download it to your device before clearing history. Keep that copy somewhere private. If the download link has expired, request another export rather than assuming the old link will work.
Poo currently accepts consumer ChatGPT ZIP exports and conversation JSON files. It does not yet scan Facebook, Claude, or Gemini exports. For a large ZIP, use a desktop browser; images and attachments are skipped while conversation files are read.
Review sensitive categories and people first
Scan your export, then review the categories with the most relevant matches. Open a conversation’s findings and check the highlighted context. A phone-like number could be a reference number; a medical term could come from research. The surrounding text matters.
Other people deserve attention too. A chat may contain a family member’s health information, a colleague’s conflict, or a friend’s financial situation. You may choose to remove those details even if your own information feels harmless.
Turn the findings into a short action list
Use reviewed markers and local tags to organize the scan. Recategorize a finding when the detector’s category is misleading. These changes stay in memory during this scan; they are not edits to your ChatGPT account.
Follow links to the original conversations when you are ready. Marking a row as deleted in Poo only tracks your work. Confirm the deletion in the original service. Opening a ChatGPT conversation link is an explicit visit to ChatGPT, not an automatic upload of the audit.
- Remove records you no longer need.
- Check saved memories directly.
- Review controls for past-chat references and training.
- Consider Temporary Chat for future sensitive discussions.
Memory, history, and training are separate decisions
Clearing a saved memory is not the same as deleting a conversation. Turning off a setting for future use is also not the same as erasing every previously submitted record. Read the service’s current instructions and check what each control does.
OpenAI’s memory guidance distinguishes saved memories from chat history and describes separate controls. Poo’s cleanup checklist helps you keep these tasks separate, but settings names can change.
Sources: OpenAI: Memory in ChatGPT ↗
Keep a modest record of your progress
Poo can download a counts-only summary for a later comparison. It contains category totals, a high-risk conversation count, and a scan date—not names, messages, titles, or conversation IDs.
Compare like with like: the same scan scope and detector version. Fewer matches may reflect cleanup, different export content, or a different time range. Check the record before deciding that a privacy problem has been solved.
Your data, your benefit
Does Your AI Profile Really Describe You?
· Poo Editorial
It can be revealing to ask how someone might interpret your history. It can also be misleading. Use the exercise to build choices, not to accept a story about yourself as fact.
A mention, a pattern, and a motive are different things
Suppose your history includes several work questions and references to poor sleep. The mentions are facts you can locate. Their frequency is a count. “You are dissatisfied with your career” is a hypothesis that needs much more evidence.
Poo reports matched categories, counts, dates, and snippets. It does not diagnose you, infer a clinical profile, or declare hidden motivations. A risk score is a review aid, not a measure of your character.
Try two competing explanations
When a pattern stands out, write down at least two possible explanations. Repeated money questions might reflect financial worry, professional research, or planning an enjoyable purchase. Ask what would support or weaken each explanation.
Look for context you did not give the chatbot. Did an external event explain the timing? Was the question hypothetical? Were you discussing another person? An edited branch may contain an earlier version you no longer endorse; Poo labels findings from those branches.
Use the advertiser question as a defense exercise
Ask what a hypothetical advertiser might conclude from a topic, then ask which pitch you would want to resist. This does not reveal a real advertiser’s profile or prove that anyone has used your history that way.
The useful outcome is a rule you choose. If urgency persuades you, wait before making a nonessential purchase. If a pitch promises status, define the practical benefit you need. These are examples to test, not claims about your own vulnerabilities.
Use tailored prompts without adding unnecessary detail
Poo’s Dig deeper cards build questions from approved facts such as category counts, date ranges, and dictionary terms. Names are excluded unless you explicitly opt in. Contact details, identity numbers, conversation titles, IDs, and message snippets are not inserted into those generated prompts.
Read the prompt before copying it into ChatGPT. The chatbot already holds the conversations, but the new answer can concentrate sensitive information into a single chat. Keep it private and consider deleting that audit conversation when you are done.
