Exit one — take it back
Every platform has an export button. You've probably never seen it.
Companion to the essay — a living index of legally mandated export buttons: where they're buried, what you click, what you get back.
GDPR, CCPA, and kin forced these buttons into existence. The data is yours — you're collecting it.
A copy leaves everything in place. Deletion is a separate step per platform.
Inputs — posts, searches, plays. Rarely the profile they built and sell against. Each guide flags gaps.
Big-tech data troves
Largest personal archives outside government — clicks, photos, locations, purchases. 3 platformsGoogle (Takeout)
By volume, probably the largest single archive about you outside a government. A Google account links every Gmail message, every Drive file and Doc revision, every Photo (with face-grouping metadata), every Calendar event, every YouTube video watched or liked, every Search query, every Maps location point if Location History is on, every Android app, every Pay transaction, every Chrome bookmark and (with Sync) browsing history, Keep notes, Fit step counts, Voice calls/SMS, and Nest/Home device commands. Skim myactivity.google.com once and you'll understand why Takeout matters.
- Go to takeout.google.com and sign in.
- By default all products are selected. Click Deselect all to pick specific ones, or leave it to grab everything.
- Per product, use the dropdown to choose subsets and formats (
.mboxfor Gmail,.json/.htmlfor activity, original quality for Photos). - Scroll to the bottom and click Next step.
- Choose delivery (email link, Drive, Dropbox, OneDrive, Box), file type (
.zipmax 50 GB/file, or.tgz), and frequency (one-time, or every 2 months for a year). - Click Create export.
- Wait — a small account may be ready in an hour; a heavy one (years of Photos/Gmail/YouTube) can take several days. Archives stay available for 7 days.
Advanced Protection accounts have exports scheduled two days out for safety. Downloading does not delete anything — to remove data, use myaccount.google.com/data-and-privacy.
.mbox, Calendar .ics, Contacts .vcf, Photos JPG/HEIC + JSON sidecars, Drive originals, YouTube/Maps/Chrome as JSON or HTML/KML.adssettings.google.com, not bundled), cross-account graph signals, and deleted-but-not-purged data.Takeout currently lists 80+ products. You only see the ones you've used — but for most people that's still 40–60 separate datasets in a single download.
Stash it. Save somewhere safe and offline. If it's bigger than a few GB, this is your moment to plug in an external drive.
Apple (iCloud + Health)
Two distinct slices: the Apple Account / iCloud server-side archive (contacts, calendars, mail, notes, reminders, bookmarks, photos, videos, documents, App Store/iTunes/Books history, AppleCare, Maps activity) and the on-device Health archive — if you've worn an Apple Watch, one of the most detailed personal records you'll ever encounter: heart rate every few seconds, sleep stages, every workout with route, ECGs, blood-oxygen, cycle tracking, mindfulness, falls, and any third-party app data written to Health. Apple's posture is stricter than most — Health is E2E-encrypted and not used for ads — but the trail still exists, and you should own a copy.
Two separate exports, because Apple holds the data in two places.
- Go to privacy.apple.com and sign in.
- Select Request a copy of your data.
- Pick categories (Apple Account details, iCloud Mail, Drive, Photos, Contacts, Calendars, Notes, Reminders, Bookmarks, purchase history, AppleCare, marketing).
- Click Continue.
- Apple takes up to 7 days and emails you; you then have 14 days to download.
- Open the Health app on your iPhone.
- Tap your profile picture in the upper-right.
- Scroll to the bottom and tap Export All Health Data.
- Wait — years of Apple Watch data can take 5–15 minutes.
- From the share sheet, save
export.zipto Files, AirDrop it to a Mac, or send it somewhere secure.
.vcf, .ics, .eml, .html, .json, .csv, .pdf + media. Health: export.zip with export.xml, export_cda.xml, and .gpx workout routes.Apple splits the export by service; HealthKit alone contains dozens of distinct data types.
Stash it. Health's export.xml is an extraordinary file — open it once and look at how many years of yourself live in it.
Amazon
One of the longest-running purchase records of your life: every order since you opened the account, with item, price, date, and where it shipped — a decade-plus of revealed preference. Beyond orders: Kindle (books owned, reading progress, highlights and notes), Alexa (voice requests and, unless turned off, recordings), Prime Video history, Audible listens, plus browsing history, search history, wish lists, returns, reviews, addresses, and payment methods.
- Go to Account → Data Privacy → Request Your Information (or the data requests page).
- Choose a category. For the essentials, request Your Orders and Kindle; or pick Request All Your Data.
- Submit and confirm via the email Amazon sends.
