In August 2026 LinkedIn made two quiet changes to its feed, both about comments: it now ranks the replies each person sees by how relevant they are to that individual, and it surfaces more timely, active discussions in the feed to pull people into threads. The driver is LinkedIn's own Q2 data — an 18% year-over-year rise in time spent in post comments and roughly 10% growth in overall content consumption. Read together, the changes formalize a shift practitioners had already felt: the comment, not the like, is now the unit of distribution, and a post that starts a real back-and-forth thread travels further than one that collects reactions. That reprices content strategy. The highest-ROI post is no longer the one written to be admired and scrolled past — it is the one written to be replied to. This guide is about that reprice: the mechanics of why conversation now drives reach, what 'reply-driven content' actually means once you strip out the engagement-bait version LinkedIn's slop filter now punishes, the specific formats and post structures that reliably open a thread, the AI-comment problem sitting underneath the rosy numbers, and the part nobody mentions — that writing to provoke genuine conversation is a supply problem, because you need a steady stream of stake-a-position, end-on-a-real-question posts, not one lucky viral thread.
In August 2026 LinkedIn made two changes to its feed that read like housekeeping and are actually a reprice of content strategy. Both are about comments. The first is a comment-ranking update: instead of showing a post's replies in a roughly chronological or most-liked order, LinkedIn now ranks the comments each person sees by how relevant they are to that individual, using signals like professional interests, connections, and past engagement. The second surfaces more timely and relevant discussions directly in the feed, nudging members toward active threads rather than passive scrolling. The updates were flagged by creator-economy analyst Lindsey Gamble and reported by Social Media Today around August 9, 2026. The full announcement detail is in the news write-up of the LinkedIn feed comment update.
The reason to treat this as strategy rather than trivia is in LinkedIn's own numbers. Its Q2 performance report cited an 18% year-over-year increase in time spent in post comments, alongside roughly 10% growth in overall content consumption. Comments are where LinkedIn is seeing its most engaged behavior, so the company is reshaping the feed to feed that loop — surface the discussions most likely to draw someone in, and rank the replies so the useful ones rise. When a platform reorganizes its feed around a behavior and tells you it did so because that behavior is where the time is going, the message to creators is unambiguous: the comment section is now a distribution surface, and content that fills it is what the algorithm wants to spread.
Reach on LinkedIn has always been an early-signals game — you publish, the post goes to a slice of your network, and what that slice does in the first hour decides whether distribution widens. What the 2026 changes do is sharpen which early signal counts most. A like is cheap: one tap, no thought, no dwell. A comment is expensive: it takes attention, an opinion, and time in the post, and it often pulls the commenter's own network toward the thread. The feed has always weighted the expensive signals higher because they are harder to fake and correlate with genuine interest; the new updates make that weighting explicit by pulling active threads into more feeds and ranking the replies inside them for relevance. This is the mechanism behind the long-standing practitioner claim that comments are among the highest-ROI moves on the platform — LinkedIn just wired it deeper into the feed.
There is a compounding effect that makes conversation especially valuable. A thread is not a single event; it is a series of them. Every reply is a fresh engagement signal, every reply can resurface the post in the commenter's network, and a back-and-forth keeps a post alive for hours instead of the minutes a like-only post gets. So a post that opens a real discussion does not just clear the early-signal bar once — it keeps re-clearing it as the conversation runs, which is exactly why threads out-travel reactions. Be precise about the claim, though: LinkedIn has not published a ranking formula, so specific multipliers you see quoted — 'comments count 2x likes' and the like — are third-party estimates, not confirmed mechanics. The direction is well-supported; the exact number is not, and a first-mover page should not pretend otherwise.
The phrase is easy to misread as 'engagement bait,' and getting that distinction right is the whole game, because the same update that rewards conversation also punishes the fake version of it. Engagement bait manufactures a reflex: 'comment YES for the free template,' 'tag someone who needs to see this,' 'agree? 👇'. It produces comment counts, but the comments are contentless — and LinkedIn's relevance ranking now pushes exactly that kind of low-value reply down, while its member-reported AI-slop button flags the posts that fish for it. Writing for the comment count is now a losing move; the feed reads the resulting thread as noise.
Reply-driven content is the opposite mechanism aimed at the same goal. It invites a real answer to a real question inside your area of expertise, so the thread that forms is specific, useful, and legible to a relevance-ranking model as genuine discussion. The structural tell is simple: a reply-driven post leaves a deliberate opening — a stakeable position someone can push back on, a genuine question the reader actually has an answer to, a concrete trade-off you made and an ask for how others handled it. A reach-driven post, by contrast, is airtight — a clean, finished take that lands well and gives the reader nothing to say back. Both can be excellent writing. Only one is built to be replied to, and after August 2026 that one carries the higher ROI.