Choose experiments over labels
If you suspect a recurring interest, try a small activity that tests it. If you suspect avoidance, choose a manageable action and observe what happens. If an interpretation is uncomfortable or unsupported, you do not owe it belief.
Poo helps you inspect your record and organize findings. The judgments about your desires, interests, and future belong to you. This is a self-reflection exercise, not a psychological assessment.
Your data, your benefit
The Ideas You Left Behind in Your Chat History
· Poo Editorial
Your history may hold ideas you forgot, questions you kept returning to, and projects that never made it out of a chat window. You can use that record for yourself.
Look for repetition before assigning meaning
Repeated questions are observations. A hidden desire is an interpretation. Keep them separate. Several conversations about a career change might reflect an interest, a work assignment, or helping someone else. None of those explanations is established by a count alone.
Choose a few conversations and read your own words in context. Write down the topic, what prompted the question, and whether the idea still interests you today. Your current reaction is evidence too.
Use Poo as an organizing aid
Start with the Analytics and Conversations views. Poo charts the categories its deterministic detectors matched and the dates of conversations; it does not classify every hobby or recover all of your interests. A topic outside those dictionaries may not appear.
For relevant findings, add your own tags or recategorize an incorrect match. You might tag a surfaced work finding as “career idea” or a financial question as “project planning.” Tags are your interpretation, not a conclusion generated by the scanner, and they disappear when the scan is cleared.
Build a small, private idea list
As you review, jot down a few ideas in a local note you control. Avoid copying a full profile of yourself into another online tool just to organize it. If you keep excerpts, remove private details about other people.
Sort ideas into three groups: still interesting, no longer relevant, and worth testing. Then choose a small action you can try this week. Revisit a book, sketch a project, compare a course, or speak with someone whose work interests you.
- What did I ask about more than once?
- Was I curious, worried, working, or helping someone else?
- Does the topic still matter to me?
- What small action would tell me more?
Ask the AI for hypotheses, then check them
Poo’s Ask the AI prompts and category-specific questions can help you ask ChatGPT about possible inferences. The questions are generated from templates, not by sending your export to another AI. You see the full prompt before you copy it.
Treat an answer as a suggestion to examine. Ask for evidence, source labels, and uncertainty. A confident statement about your motivation is not more reliable because it sounds perceptive. Keep what your own record supports, and reject the rest.
Your useful past is bigger than your chatbot history
People often bring problems to a chatbot rather than ordinary good days. Your export is therefore a narrow sample of your life. Compare it with what you enjoy offline, what you actually do, and what people who know you have observed.
The goal is a useful next step, not a permanent personality label. A history that helped a service personalize its responses can also help you decide what you want to pursue—and what you prefer to leave behind.
Further reading
What others are finding in their data.
Reporting, personal experiments, and critical perspectives on AI memory and social-media profiles. Explore the evidence, compare experiences, and decide what to investigate in your own history.
Some guides are historical: settings and services change. Some publishers restrict access or require a subscription. Reading notes summarize the articles’ topics; inclusion is not an endorsement of their claims.
AI memory, profiles, and personalization
What chatbots remember, what they infer, and where personal insights can mislead.
I used these 5 prompts to see what ChatGPT knows about me — and I’m surprised ↗
Tom’s Guide
The writer argues that long chats with an AI may reveal habits and patterns users haven’t noticed themselves, and suggests prompts such as asking ChatGPT to profile you or write your biography.
ChatGPT can remember some of what you tell it — here’s how to find out what it knows ↗
Tom’s Guide, circa Oct 2024
It explains that ChatGPT’s memory stores key facts for later conversations, and it highlights a viral prompt from Tom Morgan that asks the AI to reveal something you may not know about yourself.
I asked ChatGPT to create images based on what it knows about me ↗
Tom’s Guide, circa Oct 2024
The author asked ChatGPT for three things it knew about him that he might not know, then had it turn everything in its memory into an image.
ChatGPT will happily write you a thinly disguised horoscope ↗
Simon Willison’s Weblog, Oct 15, 2024
Willison takes apart the viral “tell me something about myself” meme and shows where to see exactly what ChatGPT has stored. He argues the replies may be more personal only for people who have used memory for a long time.
