LinkedIn has changed how comments are displayed under posts in the feed, ordering them by how relevant each comment is to the individual person reading it rather than presenting one fixed order to everyone. Social Media Today reported the change on 9 August 2026 and said LinkedIn is using signals including professional interests, connections and engagement activity to decide which comments a member sees first.
The same report says LinkedIn will also surface more current, timely discussions in the feed, with the stated aim of pulling more members into replying instead of scrolling past. Taken together, the two updates move the comment section from a shared, ordered list into something closer to a per viewer ranking, where the comment at the top depends on who is looking.
What LinkedIn changed in the feed
Social Media Today reported on 9 August 2026 that LinkedIn now orders the comments on a post by relevance to each individual viewer. The report does not describe the change as a test, and it does not describe the previous ordering as switched off. What it states is the new behaviour: relevance to the reader determines what appears, and LinkedIn is pushing more recent and active discussions into view.
The practical consequence is easy to state and hard to measure. Two people opening the same post at the same moment can be shown a different comment first, because the signals LinkedIn named are properties of the reader rather than properties of the comment. Nothing in the report says a comment can be pushed out of view entirely, and nothing says every comment remains reachable, so treat both as unconfirmed.
The signals LinkedIn named, and the ones it did not
According to the Social Media Today report of 9 August 2026, LinkedIn is using signals including professional interests, connections and engagement activity. Each of those is worth reading literally.
- Professional interests point at topical alignment between the commenter and the reader. The report does not say how LinkedIn determines a member's professional interests, or whether the interests are declared by the member or inferred from behaviour.
- Connections point at graph proximity. A comment from someone inside a reader's network plausibly ranks higher for that reader than an identical comment from a stranger, though the report does not confirm that first degree connections outrank second degree ones.
- Engagement activity is the least specific of the three. The report does not say whose activity is being measured, the reader's, the commenter's, or both.
The word used in the report is "including", which means the list is not presented as complete. No weighting between the signals is given. Anyone claiming to know which of the three matters most is guessing, and that guess is not in the source.
The engagement numbers LinkedIn put behind the change
Social Media Today cited LinkedIn's Q2 performance report alongside the announcement, and set it against an outside analysis of how many comments are machine written. The figures below are the ones named in the report of 9 August 2026, with nothing added.
| Measure | Figure reported |
|---|---|
| Time spent in post comments, per LinkedIn's Q2 performance report | Up 18% year over year |
| Overall content consumption, per the same Q2 report | Up 10% year over year |
| LinkedIn comments found to be entirely AI-generated, April to June 2026 | About 30%, from a Pangram Labs analysis of 57,000 public posts |
The first two figures are LinkedIn's own reporting on its own product. The third comes from an outside detection analysis and is not LinkedIn's number, so the table is a reading convenience rather than a claim that all three were measured the same way.
The AI comment problem sitting underneath the update
The Social Media Today report of 9 August 2026 pairs the ranking change with an AI-detection analysis finding that roughly 30% of LinkedIn comments posted between April and June 2026 were entirely AI-generated, based on Pangram Labs work covering 57,000 public posts. The report also notes that LinkedIn is fighting engagement pods and AI spam in comment sections at the same time.
What follows from the source is that a large share of comment volume in that window was machine written, and that LinkedIn treats inauthentic comment activity as a live problem. What does not follow is that the relevance ranking update was built as an anti-AI measure. The report presents the two as related context rather than cause and effect, and LinkedIn is not quoted saying the ranking change targets AI comments.
Why commenting first loses value under per viewer ranking
What follows is reasoning about mechanics, not a claim about anything LinkedIn said.
Much of B2B commenting practice was built on one assumption: that the comment section is a single ordered list, so being early in it, or being the most liked entry in it, buys visibility for everyone who arrives later. Under personalised comment ordering that assumption stops holding, because there is no single list to be at the top of.
Engagement pods degrade under the same logic. A pod concentrates fast reactions from a fixed group of accounts onto a post, which lifts the entry in a shared ranking. If the ranking is computed per viewer using professional interests, connections and engagement activity, a burst from twenty accounts with no topical relationship to a given reader has a weaker path to lifting anything for that reader. That is an inference from the signals named in the report, not a finding the report states.
A generic AI-written comment sits worst of all under this reading. It carries no genuine relationship to the reader and now competes in a ranking that weighs relevance rather than recency. Set that against the roughly 30% AI-generated share found in the April to June 2026 window and the arithmetic is unkind: the format with the least reader specific value is also the most crowded.
What a B2B marketer should actually change
There is no setting to adjust here. LinkedIn did not, according to the report, ship a control that lets a page or a member choose comment ordering, so the work is in behaviour and measurement rather than configuration.
