Published Friday, 25 September 2026 | Updated Wednesday, 21 October 2026 | By LadyBugz Marketing
The LinkedIn algorithm in 2026 runs on a single AI (artificial intelligence, technology that can weigh information and make a decision without a person doing it by hand) system, a unified Generative Recommender (GR) model, which replaced the separate ranking models LinkedIn used to run its feed, search and recommendations. It now decides how far a post travels mainly by dwell time (how long someone actually stays on your post before scrolling past) and the quality of the comments it gets, not by likes, hashtags or follower count. For South African B2B (business to business, meaning you sell to other companies rather than individual consumers) marketers, the short version for 2026 is that organic reach has thinned out across the board, and the businesses still winning are the ones publishing something useful, consistently, under a real person’s name.
Key Insights
- One unified system now runs the whole platform. A single Generative Recommender (GR) model replaced LinkedIn’s old patchwork of separate feed, search and recommendation models (LinkedIn Engineering Blog, “Engineering the next generation of LinkedIn’s Feed,” 12 March 2026).
- Dwell time is the strongest signal LinkedIn now uses. Posts that hold a reader’s attention for 60 seconds or more see engagement rates around 15.6 percent, against 1.2 percent for posts read in under 3 seconds (Richard van der Blom’s 2026 LinkedIn algorithm research, based on 1.3 million posts and 80,000-plus profiles).
- Comments now outweigh likes. The same research puts a thoughtful comment at roughly twice the ranking weight of a like.
- Reach has dropped, and unevenly. Independent analysis from Saywhat and The Creator Accelerator’s ongoing State of the Algorithm research (roughly 400,000 posts tracked) has found median impressions down by more than half from the 2023 peak, with a wide and growing gap between top-performing posts and average ones.
- Hashtag stacking now hurts you. Posts stacked with more than three hashtags perform noticeably worse than posts using few or none, according to the same Saywhat and The Creator Accelerator research.
- Google now puts more links inside its AI answers. On 15 October 2026 Google said on its official blog, The Keyword, that it is showing inline links (clickable words and phrases inside the answer itself) more often, and higher up, in AI Overviews (the AI-written summaries above some Google results) and AI Mode (Google’s chat-style search).
What actually changed: one AI system instead of a patchwork
What is the LinkedIn algorithm in 2026 and how does it work?
In 2026, “the algorithm” is a single AI model, a unified Generative Recommender (GR) system, that reads your profile, reads your post, and watches how people behave after they see it, then decides whether to keep showing that post to more people. It is not a filter you can trick with a format hack. LinkedIn’s own engineering team confirmed this in detail on the company’s engineering blog on 12 March 2026, explaining that the unified GR system replaced the separate models LinkedIn had built over the years for feed ranking, search and recommendations (LinkedIn Engineering Blog, “Engineering the next generation of LinkedIn’s Feed,” Hristo Danchev, 12 March 2026).
Instead of scoring a single post as good or bad in isolation, the system builds a longer pattern of what a specific person tends to stop for, comment on, and act on, then matches your content against that pattern. I have watched clients ask us for years to “crack the algorithm,” as if one lever existed somewhere. There was never a lever. There is now, more than ever, a model watching for lasting usefulness over time. That is a harder game for shortcut chasers, and a fairer one for anyone who actually knows their subject.
The two signals doing the heavy lifting
What is dwell time and why has it become the top ranking signal?
Dwell time is simply how long someone stays on your post before moving on, and it is now the strongest signal LinkedIn uses to decide whether content deserves wider reach. Richard van der Blom’s 2026 LinkedIn algorithm research, which analysed 1.3 million posts and more than 80,000 profiles, found posts holding a reader’s attention for 60 seconds or longer see engagement rates around 15.6 percent, against just 1.2 percent for posts read in under 3 seconds.
That gap is not small. A post someone actually reads properly can outperform a fast scroll-past like-and-move-on by more than tenfold. Length is not the point here, holding attention is. A sharp 400-character post that makes someone stop is worth more than a rambling 2,000-character one nobody finishes.
Do comments matter more than likes now?