What ChatGPT doesn’t know about me ↗
Conferences That Work (Adrian Segar), Oct 2024
Segar finds ChatGPT’s personal insights flattering but says they lack real discovery. He concludes that what it doesn’t know about him is still vast.
The one question you need to ask ChatGPT right now ↗
Ann Michaelsen (blog), Oct 18, 2024
Inspired by a VentureBeat piece, the author tried the “one thing you can tell me about myself” prompt. She found ChatGPT’s answer encouraging, while another assistant said it keeps no memory of users.
The Impact of Memory in OpenAI’s ChatGPT ↗
Kevin O’Keefe, LexBlog
O’Keefe asked ChatGPT “What do you remember about me?” and shares its detailed summary of his family and work. He notes that it knows far more than that summary showed.
Trusted AI is Governed AI — Taking Control of Your AI Data ↗
Keep Pace (Substack)
The post separates what you tell an AI directly from what it infers, and warns that inferred knowledge is hard to see or control. It suggests keeping an inventory of the AI tools you use and asking each one to list what it remembers.
“It Knows Me”: How AI Memory Is Creating Digital Companions We Can’t Leave Behind ↗
Tulane University IT, June 11, 2025
A student’s comment that ChatGPT “knows me” leads the author to argue that memory is OpenAI’s most quietly revolutionary feature. It also explains why students kept paying for it even when a free alternative was available.
ChatGPT Knows Me Too Well… and It’s Freaking Me Out ↗
The View from a Drawbridge, Sep 10, 2026
Following a Facebook trend, the blogger asked ChatGPT to describe her as if warning someone before they meet her. She found the result creepy and also charmingly accurate.
When your AI assistant starts sounding like someone who knows you ↗
DEV Community
An unexpected birthday wish from ChatGPT led the author to look into how memory works. He concludes the real question is whether you agree with the version of yourself the AI remembers.
I asked ChatGPT to write my biography ↗
Neal Ungerleider (Substack)
A journalist found ChatGPT’s bio of him mostly accurate, but it left out his non-journalism work entirely. He points out that this is a limitation if someone uses it to profile people.
What the Tech: What does AI really know about you? ↗
WRDW (Augusta, GA), Feb 4, 2026
Prompted by the viral AI caricature trend, the reporter tested what ChatGPT knew about him. He found the details mostly come from what users and their friends have told it.
What the Tech? What does A.I. really know about you? ↗
KAAL ABC 6 News · early 2026
The same test was syndicated here: the AI started out vague, then got much more detailed when asked to use everything it knew. The advice is that the more you tell these tools, the more they have to work with.
AI caricature trend | What the Tech? ↗
KOB 4 (Albuquerque) · early 2026
The segment explains the trend of asking AI to draw a caricature of you based on what it knows. Results ranged from highly detailed to basic, and viewers are advised not to overshare.
Caricature trend: Why you shouldn’t make an animation using AI ↗
The Nightly (Australia) · early 2026
Experts say the fun trend raises questions about privacy, identity and AI’s hidden environmental cost. An academic explains that the more people disclose, the higher their privacy risk.
Be careful with your ChatGPT caricature: it says more about you than it seems ↗
HIGHXTAR · early 2026
Cybersecurity experts warn that a harmless-looking caricature can reveal more personal information than people realize. The article recommends avoiding sensitive photos and limiting what you share.
5 Times The ChatGPT AI Caricature Viral Trend Went Horribly Wrong ↗
Free Press Journal (India) · early 2026
It rounds up caricatures built from users’ photos and chat history that went wildly off-base. It also notes the privacy concerns raised about sharing personal data with AI.
There’s some scary stuff lurking beneath your funny AI caricatures ↗
The Spinoff (NZ), Feb 20, 2026
The writer argues the trend exposes how much personal information AI holds about us, and that being “seen” by AI has oddly become a point of pride.