- Wait — Amazon may take up to 30 days, though common categories often arrive in a few days via download link.
Faster alternative: if your account still has the legacy Order History Reports tool, it generates an instant CSV of purchases by date range — check Account for "Download order reports."
Stash it. Order history and Kindle highlights are the two highest-signal pieces — money (what you actually spend on) and mind (which ideas you actually marked).
AI walled gardens
Conversations, prompts, memory entries. Youngest category — export tools still raw. 6 platformsAnthropic Claude
Every conversation — file uploads, attached images, project documents, titles — sits on Anthropic's servers tied to your account. Deleted chats clear from history immediately and from backend storage within 30 days. If "help improve Claude" is on, de-identified copies of your chats can live in training pipelines for up to 5 years. The 2026 memory features (search past chats, memory across sessions) write entries about you that are stored separately from transcripts — and as of now, memory entries are not included in the export.
- Sign in to claude.ai on web or Claude Desktop. iOS and Android can't initiate exports.
- Click your initials in the lower-left corner.
- Select Settings.
- Open the Privacy section.
- Click Export data.
- Wait for Anthropic's email with your download link — it expires in 24 hours; request a new export if you miss it.
On Team or Enterprise plans, only the Primary Owner can export — individual seats can't pull their own data alone.
Stash it. Save the ZIP somewhere safe and offline. This is your half of a conversation that, until now, only they kept.
OpenAI ChatGPT
Two distinct piles: the conversations (every message, uploaded file, attached image, regenerated response) and memory — facts ChatGPT was told to remember or decided to on its own. The result is a profile that can include your job, your kids' names, your medical history, your politics, your writing style. Deleting a conversation removes it within 30 days but does not delete saved memories — those clear separately under Settings → Personalization → Memory. Unless you've turned off "improve the model for everyone," your chats may train future models.
- Sign in to chatgpt.com.
- Click your profile icon in the bottom-left.
- Select Settings.
- Open Data Controls.
- Click Export under "Export Data."
- Confirm the request.
- Watch your inbox (and spam) for a mail from
noreply@tm.openai.com— usually minutes to hours, up to 7 days. The link expires 24 hours after delivery.
Available for Free, Plus, and Pro. Business and Enterprise users go through their workspace admin.
conversations.json, user.json, message_feedback.json, model_comparisons.json, and a human-readable chat.html.Stash it. Save the ZIP offline, and screenshot your Memory page before anything else — it's the part the export leaves behind.
Google Gemini
Your Gemini conversations are stored as part of your Google Account activity — the same graph that holds your search history, YouTube, location, and Maps. Prompts, responses, uploads, and generated media are saved by default. Conversations selected for human review are retained separately for up to 3 years, even after you delete them. Because Gemini is woven into Workspace, the questions you ask about your inbox or calendar are themselves logged — and if personalization is on, your AI conversations are informed by Google's existing dossier.
Footgun: do not select "Gemini" in Takeout — that only exports custom Gems, not your chats. Your conversations live under "My Activity."
- Go to takeout.google.com and sign in.
- Click Deselect all at the top of the product list.
- Scroll to My Activity and check it.
- Click All activity data included, then Deselect all in the popup.
- Check only Gemini Apps. Click OK.
- Click Next step.
- Choose delivery, file type (
.zipor.tgz), and one-time vs recurring. - Click Create export. Usually a few hours; Google reserves the right to take days.
Stash it. While you're in Takeout, grab the rest of your Google data too — see the Google Takeout guide below.
Microsoft Copilot
Microsoft retains the last 18 months of Copilot conversation history by default, tied to your account — consumer Copilot, the Windows/web apps, and (separately) Microsoft 365 Copilot inside Word, Excel, Outlook, and Teams. For 365 users, Copilot is grounded against your Microsoft Graph (mail, OneDrive, SharePoint, calendar, Teams chats), so prompts carry implicit references to that whole corpus. Then there's Recall on Copilot+ PCs — a continuously-updating local visual archive of your screen, encrypted on-device.
You export through Microsoft's Privacy Dashboard, not from inside Copilot.
- Go to account.microsoft.com/privacy and sign in.
- In the left nav, click Empower your productivity → Copilot.
- Find Your Copilot app activity history.
- Under Copilot apps, click Export all activity history.
- The CSV downloads to your device.
365 Copilot through work/school adds "Export all Copilot in Microsoft 365 apps activity history" and "Export all Copilot Chat activity history" — pull all three. Recall snapshots live locally; open the Recall app on your Copilot+ PC (no cloud export).
Stash it. Save the CSV offline. If you live in Microsoft 365, also pull your OneDrive, Outlook, and Teams data through the account export tools.