Yes. Van der Blom’s research also found comments now carry roughly twice the ranking weight of likes, while the system increasingly discounts low-effort engagement. A one-word “Great post!” from an obviously templated pattern counts for very little. A real reply that adds a view, a question or a respectful disagreement counts for a lot more, because it signals actual interest and keeps a post alive in front of more people over several days rather than just the first hour. This is also why personal branding and social selling now define B2B success on LinkedIn far more than a polished company page does.
Why your reach probably feels like it has dropped
Is LinkedIn organic reach actually declining in 2026?
Yes, and it is not just a feeling. Independent analysis from Saywhat and The Creator Accelerator’s ongoing State of the Algorithm research (roughly 400,000 posts tracked) has found median impressions down by more than half from the 2023 peak, with a wide and growing gap between top-performing content and the average post. The platform is not shrinking evenly. It is concentrating attention on fewer, better posts.
For a South African business used to LinkedIn sending a steady trickle of profile views and inbound messages, that can look alarming on a graph. In our own work running LinkedIn for B2B clients here, what it actually means is that the cost of generic content has gone up sharply, while the reward for something specific and well-written has gone up too. The real cost of staying quiet on LinkedIn has only grown as reach concentrates on active, consistent voices rather than spreading evenly across everyone posting.
We keep the exact mix and cadence we build for individual clients close, because that is the work they pay us for. But the principle behind it is simple enough to say plainly: fewer, sharper posts written by a real person now beat a high volume of generic ones, every time, under this system.
Formats and habits that still hold up
Does the algorithm favour certain content formats?
Some formats still travel further, but format alone no longer decides reach on its own. Carousels and document posts tend to out-reach plain text posts, according to the same Saywhat and The Creator Accelerator research. The gap exists mostly because a well-built carousel naturally holds someone’s attention for longer, which feeds straight back into the dwell time signal above. A carousel nobody reads past slide one earns nothing extra from the format itself.
Do hashtags still help LinkedIn find the right audience for a post?
Not the way they used to. LinkedIn’s engineering blog confirms the platform now reads the actual topic and meaning of a post rather than relying on hashtags to route it (LinkedIn Engineering Blog, 12 March 2026), and the Saywhat and The Creator Accelerator research found posts stacked with more than three hashtags performed noticeably worse than posts using few or none. One or two hashtags will not hurt you, but stacking them is now closer to a penalty than a boost.
How often should a South African B2B company post on LinkedIn?
Current 2026 practitioner guidance across several algorithm studies lands in roughly the same place: three to four solid posts a week beats either daily posting or one post a month (2026 LinkedIn algorithm practitioner research, treat as directional, exact figures vary slightly by study). Posting daily with nothing worthwhile to say tends to drag reach down rather than up, because a run of forgettable posts teaches the system to expect less from your profile. One properly considered post a week is a reasonable floor if that is truly all the capacity a small team has.
Google’s October change: your LinkedIn writing can be the link inside the answer
What did Google change about links in its AI answers in October 2026?
Google’s October 2026 change puts more inline links, the clickable words inside an AI answer, into AI Overviews and AI Mode, and higher up. On 15 October 2026 Google said it had “made continued and substantial increases to the presence of inline links.” Google is also testing a section for the one or two official websites for a task.
Does Google’s AI search cite LinkedIn content?
Google’s AI Overviews and AI Mode do cite LinkedIn content, and long-form LinkedIn articles make up most of what gets cited. OtterlyAI, a company that tracks AI search visibility, followed over 1.3 million LinkedIn citations across six AI tools from January to June 2026. Google’s two AI features accounted for 31.2 percent of them (the company’s own data).
Across all six tools, LinkedIn articles made up 72.2 percent of the LinkedIn content cited, against 26.1 percent for ordinary posts. The same study found likes and comments had almost no link to whether a piece was cited. That sits oddly next to everything above about comments. It is not a contradiction. Your writing now has two audiences. The LinkedIn feed rewards dwell time and thoughtful comments. Google’s AI answers need a clear statement to attach a link to. Our read is that because an inline link sits on a phrase, a post or article that states one precise point in plain words (a number, a named method, a clear definition) gives Google something to link. A vague post about “the future of work” gives it nothing.