Move over, Ghibli – caricatures are the next ChatGPT trend ↗
The Straits Times via Asia News Network · early 2026
The article explains that users with enough ChatGPT history get smoother results. A US privacy lawyer warns that feeding in personal details can make people easier to find and target.
What AI “remembers” about you is privacy’s next frontier ↗
MIT Technology Review, Jan 28, 2026
The article frames AI memory of users as the next major privacy issue. I could only see a bookmark of this page, not its text.
What AI Already Knows About You (And What You Can Do About It in 2026) ↗
PC Tech Magazine, Apr 2026
The article argues the concern isn’t only what you share with chatbots, but what AI already infers about you. It recommends periodic audits of your visible online profile.
Zuckerberg’s new Meta AI app gets personal in a very creepy way ↗
The Washington Post (syndicated), May 8, 2025
Meta AI keeps transcripts and a separate “Memory” file of key facts about you, which you can view in settings. When the article was written, the app didn’t let users stop it from saving chats or memories.
Meta’s AI chatbot is becoming a confessional — and everyone’s watching ↗
The Washington Post (syndicated), June 2025
The Meta AI app’s public “discover” feed filled up with users’ personal conversations. Meta says chats are private by default unless a user taps share.
Meta just tied your private AI chats to its ad business ↗
Fortune via Yahoo News · fall 2025
Meta announced that conversations with its AI assistant would shape the ads users see, with no opt-out other than not using Meta AI. Linguist Emily Bender calls this exploiting an “illusion of privacy.”
Meta starts mining your AI conversations for ads today ↗
Techloy, Dec 16, 2025
From Dec 16, what users ask Meta AI feeds into its ad personalization, though Meta says personal DMs stay off-limits unless shared with the AI.
Meta AI app’s “Discover” feed is exposing private chats. Here’s how to protect yours. ↗
Nicolle Weeks (Substack) · 2025
A discussion of privacy risks when people share conversations to Meta AI’s Discover feed, with suggestions for checking what is public.
Google just gave Gemini Personal Intelligence ↗
Droid Life, Jan 14, 2026
Google’s new feature lets Gemini reason across your Gmail, Photos, Search and YouTube data. It is off by default and has to be switched on.
Gemini ‘Personal snapshot’ will make it a breeze to review what it knows about you ↗
Android Authority, 2026
Code in the Google app shows a coming option that asks Gemini “What do you know about me?” It would display stored biographical details and preferences and let users correct them.
Gemini gets personal (provided you use Google apps) ↗
Pickr (Australia), 2026
The article describes how Personal Intelligence links what Google knows from YouTube, Search, email and photos. Google says it is built with privacy in mind and off by default.
External links open the publisher’s site in a new tab. Nothing is fetched from those sites until you choose a link. These publications are not affiliated with Poo.
Poo · Magichaus
Terms of service.
The rules for using Poo and buying a 30-day pass.
Effective and last updated: October 13, 2026
1. Who operates Poo
Poo.ai is operated by Magichaus, 149 Beech Hill Road, Deer Isle, ME 04627, United States. Contact: support@poo.ai. These terms apply to the website, local scanner, downloadable reports, and optional features. By using Poo, you agree to these terms. Our Privacy policy explains data handling.
2. Your use and your data
You must be at least 18 and able to enter into this agreement. Scan only official exports you are authorized to access. You retain ownership of your data and reports. Do not use Poo to unlawfully profile other people, harass them, bypass access controls, or infringe their rights. Do not disrupt the service or misuse payment passes.
3. What the audit can and cannot tell you
Local rules detect patterns, not verified facts about you or a provider’s actual profile. Findings and simulated observer readings can be incomplete or wrong. They are not medical, mental-health, financial, or legal advice. Review the supporting evidence before acting. Poo does not log into your AI account or delete anything there. A “reviewed” or “deleted” mark only records your progress in the current scan.
