TL;DR
Adam Mosseri just said the quiet part out loud: Instagram will lean even harder into algorithmic recommendations and away from your Following graph. Discovery will also keep shifting into DMs and small groups. You’ll get more controls to “tune” what you see, but they’re broad (topics/interest resets), not a return to a pure follow feed. Here’s what changed, why it matters, and how to adapt without burning your team.
What actually changed
- More recommendations, less Following. Meta’s been marching this way for years; by early 2024, ~50% of what you see in Instagram was already AI-recommended (not from accounts you follow). That share is still rising. (Newszii)
- Mosseri’s latest guidance: in a recent reel/thread, he reiterated that Instagram is doubling down on recommended content, expanding what it will surface (including previously deprioritized categories) with user controls to dial it up or down — but not a return to a Following-only default. In January, he made a similar call publicly when opening the door to recommended political content with intensity controls, underscoring the broader shift to recommendations. (The Verge)
- Tuning knobs (but not precision): Instagram now exposes/expands Content Preferences — things like Show more/less, Not interested, Sensitive-content controls, and even a reset recommendations option to refresh what the system thinks you like. These are helpful, but they’re broad topic-level levers, not “follow the people I choose, full stop.” (Androidsis)
- DMs as a discovery surface. Instagram keeps shipping features that push sharing and discovery into the inbox (replying to Stories with Reels, DM shortcuts, group prompts). Mosseri has been clear: people share privately more than they post, and product changes reflect that. (Popular Science)
- Trials that bypass your follower graph. Instagram has been testing Trial Reels and other experiments that decouple reach from following, letting creators audition content to non-followers directly. That’s recommendations-first distribution in practice. (SlashGear)
The net effect: Instagram’s feed and Reels rankers are optimizing for what the system predicts you’ll engage with, supported by DM-level sharing signals — not just what the accounts you follow publish.
Why this is a big deal for creators, brands & enterprises
- Your follower count matters less for reach. The ranker can hand your best post to the right non-followers — great for growth if your packaging is strong. But it also means a sleepy or off-brand post can disappear, even with a big audience. (That’s the “lottery” feeling many teams report.)
- Private shares are the new power signal. DM forwards and group chat resonance are increasingly what kick posts into broader recommendation pools. Public likes/comments aren’t the only story anymore. (Popular Science)
- Controls ≠ curation. You can nudge Instagram with topic/interests tools or even reset the algo, but you’re not getting a hard Following mode by default. Expect continued tension between user choice and platform optimization. (Androidsis)
- Meta is comfortable widening the funnel. If Instagram will even recommend political content (with user controls), it’ll recommend almost anything it believes you’ll watch — another clear sign that Following is no longer the primary gate. (The Verge)
How discoverability is changing under the hood
- Recommendations weight > Following weight. The model looks at watch time, completion, replays, saves, shares (especially DMs), and negative feedback to decide who else should see a post. That affects feed, Reels, Explore, and increasingly inbox surfaces. (Popular Science)
- Network effects move to small groups. A post that pops in two or three group chats can now outpace a post that did “fine” with a larger follower base.
- “Tuneable” feed, not “chosen” feed. You can refresh or downrank broad categories via settings, but the ranker still blends your signals with its predictions. (Androidsis)
Drawbacks & risks (let’s be honest)
- Less user agency. Many people want a pure follow feed. Instagram’s answer so far is “use controls,” not “here’s a default Following mode.” (The Verge)
- Advertising gravity. A recommendation-heavy system can feel like an ad conveyor belt, even if the organic discovery is good.
- Filter bubbles via “Show More/Less.” The new controls can unintentionally narrow what you see. Resets help, but few people use them regularly. (Phandroid)
- Creator volatility. Winners win bigger; average posts decay faster.
The operator’s playbook (brands, creators, enterprises)
1) Engineer DM-worthy posts
- Build pass-along value on-frame (checklist, map, before/after, tiny template).
- Add a line that reflects private sharing, not just public hype: “Save this” / “Send to your teammate who runs ads.”
- Track shares and saves as north stars (they’re predictive in a rec-heavy world).
2) Package for cold audiences
- First 1–2 seconds: a promise (“The 3 hooks that triple Reels saves”).
- Legible overlays so a forwarded video explains itself without captions.
- Avoid inside jokes only followers get — assume the viewer has zero context.
3) Treat Following as retention, not as distribution
- Followers are still your repeat buyers, but trial reach comes from recommendations. Run A/B creative sprints to find formats that earn DM forwards.
4) Use the tuning knobs deliberately
- Periodically reset recommendations if your feed feels off (Settings → refresh).
- Hit “Not interested” aggressively to prune junk; use Show more/less on good categories. (Androidsis)
5) Scheduling: post when you historically land
- Instagram’s native “audience online” chart is descriptive; use your own impact-by-hour data to predict when your posts hit. If you’re on Rkive, the Peak Hours clock is built on your past wins; lock winning slots and let autopilot rotate candidates so you’re always timely without babysitting. (Native analytics are fine for quick checks; use your own data for forecasting.)
A short strategy for each team
Creators
- Run a “forwardability” audit: the last 10 posts — how many DMs, how many saves?
- Double down on the format with the highest DM/share ratio and ship two more this week.
Brands
- Publish more problem-solution reels and micro-playbooks (assets colleagues share in chats).
- Instrument saves/shares per format in your weekly readout; don’t rely on likes.
Enterprises
- Build a reply-video pipeline for repeated questions (great DM fodder).
- Spin up a creatives × analytics ritual: weekly “DM spread” leaderboard, retire formats that don’t get forwarded.
Bonus: If you’re juggling multiple handles, let Rkive auto-slot winning formats at your Peak Hours, then skim the Metrics Radar Monday to decide the next tests. You keep approvals; the machine keeps cadence.
What to watch next
- More “knobs” in settings. Expect Instagram to expand visible interest/topics controls so people feel in charge — even as recommendations stay default-on. (Androidsis)
- Inbox surfaces. New DM-native discovery patterns (group prompts, save-to-DM folders, “send to…” nudges) that reward forwardable formats. (Popular Science)
- Category shifts. After re-opening the tap on political content with controls, other “sensitive or fuzzy” categories may follow a similar pattern. (The Verge)
The bottom line
Instagram’s discovery engine now optimizes for what you might love, not who you follow — and it’s increasingly listening to private shares to decide. You don’t beat that by begging for follows; you beat it by making DM-worthy posts, packaging for cold audiences, and posting at your proven hours so the algo has the best possible signal to work with. Use the platform’s new controls to keep your own feed sane, and use your analytics to keep your content machine sane.
If you want us to fold this into your weekly plan: we can spin up a 2-week test with (1) two DM-first formats, (2) Peak-Hour scheduling, and (3) a “shares & saves” KPI. You’ll know in 10 days if the recommendation engine is working for you — or what to fix next.
Sources
Read more from Rkive AI
About the author
Alberto Luengo is the founder and CEO of Rkive AI, a leading expert in AI for content automation and growth. He shares real-world insights on technology, strategy, and the future of the creator economy.
RKIVE AI