TL;DR: Meltwater analyzed 9.5 million citations of LinkedIn content inside ChatGPT, Perplexity and Gemini answers. 75% came from personal profiles. The same week, LinkedIn confirmed over 1 million people used the "Seems like AI slop" button, flagged posts lose roughly 40% of out-of-network reach, Search Engine Journal showed organic reach down 47% year over year, and Metricool's 673K-post study found a save is worth 5x a like and that a link in the post costs personal profiles 27% of reach while it adds 51% on company pages. Quietly, LinkedIn is also moving SSI out of the free account. Five shifts, and what to do about each one this week.

Seven new LinkedIn numbers landed on my desk this week, which is a lot for one week. Usually each one tells a different story. This time, once I sorted them by topic, every single one pointed at the same place.

The feed punishes content that sounds like everyone else. AI engines barely see company pages. What's left between those two facts is a personal profile with a voice you can recognize.

Five shifts, ordered by how much they matter, not by date.

1. The number of the month: 75% of citations come from personal profiles

On August 17, Meltwater published a study of 9.5 million citations of LinkedIn content inside AI engine answers. Three quarters of those citations came from personal profiles (a citation, for the purposes of this article, is when an AI engine uses your post or article as a source for its answer, usually with a link). Company pages barely register.

Think about what that means in practice. When someone asks ChatGPT "who's a good marketing expert for cybersecurity SaaS" or "how do you build an employee advocacy program", the answer gets built from posts and articles by people, with a name and a face, and almost never from what the marketing manager wrote in the brand's name on the company page.

I work with CEOs and founders on their LinkedIn, and for some of them I also write the posts (ghostwriting), so the number didn't surprise me and it makes sense to me. That's exactly why I checked it twice before putting it here. It lines up with what I saw from Semrush in March (89K LinkedIn URLs cited by AI, most of them people's posts) and with LinkedIn's own guidance from June, which talked about articles and posts from personal profiles, never about pages.

The strongest counterargument: LinkedIn has far more personal profiles than company pages, so of course they get cited more. True. But even if you normalize by content volume, a company page posting three times a week should show up somewhere, and it doesn't. The bias is more than a numbers game.

What to do now: If your client's strategy is a company page and nothing else, this is the week to open the conversation about the CEO's profile. The number does half the work for you. And it doesn't have to stop at the CEO. A post from an engineer or a salesperson is a personal profile the AI engines can see too. How much reach can an advocacy program like that add? I built an advocacy calculator that does the math in a minute.

2. A million reporters, 40% less reach, and an alert that didn't exist before

Three weeks ago I wrote here that the "Seems like AI slop" button trains models and doesn't cut reach. Still true, but now there are numbers on the other side of the equation.

LinkedIn officially confirmed that over 1 million people used the button in its first two weeks. Flagged posts get roughly 40% less reach outside the poster's network. Hari Srinivasan, Chief Product Officer, stressed that a single report does nothing, it takes several reports stacking up.

The genuinely new part: whoever gets flagged now sees a private alert in their post analytics, "Some members told us this post seems like AI." Until now there was no channel where LinkedIn told a poster what the audience thought of their text. Now there is, and for clients it becomes a metric to check like any other.

It touches the budget too. An organic post that reads as AI-generated gets pushed down before you ever promote it, and a Thought Leader Ad built on that post simply performs worse. The quality of the original post is now a threshold condition for paid spend as well.

What to do now: Go through everything scheduled for the next two weeks and filter out the generic templates. What survives the filter: specific experience, names, numbers, an opinion. And if that alert shows up on a client's post, it's a signal to stop: the next post gets written from scratch, without the template that earned the report.

3. Organic reach is down 47%, post clicks are down 42%

On August 19, Search Engine Journal published an analysis of the year since the feed moved to 360Brew, a 150-billion-parameter language model that ranks by expertise and dwell time instead of reaction counts. Organic reach fell 47% year over year, post clicks fell 42%. Company pages took the hardest hit.

This is a structural shift that keeps rolling, and now the numbers are on the table. Connect it to section 1 and you get a consistent picture: the feed and the AI engines reward the same thing, a specific person who knows something specific.

What to do now: Stop comparing impressions to last year, that comparison no longer tells you anything. Track this instead: new followers per week, and DMs that mention a post.

4. What works now, according to 673K posts

Metricool compared 673K posts from 63K accounts, early 2025 versus early 2026. A few findings I'm already applying:

• Carousels lead by a wide margin: 1,451 impressions per post on average, 3.71% engagement, 11x the interactions of a single image. Video trails at 605 impressions and 1.8%.

• A direct question in the post brings 77% more comments, an explicit call to comment brings 80% more.

• A save is worth 5x a like and 2x a comment in algorithmic weight.

