Instagram’s originality rule and what it does to reach

The rule, as documented

Dimension	What the platform states	Confidence
Window	Rolling 30 days, across recently posted reels, photos and carousels	DOCUMENTED
Test	If most of what the account posted in that window is content it did not create, it is classed as an aggregator	DOCUMENTED
Penalty	Ineligible for recommendations to non-followers — removed from Explore, Feed recommendations and Discover. Existing followers still see posts.	DOCUMENTED
Nature	Algorithmic classification — not a ban, not a strike	DOCUMENTED
Recovery	Eligibility returns after a 30-day period in which most recent posts are original; standing is visible in Account Status and appealable	DOCUMENTED
Exact threshold	The platform says "most" — no numeric ratio is published. Any tool quoting a precise floor is inventing it.	UNKNOWN

Note what the rule is not: it is not per-post, not a copyright action, and not a notification-bearing event. It is an account-level judgement that quietly changes who can be reached.

Infographic: the rolling 30-day window of posts feeding the 'mostly original?' decision — recommendations stay on when most posts are original; otherwise the account is classed as an aggregator and cut off from non-followers until 30 mostly-original days recover it — THE AGGREGATOR CYCLE — WINDOW, TEST, PENALTY, RECOVERY

What explicitly does not count as original

Meta's guidelines name the insufficient moves outright — this list is the operative part of the policy: DOCUMENTED

Adding watermarks, borders, or captions
Changing playback speed
Re-uploading a post, or a screenshot with visible credit
Stitching clips together without original material
Reacting with facial expressions, or narrating what is already on screen
Listicle / montage compilations without original narration

What does confer originality: content you wholly created, material edits that meaningfully enhance, genuine humour / commentary / analysis, and a creator on screen presenting something new. Meta's own stated test:

"If someone could remove your contribution to your post or reel, and the content would virtually be the same, it probably needs more of you in it."

Why file edits cannot reach this layer

File-level detection asks "what is this file?" — the originality classifier asks "did this account add anything?", evaluated over 30 days at account level. A transformation pipeline (crop, rescale, color, fps, codec, pitch) has no input into that question at all. Worse: every classic transformation is either on the explicitly-insufficient list above or is a strictly smaller change than things on that list — a codec swap is less of a contribution than adding a caption, and adding a caption is stated to be insufficient. DOCUMENTED

The structural consequence for new accounts

Recommendations are the only mechanism by which a new account reaches non-followers at scale. An account classed as an aggregator early has its growth ceiling capped at its existing follower graph — there is no path to acquiring the followers that would make the penalty survivable. For anyone automating publishing, this makes originality a launch gate, not a cleanup task.INFERENCE (the structural reading; each input fact is documented)

Engineering for the rule

Measure the window yourself. Keep a per-account rolling 30-day classification of every post, so the ratio is a number you read today — not a story reconstructed from a reach collapse three weeks later.
Hold a margin above "most." The platform's threshold is unpublished; a system that blocks its own scheduling at a configurable margin above the majority line fails loudly and early instead of silently and late.
Record the contribution. Capture what was added to each piece (the attestation) at publish time — it is the evidence for an appeal and the training data for your own classifier.
Watch standing, not just analytics. Account Status surfaces the classification; poll it instead of inferring from reach curves.