What is actually in a Google Takeout export

A Takeout export is a zip full of folders named after products, with everything interesting buried in JSON files nobody expects you to open. The parts that surprise people are the location history (not a map, a raw list of coordinates with timestamps and an accuracy figure, often millions of them) and the ads folder, which carries the interests that have been inferred about you. TAKEOUT reads the zip in your browser and lays it out by product, without unpacking it and without sending it anywhere. That last part is not a preference here: a Takeout export is the most revealing single file most people will ever possess.

What is in it, roughly in order of how surprising people find it

• Where you have been. Not a picture of a map. A list: latitude, longitude, a timestamp to the millisecond, and how accurate the fix was. Years of it. Every commute, every hospital visit, every night not spent at home, at a resolution good enough to say which building.
• What has been inferred about you. The ads folder carries the settings and the categories you have been sorted into. It is written down in plain language and most people have never seen it.
• Everything you searched and watched, with a timestamp on each. My Activity is one entry per action, going back as far as the account.
• Your browsing history and bookmarks, if Chrome was signed in.
• Your mail, as one mbox file: the entire mailbox in a single document, which is why it is frequently the largest thing in the export by an order of magnitude.

Location history has five different shapes

This is the part that makes "just look at your location data" bad advice. Google has restructured it repeatedly, so where your history lives depends on when you exported and from what:

An export can contain any combination of these. A viewer that only knows one of them will tell you there is no location data in a file that is full of it, which is the commonest way people conclude wrongly that they were never tracked.

Why you cannot just open it

Because of the size, and the size is not an inconvenience; it is a wall. A Records.json with three million points is hundreds of megabytes of a single JSON array. A text editor will not open it. Nor will a browser console: JavaScript engines cannot hold a string that long, let alone parse one, so the obvious approach of reading the file and calling a JSON parser on it fails before it starts.

That is the gap every "Takeout viewer" fills, and it is why they all begin the same way: upload the zip.

Which is the one thing not to do

Think about what you would be uploading. Every place you have been for several years. Everything you have searched, including the things you searched at three in the morning. Your whole mailbox. The categories an advertising system sorted you into.

There is no version of that which is worth the convenience, and there is no need for it, because none of the work requires a server.

How it is read without leaving the machine

The awkward parts are solvable and TAKEOUT solves them rather than routing round them:

• The zip is never unpacked into memory. A streaming reader hands each entry over as a run of chunks, and each chunk goes straight into a parser for its format.
• Large JSON is split before it is parsed. A byte-level splitter finds the boundaries of the elements in the top-level array and parses each element on its own, so three million points cost the memory of three million small rows rather than the memory of the file.
• The mbox is read line by line, keeping only the headers that matter: who, when, the subject and the labels. To read the messages themselves, with their attachments, POSTBAG opens the same mbox straight from the zip.
• All of it runs in a Web Worker, so a multi-gigabyte export does not lock the page up, and the page itself only ever asks for one screen of one table at a time.
• The map fetches no tiles. Points are drawn on grid lines instead, because requesting map tiles for your own coordinates would hand a tile server the very thing the page exists to keep off the network.

What it looks like

A small export read in the browser, as an example of the shape rather than of anybody's real data:

Every table is searchable, and exports as CSV or JSON. The location tracks come out as GPX, which is what a mapping tool expects.

Worth doing while you have it open

1. Read the ads interests. It takes a minute and it is the single most informative minute in the export.
2. Find out whether location history was on. Many people are certain it was off. The file settles it.
3. Check the date range against when you think you started. Exports frequently reach back further than people expect.
4. Remember what an export is not. Downloading your data does not delete it, and deleting the zip deletes nothing at the other end. Turning the collection off, and deleting the history, are separate actions in separate settings.
5. Keep the zip somewhere sensible. Having made a single file containing your whole life, do not leave it in Downloads. STRONGBOX encrypts it in the browser if you want to keep it.

Alongside: TRAIL reads the GPX that comes out of the location tracks, and does a photo show where it was taken is the same question for a single picture.