How the Facebook Algorithm Works in 2026: A Complete Breakdown

Open your Page’s Professional Dashboard and look at two posts from the same week. One reached 400 people. The other reached 40,000. Same Page, same followers, same posting time. Nothing about your audience changed between Tuesday and Thursday.

What changed is that Facebook scored those posts differently for every single person who might have seen them.

There is no single Facebook algorithm handing out reach. There is a ranking system that runs separately for each user, every time they open the app, scoring your post against every other candidate for that person’s attention. The 400-reach post did not fail once. It lost the same comparison tens of thousands of times.

The Four Stages Facebook Runs on Every Post

Meta describes Feed ranking in its Transparency Center material and in the AI system cards it publishes for Feed and for Reels. The pipeline runs in four stages, and knowing which stage your post died at tells you what to fix.

Stage one: inventory

Facebook gathers everything that could appear for this person right now. Posts from friends. Posts from Pages and Groups they follow. And, increasingly the largest bucket, recommended content from accounts they have no connection to.

If your post is not in inventory, nothing else matters. A post that breaks Community Standards is removed outright. A post flagged by an integrity classifier can stay visible to followers while being pulled out of recommendations, which is the invisible ceiling most struggling Pages hit.

Facebook has no feature called a shadowban, and Meta never uses the word. The real mechanisms have names: reduced distribution, and ineligibility for recommendations under the Recommendations Guidelines. Account Status in your Professional Dashboard tells you whether either applies to your Page, and lets you request a review.

Stage two: signals

For each candidate post, the system pulls thousands of data points. Who posted it. How old it is. How fast it is accumulating reactions relative to similar posts. Whether this viewer has commented on your Page before. Whether the post is video, image, text or link. Whether the viewer recently hid something from you.

Stage three: predictions

This is the part people miss. Facebook does not rank on what already happened. It ranks on what it predicts will happen: how likely this person is to comment, to share, to watch past the first few seconds, to hide it, to report it. Your past engagement trains those predictions. It is not the score itself.

Stage four: score

Each prediction is multiplied by a value weight, positive or negative, and the results combine into one relevance number for that post and that viewer. Highest scores go to the top of Feed; the rest drop down or out of the session.

The consequence: a post with a high predicted hide rate can score below one with weaker predicted engagement, because the negative weight swamps the positive. You do not need bad engagement to lose. Predicted annoyance is enough.

The Shift Away from Pages, and What It Means for You in 2026

For roughly a decade the deal was simple: build followers, post to them, reach some fraction of them. Organic reach shrank year over year, but the model held: your distribution was a function of your audience.

That model is largely gone.

Meta has been explicit, in product announcements and on earnings calls, that a large and rising share of Facebook Feed is content recommended from accounts the viewer does not follow. Facebook competes with TikTok and YouTube Shorts on interest-graph recommendation now.

Two things follow. The bad news: your follower count buys less than it used to. A 200,000-follower Page can post something mediocre and reach 3,000 people, because followers only see it if it beats the recommendations queued up for them.

The good news is real. A 900-follower Page can reach a six-figure audience, because eligibility for recommendation does not depend on follower count. What decides it is how the post performs with whoever sees it first.

Picture that correctly, though. Facebook has never described the fixed test batch marketers like to imagine, where a post goes out to 500 people and gets promoted if it clears a bar. Every impression is scored on its own, and interactions from the first few hundred update the predictions applied to everyone scored after them. Strong early behavior makes the next round more favorable. Weak early behavior stalls it quietly.

So your Page is an entry ticket in 2026. Followers are the fuse.

The Signals That Actually Move a Facebook Post

Not all engagement is worth the same. Meta’s Meaningful Social Interactions framework, introduced in 2018 and refined since, weights interactions by how much genuine person-to-person connection they represent. A reaction is cheap. A reply inside a thread is expensive.

The weights below are directional, drawn from how Meta describes the system and what Page operators observe. Facebook has never published exact coefficients, and anyone quoting them is guessing.

The asymmetry is the point. Many reactions are needed to offset one hide, and negative signals arrive faster, because annoyance is the quicker reflex.

