Auto commenting on LinkedIn: what works and what gets flagged

Auto commenting works only when each comment says something specific about the post and the volume stays small. Generic praise, copied lines and bursts of activity are what readers ignore and what LinkedIn's rules against automated activity target. In Sellinger the agent can write comments on posts about topics you list, capped by a daily limit you set (default 3, at most 10), with 60 seconds between actions by default, and it pauses for 24 hours after LinkedIn refuses the feature three times in a row. No tool can promise LinkedIn will never restrict an account.

Commenting is the cheapest way to be seen by a buyer's network, and the most abused. Everybody has seen the comment "Great insights, thanks for sharing!" from a stranger. This post separates the use that earns attention from the use that earns a restriction, using what the product actually does and what LinkedIn actually says.

What is auto commenting?

A tool reads posts, usually matched on keywords or from a list of people, and writes a comment under them without a person typing each one. Three versions exist, and they differ in risk and purpose:

1. Template comments. The same few phrases, rotated. Cheap and obvious.
2. AI written comments. A model reads the post and your directions and writes a comment for that post. This is what Sellinger does.
3. Engagement pods and exchange groups. People or accounts agreeing to boost each other. This is a different thing, and Sellinger does not do it.

Sellinger has two commenting modes on the LinkedIn agent, both switched off until you turn them on:

• Topic comments. You list topics. About every 30 minutes the agent scans the feed for posts that match, records each match as an intent signal, and may comment on posts where comments are enabled, within a daily limit. It can also create leads from the authors of matching posts.
• Pre invite engagement. Shortly before a connection request, the agent likes up to five of the person's recent posts (default two) and, if you choose, comments on one. The post on pre invite engagement covers the detail.

What does LinkedIn say about it?

LinkedIn's help pages state that it prohibits third party software that automates activity on the site, and that it may restrict accounts it suspects of using it (prohibited software and extensions, account restrictions). Automated comments are covered by that. LinkedIn does not publish the thresholds that trigger a restriction, so any article that gives you a safe number of comments a day is guessing, and we will not add to it.

What we can say is what the product does to keep activity modest: a low default of three topic comments a day, a hard maximum of ten, a delay between actions that you can set from 10 to 600 seconds (60 by default), and a stop. When LinkedIn refuses the feature three times in a row, engagement on that account pauses for 24 hours. These are limits, not a guarantee. The safety checklist and is LinkedIn automation safe explain the residual risk and the decision, which is yours.

What makes a comment worth reading?

A reader judges a comment in two seconds. These traits make it past that test:

• It refers to the post. A detail, a number, a phrase from the author. If the comment could sit under a different post, it fails.
• It adds something. A fact, a counterexample, a short story, a useful question.
• It is short. One to three sentences.
• It sounds like you. The directions you give the agent should carry your tone and your point of view.
• It has no link and no pitch. A comment that sells is an advertisement in someone else's room.

And these get ignored or reported:

• "Great post!" and every variant of it.
• The same sentence under ten posts.
• A comment that repeats the post's headline.
• A reply that misreads the post, which is worse than silence.
• A tag of people who were not part of the discussion.

What are the failure modes of AI comments?

A model that writes comments will, left alone, drift into a few habits. Watch for them in the first batch.

• Flattery. "This is such an important point." Remove it in the directions.
• Summary. Restating the post back to its author.
• Overconfidence. A claim about a field the model does not know. Tell it to ask a question when it is unsure.
• Bad targets. Posts about a death, a layoff, a political fight or a personal struggle. Exclude those topics in the rules, because a cheerful comment under a sad post lands badly.
• Repetition. The same opening across a week. Read ten comments side by side and you will see it.

How should you set it up?

1. Start with three topics, narrow ones that match your buyers' real conversations. The post on intent signals worth acting on helps you choose.
2. Write the directions like a brief to a colleague. Your view, your tone, what to avoid, when to stay silent.
3. Leave the daily limit at the default of three for the first two weeks.
4. Read every comment for the first week. Edit the directions, not the comments, so the fix lasts.
5. Check what came back: profile views, follow requests, replies to your comments, and connection requests that mention one. Count them; do not assume.
6. Raise the limit one step at a time, and stop if replies fall or if LinkedIn refuses the feature.

Does it produce leads?

The data we have is about the signal, not about the comment. In our own LinkedIn outbound, the largest share of prospects came from an intent signal, people who had liked or commented on a post about sales automation, and those lists replied at 54.6% against 37.2% for untagged lists in the October 2025 cohort (see the Sellinger case study). That is evidence that a person engaging with a topic is a better lead than a person on a list. It is not evidence that commenting on their posts makes them reply. The case study also notes that only 5 of 30,000 openers mentioned the post that put the person on the list, so referring to the post is an untested idea, and we say so.

For your own account, the simplest test is a comparison. Take two similar lists. Comment before the invite for one, and not for the other, and compare acceptance and reply rates. The calculator does the arithmetic.

Who should not use it?

Skip automated comments if your account is your only route to customers and you cannot afford a restriction, if the account belongs to a person who will not read what is written in their name, or if your market is small enough that everyone will notice a pattern. In those cases, a few minutes of real commenting a day beats any tool. Where the account owner agrees, the agent is a way to keep up a presence at a modest pace, with directions that sound like them. It is not a way to be everywhere at once.

A checklist

• [ ] Commenting is on only for accounts whose owner agreed to it.
• [ ] The topics are narrow and the directions carry my tone.
• [ ] The daily limit started at the default and rose in steps.
• [ ] I read the first batch and edited the directions.
• [ ] Sensitive topics are excluded.
• [ ] I count results, and I claim none I have not measured.
• [ ] I accept that no tool removes the risk of a restriction.