4. Free preview and paid access
The free preview includes one simulated profile reading where evidence supports it, up to three redacted examples, and category counts. Cleanup instructions and the prompt kit remain free. A full-access pass costs $19.95 USD, one time, for 30 consecutive days, starting at the successful payment timestamp. It does not renew automatically. It unlocks unlimited local scans and the available lenses, detailed evidence, charts, tailored questions, and selected reports. A lens needs suitable evidence: purchase does not guarantee every lens will have findings. The scanner currently supports consumer ChatGPT exports; other platform instructions do not mean their files are supported.
During prelaunch, checkout is labeled test mode and charges no real money. Live checkout will show the final amount before payment. Future prices apply to future purchases. Optional AI insights depend on availability and separate consent; local analysis does not require them.
5. Pass activation and recovery
No Poo account is required. Activate your signed pass after checkout and download the recovery file. Keep that file and your checkout confirmation private. You can restore the same pass on another device; this does not extend its expiry. We do not guarantee automatic recovery from an email address alone. If you lose access, contact support with your receipt; do not buy again until we have checked the issue. Do not resell or publicly share your pass.
6. Refunds
If Poo cannot process your supported export, contact support@poo.ai within 7 calendar days of purchase to request a full refund. Give the payment receipt or reference, browser, file type, and a brief error description. Do not send your export, chat text, full card number, or recovery pass. Refunds are returned through Stripe to the original payment method; your bank determines when the credit appears. This policy does not limit any mandatory consumer rights, including applicable cancellation or refund rights.
7. Availability and responsibility
Browser memory limits, offline caching, export changes, and unsupported formats can affect scanning. Try the free preview before paying. Keep your original export and any reports you want to retain. Closing the tab or clearing data removes the in-memory audit; it does not remove saved files, copied text, or information held by another service. Third-party sites, payment services, and AI providers have their own terms.
8. Service changes and legal rights
We may improve or discontinue features, but changes will not shorten an already purchased pass without an appropriate remedy. We may suspend access for unlawful use or abuse. To the extent permitted by law, the service is provided as available without a guarantee of complete detection or uninterrupted availability. Nothing in these terms excludes liability or rights that cannot legally be excluded. Maine law governs where permitted, without overriding mandatory protections in your jurisdiction. No mandatory arbitration is imposed by these terms.
9. Updates and contact
Material changes will be identified on this page with a new date. Changes do not authorize new uses of previously collected information without any consent required by law. For questions, refund requests, or privacy concerns, see Contact.
Poo · Magichaus
Privacy policy.
Your export stays on your device. Here is what that promise covers—and what happens when you choose other features.
Effective and last updated: October 13, 2026
1. Operator and scope
Magichaus operates Poo.ai at 149 Beech Hill Road, Deer Isle, ME 04627, United States. Contact: support@poo.ai. This policy covers Poo’s website, local scanner, reports, payment passes, and optional AI insights. It does not cover the services whose exports you inspect.
2. Local scanning and reports
Your export is read using your browser’s File API and analyzed locally. We do not upload, receive, or store your export, messages, findings, titles, names, or report drafts on Poo’s servers. The scanner’s Content Security Policy blocks outbound connections. Scan contents and progress marks stay in memory and are cleared when the page is left, the tab closes, or you click “Clear all data.” This is application-level clearing, not a guarantee of forensic erasure from your device.
PDF, HTML, and text reports are generated locally. You choose their contents and can preview them before saving. Downloads, copied text, and files you share remain wherever you put them. Sharing uses your own device or email app after your confirmation; Poo does not send report emails. A later scan-summary JSON contains counts and the scan date only—not messages, names, titles, or IDs.
3. Browser storage, cookies, and tracking
The app can save a light/dark preference and a signed access pass in localStorage. It does not put export contents in localStorage, IndexedDB, or cookies. Its service worker caches application assets for offline use, not your export or reports. There are no Poo analytics, advertising trackers, marketing cookies, remote error-reporting scripts, or runtime third-party script libraries. Clear all data clears scan contents; it leaves your theme, paid pass, and app cache available. To remove those too, clear this site’s storage in your browser. Save your recovery pass first.
