How to comment on LinkedIn so the right people see you
Short answer: comment where your audience is already standing, keep it to one sentence, and add something the post did not say. A comment under someone else's post reaches the audience they spent years building. That is the whole trade. Posting asks people to come to you. Commenting goes to them.
Why commenting outperforms posting for most people
If you have two thousand followers, a post reaches a slice of two thousand people who mostly already know you. A good comment under a post from someone with fifty thousand followers sits in front of a crowd you do not have access to any other way.
The person you want to reach is not searching for you. They are reading their feed. Commenting is the only reliable way to appear there without paying for it.
It is also the lowest risk activity on the platform. You are not sending anything to anyone. Nobody has to accept a request. Nobody gets an unsolicited message. The worst outcome is that nobody replies.
Comment where your audience already is
Most commenting advice stops at "engage more". That is useless because it does not say where.
The choice of post matters more than the choice of words. A brilliant comment under a post nobody read is worth nothing. A plain comment under the right post is worth a lot. Work out where your people gather before you write a word.
- Pick topics, not personalities. Following ten influencers puts you in the same comment section as everyone else doing the same thing. Searching a phrase your buyers use puts you somewhere less crowded.
- Go where the post is fresh. A comment on a two-day-old post is invisible. Comments sort by engagement, and early comments accumulate it. Same words, ten times the reach, purely from timing.
- Prefer mid-sized accounts. Under a megapost your comment is one of four hundred. Under a good post with thirty comments you are visible, and the author will actually see you.
- Go where the people you want work. If you want to be known by engineers at three specific companies, comment on what people at those three companies post. This is narrower and better than any topic.
Be short
A comment is not a post. It competes with the post it sits under, and a long comment reads as an attempt to take the thread over.
One sentence. Ten words is a good ceiling, and it is a harder discipline than it sounds. Ten words has room for exactly one idea, which is the correct number of ideas for a comment.
Two habits to drop. Do not open by restating the post, because everyone reading already read it. Do not end with a question aimed at the author, because it reads as a bid for a reply rather than a contribution.
Add something
The test is simple. If your comment could sit under any post on the platform, it is worth nothing. "Great post." "Well said." "This resonates." Those cost you nothing and return nothing.
Useful comments do one of four things:
- Add a fact. A number, a source, a case the post did not mention.
- Add the counterexample. Where the claim breaks. Do it plainly, not as a takedown.
- Add the specific. The post said "hire slowly". You say what happened when you did.
- Say the useful obvious. Sometimes the value is confirming from your own experience that the thing is true, with the detail that proves you were there.
Match the register of the post. A milestone announcement wants a short congratulation and nothing clever. A technical post wants a technical response. A joke wants you to be funny or quiet. Reading a meme as an invitation to comment on hiring strategy is the most common way to look like a bot.
Do it every day, at a human rate
Ten comments a day for a month beats a hundred in one afternoon. The compounding is in the repetition. People start recognising your name in the third week, and recognition is the whole product.
Rate matters for a second reason. Twenty-five comments posted in ten minutes is not a person, and the platform can tell. Spread them out.
What Iridium does with this
Iridium is a LinkedIn MCP server. Your agent, usually Claude, gets the commenting pipeline as a tool. The method above is the hard part. This is the part that runs it.
It finds the posts
You give it topics. It searches exact phrases and batches keywords together, over the past week only. Recency is not a setting, it is fixed, because a comment on anything older is a comment nobody sees.
It never shows you the same post twice
Every candidate is checked against everything you have ever drafted or commented on. Approved, edited, skipped, it does not matter. Once a post has been through your queue it never comes back. The dedup record stores what you did, not the other person's data.
It ranks what survives
Two layers of scoring narrow the list, then a relevance model decides whether a post is worth your name. Each surviving post is tagged with the register it is written in: milestone, meme, technical, question, or hottake. That tag is what stops a congratulation landing under a technical argument.
It writes ten words
Generation is capped at ten words. No dashes. No questions. The constraint does most of the work: at ten words there is no room for a preamble, no room to restate the post, and no room for the throat-clearing that makes AI text obvious.
You still have to edit. Approving AI comment text unedited is rejected by the server, with an error telling you to make it yours first. That is deliberate. A comment carries your name in a public feed and should have your hand in it.