Reply-driven-ness is mostly a property of structure, not format, but some formats make the opening easier to leave. Below are the reliable ways to build a post that earns a substantive reply rather than a reflex tap — all of which assume the content underneath is specific and first-hand enough to clear the slop bar.
The most dependable conversation-starter is a specific, defensible point of view that a reasonable person could disagree with. 'Most onboarding emails are too long' invites nothing; 'we cut our onboarding sequence from seven emails to two and activation went up' invites a dozen people to tell you why that did or didn't work for them. The move is to take a real stance from real work and leave room for pushback — not a hot take for its own sake, which the knowledge graph and slop filter both read as low-authority, but a considered position with a soft edge someone can argue with.
Ending a post with 'thoughts?' is the tell of a post that has nothing to ask. A real question is one you actually want the answer to and that the reader is qualified to give: 'how are you handling attribution now that most reach is out-of-network?' works because it names a specific, current problem your audience is living. The distinction the feed rewards is whether the question opens a useful thread or fishes for a reflex — write the one you'd genuinely read the replies to.
Carousels are strong dwell-time objects, and they become reply-driven when the last slide opens rather than closes — a framework with a deliberate gap, a teardown that ends 'here's where I'm still unsure,' a data walk-through that asks what the reader is seeing in their own numbers. The swipe-through earns the dwell; the open ending earns the comment. A carousel that ends on a tidy conclusion is admired and saved but rarely discussed.
Short vertical and avatar video is the format LinkedIn is pushing hardest, and it converts to conversation when the creator poses a real question or stakes a position out loud rather than delivering a sealed monologue. Captions are non-negotiable because most of the feed is watched with sound off, which is the argument of a captions-first video strategy. The reply-driven version ends by handing the viewer something to answer, not by wrapping up neatly.
Across all four, the same second half applies: be present to reply early. The updates reward timely, active discussions, so showing up in your own comments while the feed is still surfacing the post keeps the thread alive at the moment it matters most. The opening you engineer into the post is only half of it — the conversation has to be worked while it is live.
One caveat belongs next to the rosy engagement figures. An analysis from AI-detection startup Pangram Labs estimated that roughly 30% of LinkedIn comments posted between April and June 2026 were entirely AI-generated. So part of the measured rise in comment activity may be automated rather than human — which is precisely why LinkedIn is leaning on relevance ranking in the first place. The relevance filter is, in part, a slop filter for the comment section: automated 'insightful post!' fluff is exactly what it is built to demote. That has a direct implication for how you write, and it is the opposite of what the automation crowd will tell you. Do not automate your comments. The value of a reply is that it is specific, experienced, and yours; an AI-generated comment is the thing the update is designed to bury, and shipping it risks the slop-report flag on your own account.
The broader context is that when everyone can generate volume, specificity becomes the only durable differentiator — the same conclusion reached in AI content saturation on LinkedIn and X. Reply-driven content is not exempt from that; it is the sharp end of it. A generic question opens a generic thread that the relevance ranking ignores. A specific question rooted in your actual work opens a thread the feed wants to spread. The AI-comment problem does not argue against conversation-first content — it argues for making both the posts and the replies unmistakably human.
Here is where the strategy meets reality. You cannot manufacture one viral thread and coast; the feed rewards a running behavior, not a single hit. Winning the conversation reprice means a steady cadence of posts, each engineered to leave a real opening, held to one voice and one lane so the knowledge graph keeps reading you as credible — and then the hours to actually work the threads those posts open. That is two distinct jobs with a shared time budget, and they compete. The reply is the part you cannot outsource; it has to be yours, in your voice, from your experience. So the pressure lands on the other job: producing a reliable supply of genuinely reply-worthy posts, week after week, without the quality sliding into the bait the feed now punishes.
This is the wall most conversation advice skips. 'Write posts that start discussions' is correct and useless at cadence, because engineering a real opening — a stakeable position, a question worth answering, an open-ended carousel — is harder to do repeatedly than writing a clean, finished take. Finished is the default; open is the deliberate craft. And it has to happen across formats, in one voice, on a rhythm consistent enough for the feed to learn your lane. The bottleneck is not knowing that comments matter. It is supplying the conversation-starters at the volume the reprice demands while keeping your own time free for the replies it actually pays out on. For where this sits in the fuller LinkedIn picture, the 2026 content playbook covers the format mix and the interest-graph reach playbook covers where distribution now comes from.
