Your saved posts are only links in your export

There is a recurring, quietly urgent search: how do I see all the posts I saved on
Instagram. The urgency is real — the in-app saved collection is a private shelf people
fill for years, and some apps and services have appeared around exactly that need. The
answer this post gives is simpler than any tool: the list is already in the export you
can request for free, it has always been in the saved area, and understanding what it
contains takes about a minute — because what it contains is disappointingly small. A
save, in the machine-readable export, is a link. A title, a URL, a timestamp. Not
the post.

That is the honest shape of one of the archive's most sought-after folders, and it
deserves the careful treatment this series gives everything: what the record actually
holds, why the format you picked changes what you see, what happens when the linked
post no longer exists, and what the shelf never had any way of containing.

1. 1minwhat one saved record contains (four fields, honestly).
2. 4minwhy format choice changes this folder, and what "link rot" does to it.
3. ongoing — where saves surface in the app, and the two lists that never exist.

What one save looks like in the file

Under the hood, saved posts arrive as a generic activity record: the parser walks
saved_posts.json, reads each entry's map fields, and produces four things — a
value (the label or link the record carries), an href (the URL, when the record
has one), a timestamp (when you saved it), and a title (often the post's address
itself).¹ That is the whole object. There is no caption field in the JSON
shape, no media reference the parser can rely on, no like count, no comment, and no
thumbnail. If you open the raw file yourself you will see the same thing in its
native form: a small map per entry saying this URL, at that time.

The music shelf is its own file — saved/saved_music.json — and it is the richer
sibling: track and artist details survive there, which is why the app's Features list
advertises "posts you saved plus music tracks you bookmarked, with artist info and
dates" as two data sources rather than one.² A save of a song is a different
record than a save of a post, and the export treats them accordingly.

Notice what kind of list this makes yours: a collection of references, not a set of
actions you took at someone. The app's own taxonomy marks this distinction
deliberately — saved posts and saved music sit in the group labelled "collected," next
to likes and comments which are interactions you directed at people's content.³
The shelf is yours in the private sense: it recorded what you kept, not what you said.

Why the format you picked changes this folder

The request post's format rule has exactly one exception, and this is it. In JSON —
the complete, machine-readable record — a saved post is the reference described above:
link and time, no caption. In HTML — the human-readable rendering — saved posts
carry their captions. The same shelf, two depths: HTML shows you what the post
said; JSON shows you where it lived.⁴

Which makes the choice depend on what you are saving the archive for:

• Rebuilding the shelf as you remember it — captions, context, the reading you

  were doing — favours HTML for this folder, at the cost of the structured detail
  everywhere else in the export.

• Counting, sorting, diffing two exports, or feeding any tool favours JSON, and

  the caption is the price.

• Both is available the way everything in this series recommends: two requests,

  two formats, one archive each.

There is a quiet corollary. If your export arrived as JSON and you are staring at a
wall of bare links wondering whether the captions were dropped by accident — they were
not; that is the shape of the raw record. And if you arrive at the opposite
conclusion, that the HTML shelf looks like the "real" one, remember it is a rendering:
the structure that makes saves countable, comparable and searchable lives in the JSON
file that looks emptier. Neither format is the complete shelf; they are complete in
different directions.

When the link breaks

A reference is only as durable as its target, and the saved list is where link rot is
most personal. Three ways your shelf loses buildings while keeping every address:

1. The post was deleted. You saved it; the author later removed it. Your record

   still points at the old address — and the address now resolves to nothing. The
   eight-things post's first entry is exactly this mechanism operating on the other
   side: a deletion before the snapshot takes the content everywhere, including out of
   every saved list pointing at it. Your export will faithfully contain the link to a
   post that no longer exists.

2. The account changed. A renamed account moves its posts' addresses; a

   deactivated or removed account takes the whole shelf section with it. The record
   does not update itself — it was written when the save happened.

3. The save predates your date range. A request bounded to a window returns the

   saves inside it. Saves outside the window are absent from that export, read like
   "nothing happened there," and come back only when the request covers them.

None of this corrupts the file. The record is doing its job — remembering where
something was — and the world underneath it moved. The honest reading of a saved list
is therefore a list of intentions to revisit, some of which have already outlived
their targets, and no export can tell you in advance which.

What the saved shelf never held

Two absences frame this folder better than any feature description:

• Nobody else's saves, anywhere, ever. Your saved posts are invisible to other

  accounts — the shelf has no audience — and by the same custody wall, every other
  person's saves are invisible to you and absent from your archive. There is no
  "who saved your posts" file, at any scale, in any format, for the same reason there
  is no profile-views file: it would be other people's private collection, and your
  request only reaches your own records. The five-questions and eight-things posts
  own the general case; this folder is its most common specific one, because the
  thing I want is "who saved my post" is the natural follow-up to the thing I got is
  "my saved posts."

• The posts themselves, as content. You get the reference; you do not get a

  personal copy of everyone else's media to redistribute. What media your archive
  contains is your own — profile and post media, DM media you were part of — which is
  why the shelf reads as links rather than as a private mirror of other people's
  work.⁵ The distinction matters beyond bookkeeping: a list of addresses you
  collected is yours to keep forever; a folder of other people's files would be a
  different product with different problems, and it is not the one you requested.

Where saves surface in the app

Once the records are parsed, the shelf shows up in four places, each using a different
slice of the same four fields:

• The Activity view carries saved posts and saved music as searchable sections —

  the music one searchable by artist or track, because its record carries those
  details while the post record does not.

• The Accounts view turns saves into a filter: *saved posts — accounts whose posts

  you saved*. That works by attributing each reference's link back to the account it
  points at, which is the href field earning its keep.⁶

• Compare shows a saved-posts metric card on each side and a row in the changes

  table, so two exports taken months apart show the shelf growing or shrinking — one
  of the few metrics where growth is unambiguously just growth.

• The heatmap plots save events over time as their own event kind, alongside

  saved-music events, so the collecting habit itself gets a shape in time.

Every one of those views is reading the same links. None of them can show you the
post behind a dead one — and none pretends to.

Questions this comes up

The activity post for the full map of which file each action lives in; the request
post for choosing the format that gives this folder captions; the eight-things post
for why some links on your shelf will never open again.

¹: packages/shared/src/parsers/activity.ts — saved_posts.json is read via
the generic record extractor: string_map_data entries resolve to value, href,
timestamp and title, with the timestamp preferring an explicit time/date field.
The HTML variant saved_posts.html is collected the same way after conversion.
²: The Features view's "Saves & Music" row — posts saved plus bookmarked
music with artist info and dates — as backed by saved_posts.json and
saved/saved_music.json.
³: packages/shared/src/analytics/activity-taxonomy.ts — savedPost
("Posts you saved") and savedMusic ("Music you saved") are classified yours: true
in group collected, separate from interaction kinds; the people-mix and
classification code keeps collected items out of "who you reached" totals.
⁴: This site's request guide, format section: saved posts carry captions in
HTML and lose them in JSON — the single folder where the human-readable format is
richer than the raw record.
⁵: The archive's media is organised around your own content (profile and post
media, media from your message threads); the saved shelf parses to reference records
without a media bucket behind it — see the content and activity posts for the
file-by-file split.
⁶: The Accounts view's saved-posts filter matches accounts whose parsed
interactions include savedPosts, attributing each reference record to the account
its link points to.