MarginaGet Margina
記事

公開

What Happens to Your Highlights When a Reading App Shuts Down

A reading app shutting down is rarely about the books — most of those can be found again. It's the highlights and notes, quietly stored in one company's database, that tend to disappear without a trace. Here's what actually happens, and what a genuinely portable highlight would look like.

A reading or read-it-later app rarely announces its own shutdown as a loss of your books. Usually there's still a way to get the files back, or the files were never really the app's to begin with. What the shutdown notice threatens is quieter: the highlights, the margin notes, the passages you underlined because they mattered. Those live only inside the app's own database, keyed to an account, and when the app goes, the notice you get — if you get one at all — is a countdown to export them or lose them. This is what actually happens on that day, why it keeps happening, and what a highlight would have to look like to survive it.

The thought

A highlight is not really about the book. It's evidence of the reading you actually did — closer to a diary entry than to the text it's attached to. You could lose access to the book itself and, in principle, find another copy. Lose the highlight, and there's no other copy; it only ever existed in one place. Most reading apps quietly assume you'll never need to leave, so that record lives entirely inside their database, tied to their account system, in a format only their software can open. Framed that way, owning your books as files, not apps was always half the argument. The other half is what happens to the marks you made on those books once the app that held them disappears — which turns out to be less a reading-app problem than a data-portability problem wearing a reading-app costume.

The longer version

What actually happens the day an app shuts down

The best case is a fixed export window. When the read-it-later app Pocket shut down in 2025, users had until October 8 to export their saved articles, lists, and highlights before the data was gone for good. A handful of competitors — Instapaper, Readwise Reader, Wallabag — offered direct imports, which softened the landing for anyone who acted in time. That "in time" is the catch: a deadline only helps the people who know it exists. The worse case is no export path at all, just a login that stops working one day. There's rarely a warning for that kind, because there's seldom anyone left at the company to send one.

Owning the book file isn't the same as owning your notes of it

In 2009, Amazon remotely deleted copies of George Orwell's 1984 from customers' Kindles after a licensing dispute — an irony that wasn't lost on anyone. The Electronic Frontier Foundation's response to the incident went further than objecting to the deletion itself: among its proposed protections was the ability for readers to back up their own annotations, treated as a right distinct from owning the book file. That distinction holds. A platform that can revoke a book under a licensing clause has even less reason to preserve your notes on it — and the same logic that keeps a book from opening, explained in why some books won't open at all, tends to keep its annotation layer just as closed.

Why most highlights live in a proprietary database, not an open file

Digital reading platforms don't stay accessible on their own. The Internet Archive's Brewster Kahle has argued that digital books require continuous, active maintenance to remain readable at all — format conversion, hardware migration, the unglamorous work of moving data forward every few years. Most companies won't fund that indefinitely, and annotations sit at the bottom of the priority list when they do: low value to the vendor, high value to the reader, rarely worth building an export tool around until users start asking loudly.

The standards trying to fix this

There's a real effort underway to change that. The W3C's EPUB Annotations 1.0 defines a shared, JSON-based format for notes, highlights, and bookmarks — built on the existing Web Annotation Data Model, so an annotation exported from one reading app could, in principle, be understood by another. It's a sound idea with an honest caveat attached: broad adoption by mainstream reading apps is still years away, and a standard nobody implements doesn't help the person facing a shutdown notice this week.

The broader pattern

None of this is new territory for consumer protection generally — it's just new to reading. Data portability, the right to export and transfer your own data in a structured, reusable format, is already law in the EU under GDPR, was folded into California's CCPA, and Utah becomes the first U.S. state to require it of social platforms in 2026. Reading-app annotations are exactly the kind of personal data those laws were written to cover. They just haven't been named specifically yet, so most reading apps aren't held to it.

A useful corollary

The practical version of all this is simple: check for an export option, and check what format it produces, before you commit years of notes to an app. The mechanics of doing that today — app by app — are covered in export your highlights before you need to. The format worth aiming for is a plain-text format: no proprietary database to lose access to, readable in anything from a basic text editor to whatever comes after the app that made it.

That's the design Margina starts from rather than arrives at. It has no account and no server, so your highlights aren't rows in a database somewhere — they're stored in the files you already have, backed up to your own iCloud or Drive if you choose to. The measure of a good reading app was never what it lets you do while you're using it. It's what it lets you keep when you leave.