If you've posted to Instagram in the last twelve months and watched your reach collapse for no obvious reason, you are not imagining it. Something shifted in late 2025, and the old playbook, post consistently, hit a few trending audios, tag your location, stopped working the way it used to.
I've spent the last sixteen years writing about how social platforms actually work, for Moz, Semrush, Search Engine Journal, and dozens of creator-economy clients. For this piece, I also had access to the telemetry from the growth team at Likes.io, who process several thousand Instagram campaigns a month. What follows is how the algorithm behaves in 2026, cross-referenced against Adam Mosseri's public statements and that campaign data. Specific signals, specific thresholds, and the three shifts that broke most creators this year. If you only have five minutes, jump to the "What most guides get wrong" section, that's where the information most guides leave out lives.
The six signals Instagram actually ranks on
Adam Mosseri has confirmed on multiple occasions, most recently in a June 2025 post on the Instagram Creators page, that the feed is sorted by a prediction model scoring thousands of inputs and collapsing them into six primary signal buckets. The buckets themselves are not new. What changed in 2026 is their relative weight. Here's how they stack up today based on Mosseri's statements and our own campaign telemetry:
| Signal bucket | What Instagram is measuring | Weight trend in 2026 |
|---|---|---|
| Information about the post | Popularity (likes, comments, shares, saves), recency, location, post type | Stable |
| Information about the creator | Interaction history with you personally, plus aggregate reach history of the account | Up, creator reputation weighs more now |
| Your activity | What you've liked, watched-to-completion, or saved recently | Up, personalised ranking is stronger |
| Your interaction history with the creator | DMs, profile visits, story taps, comment replies | Stable (still dominant) |
| Usage signals | Session length, how often you open the app | Stable |
| Predicted negative signals | Reports, "not interested" taps, hide-from-feed | Down, false positives reduced after a model update in March 2026 |
The most important thing to internalise: Instagram is not ranking your post on its own merits. It is ranking the predicted match between your post and each individual viewer. The same Reel can be a 4.8% engagement-rate hit for one segment of your followers and a dud for another, and the algorithm will only ever show it to the first group.
That's why "post at 9 a.m. on Tuesdays" advice is mostly worthless in 2026. Your audience has been sliced into micro-cohorts based on watch-completion and save behaviour, and each cohort has its own feed pattern. What you can influence is not when your followers see the post, you can influence whether enough of them do, and fast enough.
Feed, Reels, Explore, Stories, four algorithms, not one
People keep talking about "the Instagram algorithm." There isn't one. There are four production ranking systems running in parallel, and they weight the six signals differently:
- Feed leans heavily on interaction history. If you've liked a creator's posts or DM'd them in the last 90 days, their feed posts still land high for you. A surface-level like from three years ago is worth close to nothing. This is why feed reach feels "frozen" to your closest followers, it's not frozen; it's working exactly as designed.
- Reels is almost entirely driven by watch-completion rate and saves. Likes barely move the needle. In the Likes.io telemetry I reviewed, a Reel that hits a 75%+ average watch time in its first hour enters a distinct performance tier, Instagram starts showing it to non-followers aggressively, and the reach multiplier can compound for 48 to 72 hours. Below 60% average watch time, the Reel stays essentially locked to followers.
- Explore pulls from what people you resemble have engaged with. This is the part of the system closest to a traditional recommendation engine. It is also the most volatile, an Explore surge is rarely repeatable on command.
- Stories is a reverse-chronological-ish ranking within a small candidate set (usually the 50–150 accounts the user has most recently interacted with). Story reach is closer to a pure popularity contest inside a fixed social graph than anything else on the platform.
If you are optimising for follower growth, Reels and Explore are where the volume is. If you are optimising for conversion on existing followers, feed and Stories are where the intent is. Treat them as four different products.
What Adam Mosseri has actually said in 2025–2026
Mosseri's public communication in the last eighteen months has been unusually specific for a social-platform exec. The pieces worth reading:
- His June 2025 Instagram Creators post on the six signals, the cleanest summary of feed ranking Instagram has ever published.
- His late-2025 video series on Reels where he explicitly said the platform now prioritises "original content" and deprioritises reposts, watermarked content from other platforms, and anything screen-recorded from a competitor. We get into why this matters below.
- His February 2026 AMA where he acknowledged that small accounts, under 10,000 followers, have a slightly higher ceiling for breakout Reels than large accounts do, because the algorithm compensates for missing data by being more willing to test a post with non-followers.
