What your screenshots say about your year

Most people’s camera roll has two libraries inside it. There’s the one they think of as their photos — trips, people, food — and there’s a much larger, mostly invisible one made of screenshots. The second one is usually the more honest record.

Screenshots are a to-do list you forgot to write

Nobody takes a screenshot for the memory. You take one because something needs to be dealt with later: an address, a booking code, a recipe, a size chart, a message worth keeping, a bug to report. It’s the most utilitarian thing on your phone.

Which is exactly what makes it interesting in aggregate. Photos record what you chose to commemorate. Screenshots record what you were actually doing. A cluster of apartment listings in March, flight confirmations in June, and job postings in September is a year narrated by its logistics.

The volume is higher than you think

Screenshots accumulate without any of the friction that limits real photos. You don’t frame them, you don’t review them, and you almost never delete them. For a lot of people they end up being a double-digit percentage of the entire library, which is why any honest analysis of a photo library has to treat them as their own category rather than lumping them in with photos.

That separation matters practically, too. Left uncategorized, a few thousand screenshots will distort every statistic about your year — your “most active day” becomes the afternoon you screenshotted forty product pages.

Reading them requires reading text

Sorting a screenshot by what it is means reading the words in it. A boarding pass, a bank statement and a conversation are visually similar and semantically nothing alike. Distinguishing them needs optical character recognition, not just image classification.

This is where it gets sensitive. Your screenshots contain more directly personal information than your photos do — account balances, addresses, medical results, private conversations. Text recognition over that set is a meaningfully different operation than recognizing that a picture contains a dog.

Apple’s Vision framework can do this recognition entirely on the device. That’s not a minor implementation detail. It’s the difference between your bank balance being read by a chip in your pocket and being uploaded to a company’s servers to be read there.

A useful exercise

You don’t need an app for the basic version of this. Open Photos, search for “screenshots,” and scroll through by month. Look for the clusters. The dense weeks are almost always periods when you were figuring something out — moving, planning, buying, applying, or troubleshooting.

Then consider deleting the ones whose task is done. Most of them are complete. The booking happened, the address was used, the recipe was cooked or wasn’t. It’s the easiest storage you’ll ever recover, and it makes everything else in your library easier to see.

How WrappedMe handles them

WrappedMe treats screenshots as their own category rather than letting them skew your photo statistics, and it uses on-device text recognition to sort them by what they actually are. Nothing is uploaded — no server ever sees the text in your screenshots, because the recognition happens on your iPhone and stays there.

If you want the technical version of that claim, what on-device analysis really means goes through the frameworks involved. If you’re wondering why an app like this needs your whole library, we explained that too.