The practical step is small. Write the key sentence of each LinkedIn article so it still makes sense lifted out on its own, and turn your strongest posts into longer articles. Keep expectations level, though. Google has published no click figures for the change. Clicks from AI Overviews and AI Mode are counted inside the normal Search Console Performance report but not shown separately, according to Google Search Central’s “AI features and your website” guide. Google says it will “introduce additional metrics over time”, as Search Engine Roundtable reported on 15 October 2026. More links is good news. It is not yet a measured traffic gain.
What this means if you are running B2B marketing from South Africa
None of the mechanics above change because you sell into South Africa rather than into the US or the UK. The algorithm does not know or care where your buyer sits. What does matter locally is that South African B2B buying cycles are already long and relationship-led, so a platform that now rewards depth over noise plays to a strength local sellers already have, if they actually use it. We have watched a single well-argued post about a live industry problem earn more replies from local executives than months of generic “thrilled to announce” updates ever did.
If you are still weighing whether LinkedIn deserves that priority over other channels, our comparison of LinkedIn versus Facebook for B2B marketing in South Africa lays out where each platform actually earns its place.
The businesses we see doing well right now treat LinkedIn as a place to demonstrate real judgement, not a billboard. That lines up with where AI search is heading too, since the tools buyers now ask questions to lean on the same kind of specific, well-written LinkedIn content. How B2B marketers are getting cited in AI search through LinkedIn covers that shift in more depth, and it is worth reading alongside this piece if AI visibility is on your radar. If your content strategy is still running on last year’s assumptions, your 2027 LinkedIn results start now is the straight version of what needs to change. And if you are wondering whether AI-written filler is working against you, LinkedIn’s new AI slop button explains what LinkedIn itself is now doing about that.
This same shift is what we track under answer engine optimisation, since the specific, well-written content LinkedIn now rewards is exactly what AI answer engines pull from when they cite a source.
The bottom line
The 2026 algorithm is less forgiving of noise and more generous to deep expertise, delivered consistently under a real name. That is a harder standard to fake, and an easier one to meet properly if you are willing to do the work. Our founder has been active on LinkedIn since 2009, watching the platform shift under South African businesses’ feet the whole way, and this is the first change we have seen that actually rewards patience and knowledge over volume and spend.
If building that consistency in-house is not realistic right now, our LinkedIn marketing management service handles the posting rhythm so it never slips.
Frequently asked questions
What are inline links in Google’s AI answers?
Inline links are clickable words or phrases inside the text of a Google AI Overview or AI Mode answer, rather than in a list of sources at the end. On 15 October 2026 Google said it is showing them more often and higher up, so readers can click through to a source while they read.
Can a LinkedIn article be linked in a Google AI answer?
Yes. Public LinkedIn articles and posts are cited in Google’s AI Overviews and AI Mode. In OtterlyAI’s 2026 study of over 1.3 million LinkedIn citations across six AI tools, long-form LinkedIn articles made up 72.2 percent of the LinkedIn content cited, against 26.1 percent for ordinary posts.
Do likes and comments help content get cited by AI search?
Not on their own. OtterlyAI’s 2026 study found likes and comments had almost no link to whether LinkedIn content was cited by AI tools. Comments still matter for reach in the LinkedIn feed, so they remain worth earning, but the study found no sign that they influence which pieces AI answers cite.
Our LinkedIn training shows your team how the 2026 algorithm ranks content, and how to earn reach and engagement without ad spend, built around your business and your audience.
Talk to us: noleen@ladybugz.co.za
LadyBugz Marketing
Editorial note
This article reflects what is publicly documented about LinkedIn’s algorithm as of late September 2026, with an update on 21 October 2026 for Google’s change to links in its AI answers. It draws on LinkedIn’s own engineering blog, independent research from Richard van der Blom and Saywhat with The Creator Accelerator, Google’s own announcement and OtterlyAI’s citation study. LinkedIn does not publish every detail of how its GR model weighs each signal, and the platform keeps tuning it, so the numbers here are a point-in-time read rather than a permanent scoreboard. What is unlikely to change is the underlying logic: usefulness, consistency and a real human voice keep winning over shortcuts.
Keep reading