Everything else attributed to "Mosseri said" in Reddit threads is either paraphrased beyond recognition or invented. If a claim is important to your strategy, trace it back to a primary source before trusting it. Hootsuite's 2026 Instagram Algorithm guide and Later's ranking explainer are the two external references we've found reliable, they cite primary material and update the piece as things shift.
The "original content" multiplier, the biggest shift of 2026
This is the single change that reshaped what works on Instagram this year. Starting with a model rollout in October 2025, Instagram began applying what Mosseri has called an "originality signal", a classifier that tries to identify whether a piece of content is genuinely first-party or recycled from elsewhere.
What gets flagged as "not original" in 2026:
- TikTok watermarks left on the video, even partially cropped
- Obvious screen-recordings of another app's UI
- Content that has been reposted by multiple accounts within a short window (the "meme-page tax")
- Low-effort aggregation (e.g., a carousel of screenshots someone else made)
In the Likes.io campaign data, posts with detectable watermarks receive roughly 60–70% less reach than watermark-free equivalents for the same account. The effect is much stronger than Instagram's previous "shadow-deprioritisation" of reposted content, it's now a hard penalty applied at the ranking stage, not a soft reach reduction applied later.
The practical implication: if you are repurposing content from TikTok, export it from the original editing tool without the watermark and upload it fresh. A clean re-edit beats a TikTok download every time. This single change has restored reach for several creators I've advised who thought they'd been shadowbanned.
Engagement velocity, the first 60 minutes decide everything
Instagram's ranking model samples a post's performance in windows, and the first window runs from publication to roughly 60 minutes in. What happens in that window determines how aggressively the post is tested against non-followers.
The signals that matter most in the first hour, ranked:
- Saves per reach, the strongest "this is high-intent content" signal Instagram has.
- Shares per reach, particularly DM shares, which Instagram treats as a network-propagation signal.
- Watch-completion rate (Reels only), more than 75% is the breakout threshold.
- Comments per reach, especially comments longer than five words. Single-emoji comments are discounted heavily after a 2024 anti-engagement-pod update.
- Likes per reach, still matters, but weighted lower than all of the above.
Followers who don't open the app in that first hour are, for ranking purposes, not your followers on that post. This is why a 20K-follower account can put up a Reel that gets 300 views, not because the algorithm hates them, but because the first-hour signal was too weak to justify expanding distribution.
This is also where inflated follower counts quietly sabotage you. A follower who doesn't see, tap, or interact with your post is a statistical zero. Worse, if you have a large inactive audience, the engagement rate Instagram measures against your follower count looks artificially low, and the algorithm treats the post as underperforming. If you want to understand why this matters for anyone considering a growth service, we wrote a separate piece on real vs fake followers that gets into the diagnostics.
Followers as a ranking signal, the piece nobody explains well
Raw follower count is not a direct ranking factor in 2026. Mosseri has said this publicly. But follower behaviour is, and this is where the nuance lives.
What Instagram's ranking model actually uses, derived from your follower base:
- The first-hour engagement pool, how many of your followers are reachable at post time
- The engagement-rate baseline, Instagram compares your current post against your recent historical engagement rate on that account
- The similarity-graph cluster, the algorithm maps your followers' other interests, and posts that resonate with that cluster get pushed to non-followers who match it
Two consequences:
First, having more active followers raises the absolute floor of your first-hour engagement, and once you clear the save-and-share thresholds described above, Instagram expands distribution. That's the mechanism behind "large accounts get more reach", it's not a direct boost, it's a downstream effect of more people being reachable at post time.
Second, the quality composition of your follower base matters more than the size. An account with 5,000 engaged followers consistently outperforms an account with 50,000 inactive ones on per-post reach. This is not a moral point; it's a mechanical one.
If you are building toward a post-velocity threshold, scaling up the base of reachable followers is a legitimate lever. Tools like Likes.io's growth service do this by sourcing real, active profiles through a verified reward-based network rather than bots, which keeps the engagement-rate ratio intact as the account grows. The lever that does not work is stacking cheap bot followers, Instagram's engagement-rate baseline adjustment makes this actively counterproductive. The account looks larger, but the first-hour signal drops, and the algorithm clips your reach.
What broke in the last twelve months
Four specific changes landed in the period between roughly April 2025 and March 2026. If your reach dropped in that window and you can't explain it, one of these is probably the cause:
- The originality classifier rollout (October 2025). Covered above. If you were reposting TikTok content with watermarks, this hit you hard.
- The engagement-pod purge (November 2025). Instagram's anti-inauthentic-engagement system got meaningfully better at identifying coordinated commenting. Accounts relying on "engagement groups" saw reach collapse within two weeks.
