Everyone has a theory about the Facebook Reels algorithm. Few bother checking what Meta actually said. That’s a shame, because in the first half of 2026 Meta published an unusually detailed paper trail — two Newsroom posts, two engineering write-ups, a peer-reviewed paper, and a ranking explainer updated in June — covering three separate changes to how Reels get found, who sees them, and what gets pushed down.
Here’s what Meta confirmed, what marketers invented, and what to actually do about it.
What Meta confirmed
SilverTorch: the retrieval engine. SilverTorch is the system that narrows millions of candidate posts down to thousands for ranking, running at 23.7x the throughput of Meta’s old setup. Important nuance: it’s plumbing, not a penalty. It decides how much content gets considered, not who gets punished. If someone tells you “SilverTorch demotes X,” they’re guessing.
Friend Bubbles: your viewers’ friends are a distribution channel. Not new, but newly explained — Reels that a viewer’s close friends reacted to get pulled into that viewer’s feed, with one tap opening a conversation. Content that sparks friend-to-friend interaction gets a second distribution wave beyond the initial audience.
Fresh and original is winning. Meta reported surfacing over 25% more same-day Reels than in Q3 2025, and said views of original Facebook Reels roughly doubled in the second half of 2025 compared to the same period in 2024. The accompanying guidance is explicit: prioritize original production and substantial creative contributions; expect reduced distribution for copied content and low-value edits.
Lazy reactions are the target. Meta is specifically going after low-effort reaction formats: watching along, stitching clips with minimal additions, slapping borders or speed changes on someone else’s video. Repeat it and your account can become non-recommendable — and demonetized. This is the enforcement arm of the originality push.
The myths Meta never said
“DM shares are the #1 ranking signal.” Meta’s published list of Facebook Reels ranking predictions has ten items. None mentions sharing. Shares count — nobody outside Meta knows exactly how much — but anyone citing a specific share multiplier is inventing it.
“The algorithm punishes AI content.” Meta never said this. What it penalizes is unoriginal and deceptive content, whether AI-made or not. Labeled, original AI content is fine; unlabeled AI persona accounts lose recommendations — but that’s a disclosure rule, not an AI penalty.
“Post X times per day at Y o’clock.” Meta’s guidance points to your own audience data, not universal schedules. The new Creator Studio app answers timing questions from your page’s actual behavior — that’s more useful than any generic timetable.
What the confirmed facts mean for your strategy
1. Original production is the whole game. “Original” doesn’t mean expensive — it means you made it. Film it yourself, edit it yourself, add genuine creative input. Meta’s data shows original Reels views doubled year over year; the algorithm is quite literally tuned to find and reward this.
2. Freshness has a measurable boost. With 25%+ more same-day Reels being surfaced, recency matters more than it used to. This favors consistent posting over occasional masterpieces — and it means trending topics have a shorter half-life. Post while the trend is warm.
3. Design for the friend bubble. Content that one friend sends to another — funny, relatable, useful, surprising — gets the second-wave distribution. Ask yourself: would someone share this with a specific friend? “Tag someone who…” bait doesn’t count (engagement bait is separately penalized); genuine shareability does.
4. Kill the lazy formats. If your workflow is downloading others’ videos, adding captions or borders, and re-uploading — stop. That’s exactly the pattern Meta is now willing to demonetize accounts over. Re-uploaded content is already being downranked across platforms.
5. Watch time still matters, but interest matters more. Completion rate and rewatches remain core signals, but Meta’s ranking now also weighs whether the reel genuinely matches viewer interests — passive completions with low interest ratings get distribution capped. A 7-second reel everyone finishes beats a 60-second reel everyone abandons.
A 7-day playbook built only on confirmed facts
Enough theory. Here’s a week of actions derived strictly from what Meta confirmed:
Day 1 — Audit for originality. Scroll your last 20 reels. Flag anything that’s a re-upload, a border-and-caption job on someone else’s video, or a reaction with minimal added value. Those are now demonetization risks, not just low performers. Delete or unlist the worst offenders.
Day 2 — Fix your hooks. Since completion rate and interest-match drive distribution, your first 2 seconds decide everything. Re-edit your next three reels to put the payoff, the surprise, or the question up front — no slow logos, no “hey guys welcome back.”
Day 3 — Batch original production. Film 5–7 short, original reels in one session. Keep them tight (7–30 seconds), one idea each. Freshness is rewarded; a backlog lets you post consistently without daily filming stress.
Day 4 — Mine your comments for friend-bubble fuel. Read your comments and DMs for the questions and jokes your audience already gives you. Content that answers a real follower question is inherently shareable — it’s the organic version of the friend-bubble mechanic.
Day 5 — Test length deliberately. Meta’s own data points toward testing longer cuts where the content earns it, while short reels win on completion. Post one 7-second and one 45-second original reel on similar topics and compare retention in insights. Your audience’s data beats any guru’s rule.
Day 6 — Check your disclosure hygiene. If any of your content uses AI generation, make sure disclosure is handled. Clean compliance means the algorithm judges your content on merit, not on policy flags.
Day 7 — Review and double down. Look at reach, 3-second views, and average watch time across the week’s posts. The two best performers reveal your current winning formula — make three more variations next week. The algorithm rewards iteration on what works, not random experimentation.
Official source: Meta Business Help
Frequently Asked Questions
Does the Reels algorithm favor followers or discovery?
Discovery. Unlike Feed, which leans on your network, Reels is the discovery surface — it shows your content to non-followers based on content signals. Follower count matters less here than content quality.
How many reels should I post per day?
Meta hasn’t given a number. The freshness boost rewards consistency, but quality gates everything — three original reels beat ten lazy ones, especially now that lazy formats carry demonetization risk.
Do hashtags still matter for Reels?
Less than they used to. Meta’s systems read the actual content — visuals, audio, captions — to understand what a reel is about. A few relevant hashtags don’t hurt, but they’re not a ranking lever.
Will using trending audio boost my reel?
Trending audio helps the system categorize your content and can ride a wave of interest, but it’s not a substitute for originality. Using a trending sound on an original video: good. Re-uploading someone’s trending video: penalized.
Where can I read Meta’s own statements?
Meta’s Newsroom posts, engineering blog write-ups, and the ranking explainer (last updated June 2026) are the primary sources. Marketer breakdowns are useful but always check which claims trace back to Meta and which are inference.