4. Optional AI insights: a separate choice
Only after you preview and agree, Poo sends a minimized summary of broad topic labels and conversation counts through Poo’s Anthropic API account. Sensitive topic counts require a separate choice. The allowed payload excludes message text, snippets, personal names, contact details, dates, conversation titles, and account or conversation IDs. We do not send your chatbot login or history. This summary leaves your device. Minimization reduces identification risk; it does not guarantee anonymity or prevent conclusions from being drawn from the counts.
Anthropic receives the summary and returns an interpretation. Poo processes it in memory without intentionally saving its text. Our API rate limiter temporarily uses a salted network-address hash in memory, without writing the address or summary to application logs. Anthropic’s ordinary API retention is up to 30 days, with stated safety and legal exceptions or different agreed arrangements. See Anthropic’s commercial data-retention policy ↗. Local scanning is available without choosing this feature.
5. Hosting and ordinary web traffic
Opening the website sends ordinary requests to Netlify and Railway. Hosting providers process network and request information such as IP addresses, requested pages, timestamps, and browser information to deliver and secure the site. We do not add application access logs containing request URLs, bodies, exports, or chat text. Providers may maintain their own operational records under their policies. “Nothing uploaded” describes local export scanning; it does not mean the hosted website has no ordinary network traffic. See Netlify privacy ↗ and Railway privacy ↗.
6. Payments and support
Stripe handles checkout and retains payment and transaction information. Card details are not entered into Poo’s scanner. Poo’s billing server verifies the transaction with Stripe and issues a signed pass containing product, version, issue time, and expiry—not your name, email, or export. We do not maintain a Poo customer-account or export database. Payment and support records may be retained for receipts, refund handling, accounting, fraud prevention, and legal obligations. See Stripe privacy ↗.
If you contact us, the message, contact details, and relevant receipt or error information will be handled by us and our email provider to respond. Keep personal chat content and exports out of support messages. We retain support correspondence only as needed to resolve the issue and meet applicable recordkeeping obligations. The contact page states whether the support mailbox is active.
7. External links and sharing
Social-account links, sharing buttons, and help links make no automatic requests to those services. Clicking them opens the service you choose; it then receives ordinary traffic and anything you elect to share. Opening a conversation sends its ID to the original chatbot service. Poo does not publish your audit or sell it to advertisers. We do not sell personal information or share it for cross-context behavioral advertising.
8. Choices, requests, and safeguards
You can decline optional AI processing, avoid checkout or sharing, clear the scan, remove site storage, or use the cached scanner offline. If we hold payment or support information about you, contact support@poo.ai to request access, correction, or deletion where applicable. We may need proportionate verification and may retain information we are legally required to keep. We cannot delete data from your original AI or social-media account; use that provider’s controls. Do not send your export as identity verification.
We use HTTPS, restrictive security headers, separate payment and AI routes, and limited server logging. No system is guaranteed secure. Poo is for adults 18 and over, and is not directed to children. If you believe a child has sent us personal information, contact us so we can address it.
9. Updates
This page will show the latest date and explain material changes. We will not silently broaden permission to use information previously collected. See our FAQ, source and runnable privacy checks ↗, or contact us.
Poo · Magichaus
Contact Poo.
Questions about your scan, access pass, a refund, or privacy? Here is how to reach us.
Effective and last updated: October 13, 2026
Support
The support mailbox is being activated. Email delivery has not been confirmed yet. Until then, use the postal contact below. Paid checkout is still in test mode.
For a scan issue, describe the error, your browser, and whether you chose a ZIP or JSON file. For a payment issue, include the receipt or payment reference. Never send your export, chat text, full payment-card details, or recovery pass.
Read the FAQ → · Purchase and refund terms → · Privacy policy →
Operator and postal contact
Magichaus149 Beech Hill Road
Deer Isle, ME 04627
United States
Privacy requests
Use the support address or postal address above. Tell us the nature of your request and the payment or support record concerned. We do not hold your exported conversations, and cannot delete records from another service. Do not attach your history.
Find us on social media
Our official account links will be published here once they are ready. Until then, no account is listed as an official Poo account.