Comments, and Why Replies Are Worth More Than the Comment Itself

When someone comments, Facebook logs one meaningful interaction. When you reply and they reply again, it logs a conversation, a back-and-forth between two accounts. The second is worth more, because that back-and-forth is what MSI was built to promote after Meta decided passive consumption was hurting the platform.

Timing compounds it. Early interaction matters more than late interaction because it feeds the predictions applied to whoever gets scored next. A comment in the first hour changes how your post is rated for the following viewer. A comment on day three arrives after the distribution has largely settled.

So the highest-leverage half hour of your week is the one right after you publish:

• Block 20 minutes immediately after every post goes live, and reply to every comment with a genuine question of your own.
• Write posts that make one-word answers impossible. “Which of these two would you pick and why?” beats “Do you agree?” every time.
• Never delete a critical comment unless it breaks your rules. Disagreement generates more replies than applause, and the ranker reads volume.
• Post when you are actually available to respond. A perfect posting time you sleep through is worse than a mediocre one you work.

If you run several Pages and cannot cover every launch window yourself, a supplementary distribution service such as LitFame can give a post enough early surface area for real conversation to start, but only if the content is worth commenting on. Engagement bought onto a post nobody wants to discuss produces reactions and silence, close to the worst signal mix available.

Shares to Private Surfaces: The Signal You Cannot See in Insights

Facebook’s share button is not one action. Sending a post to a specific person in Messenger, dropping it into a Group, and resharing to your own timeline are three separate behaviors, and the ranker does not treat them equally.

A send to one person is the strongest voluntary endorsement on the platform. It costs the sender something: they are vouching to someone they know that this is worth their time. That cost is why the signal is hard to fake.

You can barely measure it. Insights shows shares in aggregate, but sends into Messenger threads and closed Groups are effectively invisible. When a post has modest public numbers and outsized reach, quiet forwarding is usually the explanation.

What gets sent privately:

• Something specific enough that a person thinks of one particular friend: a local reference, a niche complaint, a role-specific frustration.
• Practical information with a deadline: price changes, closures, application windows.
• Something mildly contentious that a person wants a second opinion on before commenting.

What never gets forwarded: generic motivational content, announcements in marketing voice, anything that reads as an ad. Design one post a week for private forwarding, and name the two people you imagine sending it before you write a word.

Dwell Time and Video Watch Time

Dwell time is how long your post stays on a viewer’s screen before they scroll past. Nobody decides to do it, which is what makes it one of the more honest inputs the system has. You cannot ask people to dwell.

It separates what people consumed from what they liked reflexively mid-scroll. Fifty reactions with no dwell behind them read very differently than twenty that each held someone for several seconds.

Things that reliably extend dwell:

• Text posts long enough to trigger the “See more” cut. Tapping that expansion is a small commitment, and it buys several seconds.
• An opening line that creates an unresolved question the rest of the post answers.
• Images with readable detail, and multi-image posts that require deliberate swiping.

For video, watch time is the same idea with better instrumentation. Meta consolidated its headline video metric to Views across Facebook and Instagram, but ranking cares about retention: how many viewers survive the first three seconds, and average watch time as a fraction of length.

The most reliable fix is cutting your first two seconds. Most Page video opens with a logo, a slow zoom or someone saying hello. Delete it and start on the most interesting frame in the clip.

Short and fully watched beats long and abandoned, because retention is judged against what the model expected for a video of that length.

What Gets You Demoted

Meta publishes Content Distribution Guidelines in its Transparency Center, listing categories that get reduced distribution without being removed. It is a documented list. Three parts matter most to business Pages.

Hides, “Show less”, and unfollows

Facebook gives viewers direct controls: hiding a post, choosing Show less on a Page’s content, snoozing an account for 30 days, unfollowing, reporting. Each is an explicit instruction to the ranker, and each carries outsized negative weight because it is a stated preference rather than an inferred one.

Hides scale with frequency more than with quality. A Page posting five times a day gets hidden not because any one post is bad but because the cumulative presence irritates. If reach fell off a cliff and your content did not change, check your cadence over the last two weeks. The fix is unglamorous: post less, make each post carry more.

