The experiment nobody planned
Every working day, as part of building V8’s presence on LinkedIn, I commented on posts by people who write about AI. Axia picked the posts and drafted each comment; I read every one, approved it and posted it myself. Thoughtful comments, not “great post”. Each one was logged, watched for replies for seven days, and tracked back to whoever replied.
Ninety days later I had a clean data set, so I asked the question I should have asked in week one: did any of it turn into a meeting?
The numbers
| Comments posted on other people’s posts | 482 |
| Different authors I commented on | ~180 |
| Authors where I commented exactly once | 133 |
| Authors who produced a conversation of some kind (a reply, a thread, a signal) | 68 |
| People I actually sat down with in the same period | ~72 |
| Of those 72, how many I had commented on before the meeting | 3 or 4 |
The ten most-commented accounts took 173 of my comments, about 40%. Between them they produced a lot of conversation: 60 reply threads. They produced no meeting.
Where the meetings did come from
Almost all of them came through channels with a human behind them:
- BNI chapters and the people I met there
- introductions, including through Gina
- email to people we already had context on
- direct LinkedIn outreach to a specific, qualified person
- events
None of those is clever. All of them involve a person who knows another person.
What the engagement actually was
The AI-content accounts I followed did well. Their posts get dozens or hundreds of comments. But read the comment section of a post about “AI adoption for small business” and ask who is in it:
- people who sell AI adoption to small businesses
- freelancers and agencies
- other people building content audiences
- very few business owners in the town where I work
The post is a stage. The audience is mostly other performers. Everyone is engaged, the numbers look healthy, and nobody in the room is buying.
What I am not saying
- Not “social media does not work”. Direct LinkedIn outreach to a named, qualified person produced about a fifth of my meetings in this window. The channel works. The audience I was borrowing from other people’s posts did not.
- Not “those creators are doing it wrong”. They are building an audience, and they have one. It is just not mine.
- Not a proven zero. My attribution is imperfect. About 40% of the meetings carry no recorded source, and I matched people by name, which loses some. A comment can raise your visibility without being the thing that books the call. The honest claim is that I found no evidence it converts, not that it cannot.
A few real exceptions
Three or four people I commented on did later sit down with me. Looking at them, none was a creator. They were builders and operators who happened to post something relevant, and the conversation moved off the feed quickly: a DM, then a call. The comment was an introduction, not a funnel.
What I changed
- Comments move to people who might buy. Most of the daily comment budget now goes on posts by prospects and existing contacts, where a comment is part of a relationship that already exists.
- A small number of peers who reply keep a place. Three or four people reply thoughtfully and consistently. That is a conversation, not an audience.
- Mining influencer audiences for prospects is cut to one account a day. In August the same exercise returned zero qualified prospects seven times running, while a plain targeted search returned five to nine a run. I kept one a day so that if the pattern ever changes I will see it.
- Record the source at the moment of booking. I could only answer this question after the fact, and imperfectly, because the calendar does not know where a meeting came from. Fixing that is the real lesson.
The bit that is bigger than my numbers
If engagement in a niche is dominated by people who sell into it, then engagement is measuring who is talking, not who is buying. That is true of AI in 2026. I suspect it is true of most “how to use X” content in most niches.
My meetings came from people who know people. So the first hour of my day now goes to the people I might actually meet, and to asking the people I already know who they would like to meet.
Method note, in case anyone asks: 482 comments is the count of presence-monitor items created in my action queue between 14 July and 12 October 2026 (415 with a parseable author). Meetings are post-meeting items logged in the same window (about 72 distinct people). “Commented first” means a monitor on that person existed before their meeting item was created.