It sends at a human rate
Approved comments are spread across an eight hour window with randomised gaps of at least eight minutes. The daily cap is a hard 25, enforced server side rather than suggested in a prompt.
The reasoning for 25 is written into the code rather than invented for a landing page: above roughly 20 to 35 quality comments a day, LinkedIn's bot detection risk rises sharply. Twenty-five sits inside that band with room to spare. If you breach a limit the send is deferred to the next morning rather than failed, so nothing is silently lost.
Commenting at named companies
This is the version most people should run, and it is the one nobody thinks of.
You name up to five companies. The pipeline finds what people at those companies are posting and drafts comments on that. The case it was built for is stated plainly in the tool itself: being visible to the engineers at the companies you care about.
Think about what that replaces. The normal way to get on the radar of a company you admire is to send cold messages to people who did not ask for them. The alternative is to be a familiar name in their feed for a month first. When you eventually message someone, you are not a stranger. You are the person who has been saying sensible things under their team's posts.
It works for hiring in the same direction. If you want engineers from three specific companies to consider joining yours, being a known name in their feed does more than a recruiter sequence and costs you nothing but attention.
Five is the ceiling. That is enough to cover the companies that genuinely matter and few enough that you stay recognisable in each one.
The filters, stated honestly
Three filters work, and all three do the same kind of thing. They remove posts from consideration.
| Filter | What it does | Who turns it on |
|---|---|---|
| Skip certification posts | Removes "I just earned a certificate" posts | Almost everyone. There is nothing to add. |
| Skip hiring posts | Removes "we are hiring" posts | Anyone who does not want to look like they are fishing for a job |
| Skip open to work posts | Removes "open to work" posts | Job seekers who do not want to be visible in that crowd |
These filters exclude and only exclude. There is no mode that comments only on hiring posts. Turning off the hiring filter means hiring posts are eligible alongside everything else. It does not make them a target.
One more thing worth saying plainly for job seekers. Skipping "open to work" posts controls what you engage with. It does not control what LinkedIn shows an employer about you. Nothing in any tool controls that.
What a week looks like
- Set your topics, or name up to five companies. Turn on the filters you want.
- Run the pipeline. You get a queue of ranked posts with a ten-word draft on each.
- Read them. Skip the ones where you have nothing to add, which will be several.
- Edit the rest. Ten words is fast to fix and the edit is required anyway.
- Approve. They go out across eight hours, at least eight minutes apart, up to 25.
- Repeat tomorrow. The result shows up in week three, not day two.
Steps two through five take about ten minutes. That is the actual claim.
Questions people ask next
Why only ten words?
Because that is the length people read. A comment sits under a post competing for the same attention, and anything longer reads as an attempt to take over the thread. Ten words holds one idea. The cap also removes the room where AI text usually gives itself away, in the preamble and the summary of the post you are replying to.
Can I comment only on hiring posts to find a job?
No. The three filters exclude categories and nothing inverts them into a target. If you want to be visible to people who are hiring, the working approach is the company pipeline: name the companies you want to work at and comment on what their people post. That puts you in front of the team rather than in the comment section of a job ad with three hundred others.
Is every comment checked for quality before I see it?
A relevance model decides which posts are worth commenting on and tags each with its register, and the ten-word cap plus the no-dashes and no-questions rules constrain the text. There is no separate quality score written on live comment drafts. The real check is you: the server refuses an approval of unedited AI text, so a comment cannot reach the feed without your hand in it.
Do the comments sound like me?
Not automatically. Comment generation is not voice matched against a stored profile. What you get is a short, on-topic, register-aware draft. Editing it is where your voice enters, and editing is mandatory before approval. At ten words that edit takes seconds.
What happens if I hit the daily cap?
The send is deferred to the next morning rather than failing. Nothing is dropped and nothing is silently lost. The cap is 25 and it is enforced on the server, so it holds whether the request came from the web app or from your agent.
How do I see which comments went out?
The activity log is append-only by design and records what went out and when. Ask your agent what it commented on yesterday and it reads the log rather than guessing.
What it costs
Iridium is $20 a month for one LinkedIn account, with a 7-day free trial and no card required. A week is enough to find out whether ten minutes a day of commenting changes who knows your name.
Be a familiar name in the right feed
7-day free trial. No credit card. Cancel anytime.
Start free → Nothing sends without approval