Clickbait and engagement bait

Clickbait, meaning headlines that withhold information to force a click or exaggerate what sits behind the link, has been a documented demotion category for years. Facebook detects it partly through phrasing and partly through behavior: if people click and bounce straight back, the destination did not deliver.

Engagement bait is the more common trap, because it looks like a tactic. Posts that explicitly ask for interaction (“LIKE if you agree”, “comment AMEN”, “share to win”, “tag three friends”, vote-bait with reaction emojis) are demoted by classifiers Meta trained for exactly this, under a policy announced in December 2017. It applies at Page level too, so repeat offenders lose distribution across everything they publish.

The distinction that catches people out: a genuine question is fine. “What’s your experience been with this?” is a real prompt. “Comment YES if you’ve ever felt this way” is bait, because the action asked for is decoupled from thought.

Links to low-quality destinations

Facebook evaluates where your links go as well as what your post says. Ad-heavy destinations, pages that take several clicks to reach the substance, cloaking and misleading redirects, and domains with a pattern of thin content all reduce distribution for the posts pointing at them.

This applies to your own site. A 300-word post wrapped in six ad units and an interstitial is a low-quality destination, whatever you think of it. Fast landing pages that match the promise in the post are ranking factors in practice.

Reels on Facebook Is a Separate Ranking Surface

Do not treat Facebook Reels as Feed with vertical video. The Reels player is a separate surface with its own candidate pool and predicted actions, and Meta publishes a separate system card for it.

The core difference is connection. Feed still leans on your existing audience; Reels is recommendation-first, with most of what a person sees coming from accounts they do not follow. That is why a Page with almost no followers can post a Reel and reach six figures while its photo posts reach 200.

The predictions skew differently here:

• Watch-through and replays. A loop is the strongest single signal, which is part of why 7–15 second clips often outperform 45-second ones.
• Sends. Forwarding a Reel to a friend carries far more weight than a like.
• Swipe-away speed. The Reels equivalent of a hide. Enough fast flick-aways and the predictions for later viewers drop with them.
• Sound. Reels are watched with sound on far more often than Feed video, and the track you attach shows as a tappable label leading to other Reels using it. A clip that only works muted gives up both.

Two practical rules. Never post a Reel carrying a visible watermark from another app: Meta’s guidelines put recycled watermarked video in the reduced-distribution bucket. And put your hook on screen as text in the first frame, since the interface covers part of the caption and viewers rarely expand it.

Publish nothing but Reels, though, and expect a shallow audience. That surface hands you views rather than followers who care. Pair Reels for discovery with Feed posts for depth.

Why Group Distribution Behaves Differently from Page Distribution

Groups sit on a different footing, for structural reasons rather than favoritism. Joining a Group declares active interest in a conversation; following a Page subscribes you to a broadcaster. The system treats those declarations differently because the behavior behind them differs. Three consequences:

Notification defaults. Members can set a Group to notify them on every post and add it to Favorites so its content gets prioritized. Page followers have the same controls and almost never touch them. That is opt-in distribution, and it skips part of the competition.

Member-to-member interaction. Group posts generate conversation between members themselves. That is a denser interaction graph than a Page post, where all engagement runs through you.

Topical coherence. An active Group has a narrow subject, so the model has a cleaner read on who else might want its content, and easier posts to recommend.

The caveat businesses hate hearing: a Group that works like a Page with extra steps gets none of this. If you are the only one posting and members receive announcements rather than participate, you have built a mailing list with worse deliverability. The advantage comes from member behavior.

If you run both, split them properly. The Page carries public identity, Reels, reviews and ads; the Group carries questions, members-first announcements, and discussions you seed but do not dominate. Link the Group from the Page and let customers self-select into it: the Group ends up smaller and much more active.

Turning Every Mechanic into Something You Actually Do

Weekly cadence. Three to four Feed posts, two to four Reels, one post built to be forwarded privately. Fewer and you lose momentum; more and you collect hides.

Before publishing. Read the first line alone. If it gives nobody a reason to stop, rewrite it. Check that there is a question nobody can answer in one word, and that you are not asking for likes, tags or shares.

After publishing. Twenty minutes of replies, a question at the end of each, and pin the best comment.