• "Invisible interactions", carousel swipes, video replays, "see more" clicks, now count as engagement signals. The like is becoming a weak metric.

• 50% of impressions arrive in the first 48 hours, peak window 9:00 to 12:00.

And the finding that surprised me: external links. On a personal profile, a link in the post body cuts reach by 27%, so it still goes in the first comment. On a company page, a link in the post actually adds 51% impressions. The exact opposite, which means the "link in the comments" rule we all recite is right for only half the assets we manage.

What to do now: Build your next post as a carousel with a framework or a checklist (something worth saving), a direct question at the end, link in the first comment on a personal profile and in the post body on a company page.

5. Four small updates, and one LinkedIn didn't announce

AI Event Highlights rolled out worldwide on August 19. The system spots strong moments in a LinkedIn Live recording, announcements, quotes, insights, and offers them as short clips inside the event management tool. Anyone running a Live can now get 3 to 4 clips out of one recording with no editing.

Personalized comment ranking keeps rolling out. Comments in the feed are ordered by relevance to the viewer rather than chronologically, so a comment from someone the reader knows is worth more for distribution than a generic comment from a stranger. The backdrop: time spent in comments is up 18% year over year.

LinkedIn's Marketing API changed versions. On August 17, LinkedIn shut down the old version (v202508), and every third-party tool that didn't update in time, schedulers, dashboards, campaign systems, simply lost its connection. If one of your tools is acting strange this week, check this before you go hunting for a bug.

Paid creator events are running now as a pilot with 50 creators, scaling to 1,000. Participants include Cassie Kozyrkov, Codie Sanchez and Chris Do. I covered the model in "LinkedIn Wants Your Money", the difference is that in July it was a plan and now it's happening.

Now for the one nobody announced. I opened my SSI page this week and got a notice: "Your access to the Social Selling Index (SSI) page will be discontinued soon. SSI is a Sales Navigator feature." SSI has been free since 2014, and it's becoming a Sales Navigator feature. I'm not a huge fan of SSI as a metric, it mostly measures how much you use LinkedIn, but it was convenient to screenshot at the start and end of a program. Screenshot it now, before it's gone.

What to do now: Confirm every connected tool is on a current API version, capture SSI for every client, and if you have a Live coming up, try the automatic clips.

The bottom line

LinkedIn has made the community part of the algorithm. A million people flag generic content, the feed cuts it by 40%, AI engines skip company pages, and meanwhile the platform moves one more metric behind a paywall. The only asset that both the feed and ChatGPT reward this week is a personal profile that writes like a specific human being.

If all you've run so far is a company page, which number from this article are you taking into your next conversation with the CEO?

FAQ

What share of LinkedIn citations in AI answers come from personal profiles?

According to Meltwater's August 2026 study of 9.5 million citations across ChatGPT, Perplexity and Gemini, 75% come from personal profiles. Company pages are almost never cited.

How many people have reported AI slop on LinkedIn?

LinkedIn confirmed over 1 million members used the "Seems like AI slop" button in its first two weeks. Posts flagged by several reporters get roughly 40% less out-of-network reach, and a single report has no effect.

Will I be notified if my post is flagged as AI?

Yes. Since August 2026, anyone whose post is flagged sees a private notice in the post's analytics: "Some members told us this post seems like AI."

How much has LinkedIn organic reach dropped?

Per Search Engine Journal's August 2026 analysis, organic reach is down 47% and post clicks are down 42% year over year since the move to the 360Brew model. Company pages were hit hardest.

Which post format performs best on LinkedIn in 2026?

Per Metricool (673K posts, 63K accounts), carousels lead with 1,451 average impressions and 3.71% engagement, 11x the interactions of a single image. A save is worth 5x a like, and a direct question brings 77% more comments.

Should I put a link in a LinkedIn post?

Depends on the account type. On a personal profile, a link in the body cuts reach by 27%, so the first comment is better. On a company page, a link in the post adds 51% impressions.

Is LinkedIn's SSI still free?

Not for long. In August 2026 LinkedIn began showing a notice that access to the SSI page will be discontinued and the metric is becoming a Sales Navigator feature.

Written by Hanita Yudovski. I run LinkedIn systems for B2B companies, from the CEO's profile to employee advocacy programs, and help independents build their presence on LinkedIn. LinkedIn is the field I work in every day, and the AI agents and tools I build myself are how the work gets done. Updated August 2026. Written with AI tools, with human strategy and editing.

Sources: Social Media Today: 1M AI slop reports · Slashdot: the AI slop button · Search Engine Journal: reach down 47% · Social Media Today: Meltwater study · Metricool: LinkedIn Trends 2026 · Social Media Today: AI Event Highlights · Social Media Today: comment ranking · Microsoft Learn: Marketing API recent changes · Social Media Today: creator events