Apple Intelligence
Structurally different from the rest. Most requests run on-device; heavier ones route through Private Cloud Compute — encrypted Apple-silicon servers where, per Apple's published architecture, request data is processed only to fulfill the request and not retained. The catch: when Apple Intelligence falls back to ChatGPT, those requests leave the Apple envelope (anonymized by default, no-retention by contract). So there's largely no server-side "chat history" to export — what exists is the on-device Apple Intelligence Report and your full Apple Account data.
- iOS/iPadOS: Settings → Privacy & Security → Apple Intelligence Report.
- macOS: System Settings → Privacy & Security → Apple Intelligence Report.
- Tap Export Activity to save a snapshot of recent PCC requests.
- Go to privacy.apple.com and sign in.
- Click Request a copy of your data.
- Select categories (Apple Account details, iCloud, app usage, purchases, marketing, etc.).
- Click Continue.
- Apple typically fulfills within 7 days and emails you; you then have 14 days to download.
EU, UK, and Japan users can schedule recurring downloads for some categories.
.vcf, .ics, .html, .eml, .json, .csv, .pdf.Stash it. Save the archive offline. If you use iCloud heavily, also see the Apple iCloud + Health guide for the device-side Health export.
Perplexity
Perplexity stores every thread indefinitely — not just the question and answer, but the prompts, AI responses, retrieved sources, every follow-up turn, and any files or images you uploaded. Comet browser and Spaces carry their own context; Spaces retain the documents and instructions you've attached. Like other AI platforms, your queries improve the product unless you've turned off the AI-training toggle in Settings → Privacy.
- Sign in at perplexity.ai.
- Click your avatar in the bottom-left to open Settings.
- Find the Data Controls section.
- Click Download My Data (sometimes "Export Data").
- Confirm. Your full thread history is delivered as JSON.
No date-range option — it's the whole account at once. Perplexity warns the JSON can contain sensitive data you pasted into prompts (phone numbers, API keys, addresses); sanitize before re-uploading anywhere.
Stash it. Open it once and search for anything you wouldn't want read aloud — Perplexity's own guidance flags that exports often contain pasted secrets.
Media & tools
Create, listen, move — often the most personally meaningful archives. 4 platformsSpotify
A track-by-track, second-resolution record of your listening life: every song since you signed up, with timestamps, duration listened, platform, skip and shuffle flags, and what context it came from. That's the Extended Streaming History — for most users it goes back years. Beyond playback: every playlist, saved track, podcast and how far you got, every search, every artist you follow, every device, every payment, and your inferred taste profile. There are three separate packages — most people only download the first, which is just the last year.
- Go to spotify.com/account/privacy and sign in.
- Scroll to Download your data.
- You'll see three checkboxes: Account data (last 1 year only — skip if you want everything), Extended streaming history (every track ever — check this), and Technical log information.
- Check the boxes you want and click Request data.
- Spotify emails to confirm — click the confirm link.
- Wait. Spotify reserves up to 30 days, but extended history typically arrives in 1–5 days.
- A second email brings the download link when it's ready.
Streaming_History_Audio_*.json (every play, timestamps, ms played, platform, skip data), Streaming_History_Video_*.json, plus playlists, library, searches, follows, payments, taste-profile metadata.Stash it. The Extended Streaming History JSON is one of the most surprisingly meaningful files here — open it and look at what you were listening to during a hard week years ago.
Notion
If you've used Notion as a "second brain," it knows everything you've trusted to one: notes, journals, project plans, financial records, contacts, decision logs, bookmarks, recipes. Notion stores all pages, databases, comments, page history (edit-by-edit revisions on most plans), shared spaces, and uploaded files. With Notion AI, the questions you've asked about your own workspace — and what Notion AI read to answer them — also live in your account.
For the cleanest re-importable archive of an entire workspace:
- Sign in to notion.so on a desktop browser (mobile doesn't expose this).
- Click Settings in the left sidebar.
- Go to General.
- Scroll to Export all workspace content.
- Choose format: Markdown & CSV (best for re-import — pages become
.md, databases.csv; relations export as UUIDs), HTML (best for offline browsing), or PDF (Business/Enterprise only). - Click Export. Notion emails you when ready (link expires in 7 days).
Single page: ••• menu → Export → optionally include subpages. Very large workspaces: export in batches (one teamspace at a time).
.md per page, nested to match your sidebar), every database as CSV, embedded files in subfolders, comments inline as blockquotes.Stash it. If you've used Notion seriously, this is one of the highest-value exports on the page — your written record of your own thinking.