Weekly review, 15 minutes. Sort Insights by reach and pull your top three and bottom three posts. For the bottom three, check cadence and format before you blame the topic; Meta reports very little negative feedback back to you now, so a reach drop that tracks a week of heavier posting is the evidence you will get. For the top three, name the format rather than the subject and run it again. Monthly, check where your outbound links land, and reread a month of captions for anything that would trip the engagement bait classifier.

Now the inconvenient part, written by a company that sells growth services and would rather you spent the money well.

Buying reactions does very little for ranking. Reactions sit near the bottom of the weight table, which is exactly why they are the cheapest thing to sell. Worse, engagement from accounts with no real interest in your subject muddies the signals the recommendation system uses to decide who else to show you to, and inauthentic engagement breaches Community Standards in its own right.

Supplementary services have a real use, narrower than the marketing suggests: giving a post that already deserves attention enough early visibility to escape a cold start, or making a new Page look established enough that a first-time visitor takes it seriously. If that is the job you actually have, create an account and buy the smallest thing that tests it, then judge the result on comments and watch time. Nobody can guarantee reach on a system that scores every impression separately, and any seller who does is quoting a number rather than an outcome.

Frequently Asked Questions

Does posting at a specific time still matter on Facebook in 2026?

It matters less than it did, because recommendation-driven distribution can surface a post days after you publish it. It still matters for one reason: interactions collected early feed the predictions applied to everyone scored later, and you get more of them when your audience is awake. Skip the universal best-time charts. Check Insights for when your own followers are active, and pick a slot you can be present for.

Why did my Facebook Page reach suddenly drop by half?

Check three things in order. First, posting frequency over the previous two weeks: a jump in cadence is the most common cause of a hide-driven collapse. Second, Account Status in your Professional Dashboard, which tells you whether something made your content ineligible for recommendations. Third, whether you changed format or started adding outbound links. Meta reports little negative feedback to Page owners now, so expect to diagnose this from your own cadence and reach history.

Do Facebook likes and reactions still help ranking?

They help, but far less than comments, replies, shares or dwell time. A reaction is one tap requiring no thought, which makes it weak evidence of real interest and easy to manipulate. Treat the reaction count as a rough health indicator. A post with plenty of reactions but almost no comments and short dwell reads to the ranker as something people acknowledged rather than engaged with.

Does Facebook penalize posts with external links?

Not automatically. What Meta documents is reduced distribution for links to low-quality destinations: ad-heavy pages, sites using cloaking or misleading redirects, and pages that do not deliver what the post promised. A link to a fast, substantive page you control is fine. If your link posts consistently underperform, audit the destination first, then test the same content with the link in the first comment and compare over several posts.

Is Facebook Reels ranked the same way as Feed?

No. Reels is a separate surface with its own candidate pool and its own system card from Meta. It leans far more heavily on recommendations from accounts the viewer does not follow, and it weights watch-through, replays and sends more than Feed does. Video recycled from other apps with a visible watermark gets reduced distribution in recommendations, so export clean files instead of downloading from another platform.

Should I start a Facebook Group instead of a Page?

Run both, for different jobs. Groups get structurally better distribution because members opt in actively, can set notifications to all posts, and talk to each other rather than only to the admin. That advantage disappears if you are the only person posting. A Group where members never talk is a Page with fewer features. Keep the Page for public identity, Reels and ads, and use the Group for discussion.

What counts as engagement bait on Facebook?

Asking explicitly for an interaction that is decoupled from real thought. Comment YES, tag three friends, share to win, LIKE if you agree, and vote-with-a-reaction posts all fall under the policy Meta announced in late 2017. Genuine questions are fine. The test is whether answering requires the person to think. Repeat offenses can reduce distribution across your whole Page, beyond the offending post.

How long does it take for a Facebook Page to recover from reduced reach?

If the cause was accumulated negative feedback rather than a policy violation, expect several weeks of reduced cadence and better posts before it turns. There is no reset button. Account Status lets you request a review of a specific eligibility problem, but ordinary negative feedback has no appeal path. Cut to roughly three strong posts a week, prioritize conversation over reactions, and expect dwell time to improve before reach does.