- Stop reporting comment position. If a monthly report contains a line like "our comment ranked second on the post", retire it. Under per viewer ranking that number describes what one person saw, and the person who saw it was usually the one taking the screenshot.
- Move the metric to responses. Replies received on your comment, profile visits from the post, and connection requests in the following days are all observable in LinkedIn's own analytics and notifications, and none of them depend on a shared position.
- Comment where there is a real connection to the audience. The signals named are professional interests and connections. Employees commenting inside their own professional area, on posts from people they actually know, line up with those signals. A company-wide instruction to comment on everything does not.
- Expect uneven visibility. A comment that draws replies from relevant people now beats a comment sitting under a large post with no reply, because the second may never have been shown to the audience you cared about.
- Audit any automated commenting in use. With roughly 30% of comments in a three month window found to be entirely machine written, generic AI replies are a crowded, low differentiation format whatever ranking LinkedIn applies.
The same logic applies to how a page's own content marketing programme is measured: comment counts on a post remain countable, but they no longer describe a single distribution outcome shared by every reader.
What the source did not say
The gaps here matter more than usual, because the change alters how a public metric behaves.
- No rollout date. The report of 9 August 2026 does not say when the ranking change went live, or whether it is fully deployed.
- No markets. No country, region or language is named anywhere in the report.
- No weighting. The relative importance of professional interests, connections and engagement activity is not disclosed.
- No confirmation the old ordering is retired. The report does not say whether any chronological or top comment view survives.
- No detail on account types. Nothing says whether company pages, newsletters or member posts behave differently.
- Limited detail on the AI figure. The roughly 30% share is attributed to a Pangram Labs analysis of 57,000 public posts covering April to June 2026. The report does not set out the detection method, the error rate, or how posts were sampled.
The honest limits of this finding
Personalised ranking is difficult to verify from outside. Any marketer checking a post sees exactly one ranking, their own, and cannot separate a change in LinkedIn's system from a change in their own signals. That makes anecdotal reports either way close to worthless as evidence. The Q2 figures also predate the change, so they describe the conditions LinkedIn was responding to rather than any effect of the update.
What this means for Thai marketers
The report names no markets at all, so there is no basis for saying this applies in Thailand sooner, later or differently. What follows is reasoning about implications, clearly separated from what the source states.
Thai B2B pages frequently post in both Thai and English, sometimes in the same post. If comment ranking leans on professional interests and connections, then a Thai language comment on a bilingual post plausibly reaches Thai speaking readers in the commenter's professional area more reliably than it reaches a wider English speaking audience, and the reverse holds too. That is an inference from the named signals, not something LinkedIn or Social Media Today said about Thailand.
The more portable point is about reporting. Teams in Bangkok that show comment screenshots as proof of reach are showing one viewer's ranking, and if visibility varies per reader, a screenshot is evidence of what the screenshotter saw and nothing more. Anyone running social advertising alongside organic posting already has the cleaner comparison: paid delivery reporting states how many people were served something, while a comment position never did.
FAQ on LinkedIn comment ranking
Does this change apply in Thailand?
The source does not say. Social Media Today's report of 9 August 2026 names no countries, regions or languages, so there is no reported basis for saying whether Thailand is included, excluded or on a different timeline. Anyone stating otherwise is going beyond the report.
Do I need to change anything in my LinkedIn settings?
No, because no setting was announced. The report describes a change to how LinkedIn ranks comments, not a control given to members or pages. The changes worth making are to how comment performance is reported internally and to where your team spends its commenting time.
Are engagement pods now useless on LinkedIn?
The source does not say that, and it should not be claimed. What the report does say is that LinkedIn is fighting engagement pods and AI spam in comment sections, and that comment ordering is now personalised per viewer. Reasoning from those two facts, coordinated reaction bursts have a weaker route to lifting a comment for a reader who has no topical or network relationship to the pod, but no measurement of that effect exists in the report.
Was 30% of LinkedIn really written by AI?
The reported figure is that roughly 30% of comments posted between April and June 2026 were entirely AI-generated, from a Pangram Labs analysis of 57,000 public posts, as cited by Social Media Today. That covers comments in a specific three month window, not LinkedIn content overall, and the report does not publish the detection methodology or its error rate.
Where this leaves comment strategy
The change reported by Social Media Today on 9 August 2026 does not make commenting less valuable. It makes the value harder to fake and harder to see. A reply that is genuinely relevant to a specific professional audience gains from a system that ranks on professional interests and connections. A reply produced at volume to occupy a position loses the position it was produced for.
If your LinkedIn reporting currently leans on comment counts and comment placement as proof that a programme is working, this is a reasonable moment to rebuild those reports around outcomes that survive personalisation. Relevant Audience works with B2B teams on exactly that kind of measurement and content planning, and the full report is worth reading in the original on Social Media Today.