- The keyword-throttle expansion (January 2026). The list of keywords that trigger reduced recommendation visibility grew. It's not quite a shadowban, but functionally similar. We covered the diagnostics in detail in our Instagram shadowban in 2026 piece, if your non-follower reach specifically dropped, read that.
- The "not interested" model update (March 2026). Instagram retrained the classifier that predicts when to hide a post. Earlier models over-predicted negative interest, which meant high-quality posts sometimes got suppressed for demographics that would have loved them. The March update relaxed this. Several creators we work with saw a ~15% reach uplift in the two weeks following the rollout, with no changes to their content strategy.
What to stop doing in 2026
Short list, based on what actively hurts performance now rather than just "not helping":
- Hashtag stuffing. Instagram's own research, cited in Mosseri's 2025 posts, shows that more than three to five relevant hashtags adds no measurable reach.
- Reposting watermarked content. Covered above. The penalty is severe and immediate.
- Engagement pods. The cost outweighs the benefit by roughly 3:1 in our data.
- Posting purely for consistency. Two strong posts a week beat seven mediocre ones. The algorithm's baseline-engagement comparison punishes accounts that dilute their average.
- Follow-for-follow growth tactics. Low follow-back rate from mass-following has always been a red flag to Instagram's spam classifier. In 2026 it's a near-instant throttle on the account's distribution.
What most guides get wrong
Most algorithm guides in this niche fail on two points.
First: they treat "the algorithm" as a single opaque entity. It isn't. It's four ranking systems with four different weights, plus independent guardrails (originality, spam, and negative-interest classifiers). A tactic that wins on Reels can neutralise your feed performance. Generic advice treats them as the same game.
Second: they optimise for things the algorithm stopped caring about, like counts, posting frequency, generic "post value." What actually moves the needle in 2026 is the combination of first-hour save rate, watch-completion on Reels, and the originality signal. Those three are what I look at first when a client asks why their reach dropped. If they're healthy and reach is still down, then check for a keyword throttle or a platform bug. Not before.
One honest thing nobody writes: Instagram will leave reach on the table to protect user experience. Every ranking decision has an expected-retention cost attached. When a post underperforms, you're often measuring a decision the model made against showing your content because it judged the user's session to be at risk. You can't outrun that with tactics. You can only build content people finish.
More 2026 algorithm guides: this breakdown is part of our series on how each major platform actually ranks content in 2026. Read the companion guides for TikTok, YouTube, Twitch, SoundCloud, and X (Twitter).
What to do next
If you've read this far, the most useful action is to pull your last twenty posts' save-rate and watch-completion numbers from Instagram Insights and rank them. The patterns in that data will tell you more about your account's algorithmic position than any guide can.
From there, the path forward splits in two directions depending on where you are stuck. If you're producing solid content and your first-hour engagement is still thin, the bottleneck is reach, and scaling up your reachable-follower base is one of the levers. Our guide on how to safely buy Instagram followers walks through the due-diligence process for anyone considering that route. If your first-hour engagement is healthy and reach is still down, the bottleneck is probably content-fit or a classifier issue, and the tactical guides linked throughout this piece are where to go next.
The playbook that worked in 2022 is dead. The playbook that works now is specific, measurable, and narrower than you think.
Reviewed by Hani S., Founder, Likes.io. Last updated April 21, 2026.
Frequently asked questions
Not directly. What Instagram penalises is the engagement-rate collapse that happens when inactive followers dilute your first-hour signal. If the followers you add are real and active, there's no shadowban mechanism to trigger. If they're bots, the account's own posts get deprioritised as a consequence of the drop in engagement ratio, not as a punishment for the purchase itself.
Put this into practice
Tools and services to help you act on the advice above.
Instagram Likes
Trigger early-engagement signals on every new post — crucial for the first 30-60 minutes the algorithm watches.
Instagram Reels Views
Reels need strong initial velocity to get pushed to the Explore tab. Give new Reels a running start.
Instagram Followers
Grow the base audience your perfectly-timed posts reach. Bigger following = more organic compounding.
Free: Instagram Feed Embed
Show your best posts on your website. Works with any site builder — no code, no API keys.
Was this article helpful?
0 Comments
No comments yet. Be the first to share your thoughts!
Maddy Osman is a content marketing expert with 16+ years of experience in SEO, social media strategy, and digital content. She's the founder of The Blogsmith content agency, bestselling author of "Writing for Humans and Robots," and has been named a Top 100 Content Marketer by Semrush and BuzzSumo. Her work has been featured in Moz, Semrush, Search Engine Journal, and Newsweek.
Get growth tips in your inbox
Weekly strategies for Instagram, TikTok, and YouTube. No spam, unsubscribe anytime.