Strava
A GPS-precision record of where you've moved your body for as long as you've used it. Every run, ride, swim, hike, or ski includes the full GPS track (often 1-second resolution), heart rate, cadence, power, temperature, elevation, splits, and segment matches — years of which map your routine, neighborhood, fitness, and (without privacy zones) your home and workplace. Beyond activities: kudos history, comments, clubs, route library, gear inventory, goals, leaderboard rankings, and your follower/following graph. The 2018 heatmap incident is a useful reminder of how rich this data is.
The bulk export is on the website only, not the mobile app.
- Sign in to strava.com on a computer.
- Hover your profile picture (top-right) → Settings.
- In the left sidebar click My Account.
- Scroll to Download or Delete Your Account.
- Click Get Started under "Download Request (optional)."
- Click Request your archive.
- Strava emails a download link, typically within 2–4 hours.
Single activity in GPX: open it → ••• menu → Export GPX.
activities.csv (summary metrics for everything), profile.json, plus folders for clubs, gear, and routes.Stash it. The activities.csv alone is a remarkable training log — years of your own effort in one file.
Substack
If you write on Substack, the platform holds your full corpus — every post (published and draft), every Note, comment, podcast episode, plus your full subscriber list with email addresses, paid status, signup source, and engagement metrics. If you only read, it still holds your subscriptions, reading history (which posts you opened, scroll depth, time spent), Notes activity, comments, and saved posts. The reader-side data is harder to export — the official tooling is built for writers — but Substack's privacy page lets you request your account data under privacy law for the rest.
- Sign in to your publication's dashboard.
- Click Settings in the left navigation.
- Click Exports.
- Click New export.
- Substack generates a ZIP and emails you when it's ready.
- From the dashboard, go to Subscribers.
- Scroll to the All subscribers table.
- Click the ••• menu above the list and select Export.
- Choose All columns (last-opened + post-view data) or visible columns only.
- CSV downloads to your device.
As of 2026, publisher analytics also export from the Stats page — pick a date range and download the CSV.
Stash it. Save both the post archive and the subscriber CSV offline. If you ever want to leave — and people increasingly do — these two files are what make that possible.
You've got the file. Now make it pay.
Export is exit one (self-custody). Exit two is live: charge for what you already give away while you wait on AI.
Social graphs
Posts, connections, inferences — including off-platform tracking. 3 platformsMeta (Facebook + Instagram) format ZIP · HTML or JSONwait minutes–days
The data graph spans Facebook, Instagram, Messenger, WhatsApp, and Threads. The export covers every post and comment, every Messenger/Instagram DM, every photo and video, every Reel, every Story (even ephemeral ones — Meta keeps copies), every reaction, group, event, and search, every ad clicked, every login session and IP. The surprising part is off-Facebook activity — data Meta receives from third-party sites via the Meta Pixel, often hundreds of businesses long — plus the ads-interests profile of inferred topics used for targeting.
Both Facebook and Instagram now go through the unified Accounts Center.
Meta's "Download Your Information" is itemized into 100+ toggles across Facebook and Instagram — most people never scroll the full list.
Stash it. Pick JSON if you have to choose — it re-imports cleanly. While you're in Accounts Center, read the Off-Facebook activity list; it's instructive.
X (Twitter) format ZIP · JS/JSON + HTMLwait ~24 hours
Unusually complete: every post (back to your first, including ones you later deleted up to the export date), every reply, repost, quote, Like, Bookmark, DM, photo and video, every list, every Community, every Space, and your full follower/following graph at export time. Less obvious but in there: every device and IP, every email and phone ever attached, the advertisers X believes you engaged with, and the inferred-interests profile. Under current terms, posts may be used to train Grok — though the archive captures what you put on the platform, not what downstream systems did with it.
Mobile follows the same path via your profile picture → Settings and privacy. You can request a new archive every 24 hours.
index.htmland adata/folder of raw files.Stash it. Open the
index.htmlonce just to see your own archive — it's a strangely affecting time machine.LinkedIn format ZIP · CSV + JSONwait 10 min – 24 hours
The most professionally consequential dossier in this list. Every job title, company, education entry, skill, endorsement, recommendation, certification; every connection (with date and how you connected); every message ever sent or received; every post, comment, reaction, article; every job applied to, saved search, and Learning course; every login and device. The non-obvious data: profile views and LinkedIn's inferences about who they were; connection email addresses; the inferred ad-interests profile of your seniority, function, and industry; and every interaction with LinkedIn's AI features.
Desktop only — the export isn't exposed on mobile.
Stash it. Your messages CSV is genuinely useful — years of professional context, finally searchable on your own terms.