Comments Are the Channel Now

Publishing on LinkedIn is a cold-start problem: you write, and the platform decides how many people find out. Commenting is not. Someone else already assembled the audience, already earned their attention, and you get to say something intelligent inside it.

The asymmetry is large enough that for accounts under about five thousand followers, comments routinely out-deliver posts on profile visits and inbound conversations. That is not a growth hack; it is just where the attention already is.

What a comment worth leaving looks like

Three properties, all of which are absent from the average reply:

1. It adds information the post does not contain. A number from your own work, a counterexample, a constraint the author did not mention.
2. It can be disagreed with. Agreement is invisible. A specific, generous disagreement is the single most reliable way to be read by the author and their audience.
3. It stands alone. Someone should get value from it without reading the post. That is what makes it travel.

Length is not the variable. A two-sentence comment with a real number beats four paragraphs of appreciation every time.

So what should the agent actually do?

Everything except write the comment. The work that makes a good reply possible is mostly retrieval, and retrieval is what agents are genuinely good at.

• ·Selection. Surface the ten posts today where you have first-hand knowledge and an existing relationship. This is the highest-value step and the one people do worst by hand.
• ·Context. For each: what you have said publicly on this topic, any prior interaction with the author, and whether they are in your pipeline.
• ·Evidence. The relevant number from your own data, so the reply can be specific in the thirty seconds you have.
• ·Nothing else. No drafted text. The moment there is a draft in the box, you edit it instead of thinking, and the result reads like everyone else’s.

We tested the drafted-comment version internally and killed it. Reply quality dropped measurably even with heavy editing, because a draft anchors you to its framing. The context-only version made commenting faster and better, which is the rare case where the weaker-sounding feature wins.

A cadence that survives a busy week

Ten comments a week, in one twenty-minute block, on posts from people you would take a call with. Track profile views and inbound conversations rather than comment likes. The likes are noise and the conversations are the point.

This pairs directly with a lower posting cadence: two posts and ten comments a week is roughly the same time as five posts, and produces substantially more of what you actually want. See stop posting daily.

The attribution problem

Comment-driven pipeline is nearly invisible in standard reporting: someone reads your reply, checks your profile, remembers you three weeks later, and arrives via direct traffic. Logging comment activity against contact records in your CRM is the only way to see the pattern. Content attribution covers the plumbing.

My own comment routine, thirty minutes, twice a day, is written out at alexcinovoj.com↗ (opens in a new tab). The tooling that makes it survivable at scale is built at TechTide AI↗ (opens in a new tab).