AI News
04 Oct 2026
Read 10 min
How to audit your ChatGPT data and stop hidden profiling
how to audit your ChatGPT data to reveal hidden profiling and regain control of your privacy quickly.
Want a quick checkup on what an AI has learned about you? Here’s how to audit your ChatGPT data in minutes: export your chat history, upload it to a privacy tool, and read a clear report of what the model infers about your location, job, interests, and more—then lock down your settings to stop hidden profiling.
AI remembers more than your last prompt. Across months of chats, it can infer where you live, what you do, and what you care about—even if you never typed those facts outright. This guide shows how to audit your ChatGPT data using a free analysis from Lumo AI Paper Trail, what to look for in the results, and the settings to change so your future chats reveal less.
A ChatGPT or Claude account with past chats
Time to request and receive your export (often hours to days)
The Lumo AI Paper Trail webpage (audits are free)
Open ChatGPT. Go to Settings.
Find Data controls (menu names can vary over time).
Select Export data, then confirm.
Watch your email for a ZIP download link. Save it.
Open Claude. Go to Settings or Account.
Choose Data export, then request your archive.
When the email arrives, download the ZIP file.
Open Lumo AI Paper Trail and upload your ZIP. Lumo says it encrypts the upload in transit and processes it without storing the archive.
Wait a minute or two. You’ll get a dashboard plus a shareable “data card” with a privacy score and category scores (like location, work, interests).
Open the full report. Look for:
Personal info summaries: location hints, education, job signals, interests, age range, and political leanings
Most revealing chats: specific conversations that exposed the most data
Inferences: what the model thinks is true about you, and why
Red flags: where inferred data could enable profiling or targeting
Note: Run the audit again any time. Results can differ slightly, which also indicates Lumo isn’t saving your report.
If you were searching for how to audit your ChatGPT data with the least friction, this is it: export, upload, review, then adjust your privacy settings.
Turn off training where available: look for a Chat history & training toggle and disable it.
Delete old chats you no longer need. Clear sensitive threads first.
Review Custom Instructions. Remove personal details (job title, company, city).
Limit connected apps, plugins, and browsing tools that may send data to third parties.
Open settings and opt out of training if the option is present for your plan.
Delete past conversations that reveal personal info.
Remove files containing sensitive data from your workspace.
Use separate profiles for work, personal, and research chats.
Avoid auto-filling addresses or contact info into prompts.
Consider privacy tools (containers, trackers blocked) to reduce cross-site signals.
Speak in scenarios, not specifics: “a mid-sized city” instead of your town; “a retail company” instead of your employer.
Use synthetic or masked data for examples (fake names, dummy emails, altered dates).
Remove hidden metadata from files before upload (PDFs, images, docs).
Ask yourself: “What could this reveal?” If the risk is high, don’t include it—or use offline tools.
Split sensitive tasks: do research in the AI, but apply personal details offline.
Quarterly: export your data and re-run Lumo. Track how your privacy score changes.
Monthly: delete chats you no longer need. Clear uploads and custom instructions.
Per project: create a fresh chat with minimal context. End and delete when done.
For sensitive topics: consider a privacy-focused assistant. Lumo’s Paper Trail audit is free, and Lumo’s paid plans add larger uploads and fewer limits (you don’t need a subscription to run the audit).
Mentions of neighborhoods, transit, weather, or store names can pin your region.
Posting times can hint at your time zone and work schedule.
Tools, frameworks, and jargon can reveal your field and seniority.
Alma mater or course names can leak your education history.
Hobbies, teams, media, and medical terms may indicate age range or health interests.
Policy topics and word choice can imply political leanings.
If a guess is wrong, it still matters: the model may act on that inference. Correct it by deleting revealing chats and changing how you prompt going forward.
Work vs. personal: separate logins reduce cross-contamination of topics and contacts.
Research vs. private life: keep sensitive health, finance, and family topics out of your main account.
Teaming: for shared prompts, use a team or enterprise plan with clear data-use terms.
Conclusion: If you care about privacy, make this part of your routine. Learn how to audit your ChatGPT data, cut back what each prompt reveals, and adjust your settings so hidden profiling shrinks over time. A few minutes every month keeps your digital self smaller, safer, and easier to manage.
Why chatbots know more than you think
Large models connect dots. Ask about a local café, a city event, or a time zone, and the system can guess your region. Discuss tools you use, and it can infer your role. Share pop culture tastes, medicine names, or sports teams, and it can sketch your age range or health interests. That profile improves responses, but it also expands the surface for targeting, profiling, or breaches if settings are loose.How to audit your ChatGPT data with Lumo AI Paper Trail
What you need
Export from ChatGPT
Export from Claude
Upload to Lumo and read the report
What to fix after your audit
In ChatGPT
In Claude
In your browser and devices
Reduce what each prompt reveals
Set a simple audit routine
What your report is telling you
Location and routine
Work and education
Interests and identity
When to use different accounts
For more news: Click Here
FAQ
Q: What’s the fastest way to find out what ChatGPT knows about me?
A: Export your ChatGPT chat history, upload the ZIP file to a privacy tool like Lumo AI Paper Trail, and read the generated report to see inferred details and red flags. Learning how to audit your ChatGPT data this way reveals location hints, job signals, interests, and other profiling risk factors.
Q: How do I export my ChatGPT data for an audit?
A: Open ChatGPT, go to Settings, find Data controls (menu names can vary), select Export data, and confirm your request, then watch your email for a ZIP download link. The export process can take hours to days, so save the ZIP when it arrives.
Q: Can I use the same export process with Claude and how long will it take?
A: Yes, Claude supports a similar flow: open Settings or Account, choose Data export, request your archive, and download the ZIP from the email that arrives. The guide notes exports often take hours to days before the archive is delivered.
Q: What should I expect when I upload my export to Lumo AI Paper Trail?
A: Lumo encrypts the upload in transit and processes the archive without storing it, then returns a dashboard and a shareable data card with a privacy score and category sub-scores. The full report shows personal-info summaries, the most revealing conversations, model inferences, and flagged items that could enable profiling.
Q: Which specific items in the Lumo report should I review first?
A: Check the personal-info summaries (location, education, profession, interests, age range, political leanings), the most revealing chats, and the inferences explaining why the model guessed those things. Also review any red flags that indicate where inferred data could enable targeting or profiling.
Q: What privacy settings should I change in ChatGPT after the audit?
A: Turn off options that allow your inputs to be used for training (look for a Chat history & training toggle), delete old conversations you no longer need, and remove personal details from Custom Instructions. Also limit connected apps, plugins, and browsing tools that may send data to third parties.
Q: How can I reduce what each prompt reveals about me going forward?
A: Prefer scenarios instead of specifics (for example, “a mid-sized city” instead of your town), use synthetic or masked data for examples, and strip metadata from files before uploading. Split sensitive tasks by researching in the AI but applying personal details offline, and avoid autofill of addresses or contact info into prompts.
Q: How often should I re-run audits and clean up my ChatGPT data?
A: Set a routine like quarterly exports and Lumo re-runs to track your privacy score, monthly deletion of unneeded chats, and creating fresh minimal-context chats per project. For sensitive topics, run audits sooner and consider separating accounts for work, personal, and research to reduce cross-contamination.
Contents
