In a 15 November 2026 report, Social Media Today says LinkedIn recently updated the classifiers that judge how
relevant each comment is, and now flags suspected AI comments as less valuable. The change was explained by LinkedIn's
Chief Product Officer, Hari Srinivasan. If you comment on behalf of a brand or a
client, the question is no longer "can I use AI?" but "does this comment say anything?".
What changes
- Comments are not deleted. They are re-ranked: in the default "most relevant" view, suspected AI comments show up
less.
- In testing, views of these comments dropped by more than 20%.
- Notifications change too. People who post are alerted to the comments they are likely to care about, not to
every comment.
- It has not reached everyone yet. LinkedIn says it will roll out to everyone "soon", with no date.
It follows an earlier move: in August, LinkedIn added an option to report posts as low-quality AI content, and more
than one million people used it within the first three weeks.
How to spot a generic comment
One simple test: would the comment work under any other post? If it would, it is generic, whether AI wrote it or
not.
- "Great share! 🙌"
- "Really interesting, thanks for this."
- "Totally agree, this is spot on."
- A three-sentence summary of the post, ending with an emoji.
None of these adds anything for the people reading the thread, and that is what the re-ranking goes after.
Three rules for commenting as a brand
- A point of view. What the brand thinks, and why, in one sentence. "In our industry it's the other way round:
customers ask for fewer options, not more."
- A fact or an experience. Something only the commenter knows: what they tested, what they saw happen, a number
from their own work that can be shared.
- A specific question. One that makes the author reply with more than a thank-you. "Does this apply to company
pages too, or only to personal profiles?"
A good reply has one of the three. Two beats most of the comments on LinkedIn today.
Where AI still helps
Use AI for the draft, not for the reply. Uses that work:
- Ask for three possible angles on a post and pick one.
- Ask it to shorten a comment that already has a point of view.
- Ask it whether the comment would work under any post (the test above).
What does not work is pasting the post and publishing whatever comes out. That is exactly the comment LinkedIn wants
to show less.
Next step
Pick three posts from people in your client's industry and write one comment for each using the three rules. If you
use Andie, ask for angles, not finished comments: "give me three possible points of view on this post, one of them
contrarian".
Source: Social Media Today, "LinkedIn cracks down on AI-generated comments", 15/11/2026.