This split — outsource the supply, keep the replies — is the exact shape Kompozy is built for. Kompozy is a full AI content generation and multi-platform publishing engine, not a repurposing add-on: from a single source — a talk, a client call, a founder voice memo, a rough take — it generates net-new posts across five output buckets, which is precisely the reliable supply of conversation-starters the reprice demands. The difference from a generic content tool is that reply-driven content has a specific production spec, and Kompozy's brief system is where you encode it.
Every generation descends from one Persona Brief that fixes your voice, your point of view, and a banned-phrase list — and it is the natural place to instruct the engine toward the open ending rather than the airtight one: text posts that stake a defensible position and close on a genuine question, Carousel Posts rendered brand-exact through HyperFrames that end on an unfinished argument rather than a tidy summary, and Persona Shorts that pose a question to camera. Because the brief governs voice and specificity, scaling the supply sharpens your topic signal instead of drifting into the generic output the slop filter demotes — the posts stay unmistakably yours, which is the whole point when the feed is filtering for exactly that. You review each one behind a per-post gate and Autopilot schedules the set into your best windows, not only to LinkedIn but across the eight social platforms plus blog and email from one queue.
The honest boundary is the important part, and it is the opposite of what an automation pitch would promise. Kompozy does not — and should not — write your comments. The reply is the one thing the August 2026 update pays out on that has to be human, specific, and yours; automating it is the behavior the relevance ranking is built to bury. What Kompozy removes is the other constraint: the production ceiling that makes a steady cadence of genuinely reply-worthy posts impractical by hand, so the scarce resource becomes your time in the conversation rather than your time making things to have a conversation about. That is the correct division of labor for a feed that now pays for the thread — let the engine keep the openings coming, and spend your hours where the reach actually is. For the timing layer, the best time to post on LinkedIn data pairs with this, and content repurposing covers turning one source into the many conversation-starters a week needs.
Two things, both reported around August 9, 2026. First, LinkedIn now ranks the comments each person sees by relevance to them — using signals like professional interests, connections, and past engagement — instead of a roughly chronological or popularity order. Second, it surfaces more timely, active discussions directly in the feed to pull members into live threads. Both changes are aimed at driving more conversation, and both make the comment section a bigger factor in what gets seen.
In practice, yes. A post that generates a genuine back-and-forth thread tends to travel further than one that only collects reactions, because comments cost more effort and signal real interest — and the 2026 feed is explicitly reshaping around that behavior. LinkedIn has not published exact weightings, so treat specific figures like 'comments count 2x' as third-party estimates, not confirmed mechanics. The direction is clear even if the multiplier is not.
Content written to earn a substantive reply rather than a passive reaction — posts that stake a defensible position, ask a real question the reader has an answer to, or leave a deliberate open loop that invites people to add their own experience. It is the opposite of a clean, self-contained take that gives the reader nothing to say back. The key qualifier is 'substantive': LinkedIn's relevance ranking and its member-reported AI-slop signal both demote the low-effort 'Agree?' bait version.
No, and conflating them is the fastest way to get demoted. Engagement bait manufactures a reflex — 'comment YES for the template' — and the 2026 feed's relevance ranking pushes those generic replies down while the slop-report button flags the posts. Reply-driven content invites a real answer to a real question inside your expertise, so the thread that forms is specific and useful. Same goal (a conversation), opposite mechanism: one provokes a reflex, the other earns a contribution.
Give the reader something to answer or push back on. Stake a specific, defensible point of view rather than a safe summary; end on a genuine question you actually want the answer to, not a rhetorical 'thoughts?'; share a concrete decision or trade-off and ask how others handled it; and then be present to reply early while the thread is still being surfaced. The structure matters less than leaving a real opening — a finished, airtight post gives no one a reason to comment.
Reply-driven content has a different production spec than reach-driven content — every post has to leave a real opening, and it has to do that at a cadence, not once. Kompozy is an AI content generation and multi-platform publishing engine that produces that supply from one source: text posts that stake a position and end on a genuine question, brand-exact carousels, and avatar video, all governed by a Persona Brief you can tune to end open rather than closed. It keeps the conversation-starters coming across eight social platforms plus blog and email, so your own time goes to the replies the feed now rewards.
In August 2026 LinkedIn changed its feed in two ways, both about comments: it ranks the replies each person sees by relevance to them, and it surfaces more timely, active discussions to pull members into threads. The driver is its own data — an 18% year-over-year rise in time spent in comments and ~10% growth in consumption. The effect is a reprice: the comment, not the like, is the unit of distribution, so reply-driven content — posts that stake a position and end on a real question — now carries the higher ROI, provided it stays specific enough to clear LinkedIn's AI-slop filter.
Get started → · ← All guides · Compare Kompozy vs other tools