Strategy · 2026-04-25 · 14 min

Stop Using AI Blindly on LinkedIn Comments: It Is Quietly Burning Your Brand

I am not against AI. I use it every day. I am against people using AI blindly, pasting posts into a tool and shipping whatever comes out. This Dispatch covers the seven tells buyers spot in two seconds, the real brand cost, and the manual comment system where AI is used the right way.

First, A Clear Line

I am not against AI. I use AI every day. I use it to summarise long posts, pull themes from a comment thread, clean up my tone, and check a draft for repetition. AI is part of how I work.

What I am against is using AI blindly.

That means pasting a LinkedIn post into a tool, copying whatever comes out, and shipping it as a comment under a senior buyer's post. No thinking. No editing. No judgment. That is not engagement. That is noise with your name on it.

This Dispatch is about that second behaviour. Not AI itself. The blind use of AI.

If you sell anything serious on LinkedIn, founder, CEO, VP of sales, consultant, partner at a firm, read this before your next prospect opens your activity feed.

Why I Wrote This

I started writing this after a strategy call last month.

The prospect was a founder of a mid-sized B2B company. Strong team. Real revenue. Forty minutes in, he said:

"I used a tool that wrote my comments for three months. I went from zero to forty comments a day. Nothing happened. Worse than nothing. Two clients told me my LinkedIn looked weird."

Two clients. Already paying him. Told him his LinkedIn looked weird.

The tool company shows you a chart of impressions. They do not show you the chart of every existing relationship that quietly downgrades you every time you ship another machine-shaped reply.

I have heard the same story from a SaaS CEO, a partner at a consulting firm, a head of sales at a manufacturing company, and a fractional CMO. Different industries. Same pattern. They handed a brand they spent years building to a Chrome extension for sixty-nine dollars a month.

The Chrome extension is not the villain. Letting it write under your name with your eyes closed is.

The Seven Tells of a Blind AI Comment

Pattern recognition is the whole game. Once you see the shape, you cannot unsee it. Here are the seven fingerprints I look for, in the order I scan them.

Tell 1: The Compliment Opener

Every blind AI comment opens with validation. "Great point, Sarah!" "Powerful insight, James." "This resonates so much." The model is trained to be agreeable. Left alone, it cannot help itself.

A real peer does not open with applause. A real peer opens with a thought, a counter, or a specific reference to one phrase in the post.

ai_shape: "Great post, Marco, really insightful breakdown of the SaaS metrics nobody talks about!"

human_shape: "The CAC payback math you ran assumes flat churn. In my cohort it bends at month 14. Happy to share the curve."

Tell 2: The Restatement

The tool was given the post as context. It paraphrases the post back at the author. It is the conversational version of a parrot. "You are absolutely right that scaling outbound requires a strong foundation in ICP work." The author wrote the post. They know what they wrote.

Restatement is what a model does when nobody told it to add anything. Blind use ships that restatement as is.

Tell 3: The Affirmation Ladder

Three or four adjectives stacked. "So insightful, powerful, and well said." "Brilliant, thoughtful, and timely." No human writes like this in a comment. Humans pick one word or none at all. The ladder is a giveaway because the model is hedging. Nobody edited it down.

Tell 4: The Phantom Experience

This is the most damaging one. "I have seen this exact pattern in my own work." No specifics. No company. No metric. No timeline. The model is filling in credibility on your behalf.

Buyers read this and conclude one of two things. Either you are inventing experience, or you are letting a machine invent it for you. Both conclusions remove you from the shortlist.

Tell 5: The Em-Dash Cadence

Most LLMs love the em-dash. Real humans use one per paragraph at most. When a comment uses em-dashes to punctuate every clause, the odds it was generated and shipped without an edit are very high.

This is not a stylistic preference. It is a fingerprint. Read your own recent comments. Count the em-dashes. If it looks like an AI, the buyer will think it is one.

Tell 6: The Hollow Question

"What are your thoughts?" "Curious to hear how others approach this." "Would love to know what you think." The closing question is the model's attempt to drive engagement. It costs the reader nothing to ignore.

A real question costs the asker something. It exposes a gap in their thinking. It commits to a position. Hollow questions commit to nothing.

Tell 7: The Activity Feed Audit

This is the one that ends careers. Open the commenter's profile. Click Activity. Scroll through their last fifty comments.

If the shape is identical across unrelated posts, same opener, same length, same em-dash density, same closing question, they used a tool and shipped the output without thinking. No human writes the same way under a post about a Series B round as they do under a post about toxic culture in remote teams.

Decision-makers run this audit before booking calls. I run it before responding to inbound. It takes ninety seconds.

What This Looks Like to the Buyer

Operators who use AI blindly tell themselves the same story. "I am buying time. I would write these manually if I had the hours. The tool is just a force multiplier."

The buyer reads a different story.

Three compounding effects most operators miss.

Credibility dilution. Every blind AI comment dilutes the credibility of every manual one. A buyer who has seen four templated comments from you will not believe the fifth one is real, even if you wrote it yourself at three in the morning.

Network downgrade. Peers stop tagging you. They stop forwarding your posts. They stop recommending you in DMs. None of this is conscious. It is the slow, silent reclassification of your account from peer to noise generator.

Algorithmic clustering. LinkedIn's distribution groups accounts by behavioural signature. Accounts that ship machine-shaped engagement get clustered with other accounts that ship machine-shaped engagement. Your reach drops.

None of this shows up in any dashboard. The damage is invisible until your inbound goes quiet and you cannot figure out why.

How To Use AI The Right Way On Comments

This is the part most posts on this topic skip, because it does not fit a clean "AI bad" headline. AI is genuinely useful inside a comment workflow. You just cannot let it be the last hand on the keyboard.

Here is how I use it on my own profile.

Allowed. Asking AI to summarise a long post in two lines so I can read it faster. Asking it to pull the three claims in a post and tell me which one is weakest. Asking it to check my draft for repetition or weak verbs. Translating a draft into another language when I am writing to a non-English buyer. Sanity-checking tone before I ship.

Not allowed. Letting AI pick which posts I engage with. Letting AI write the opener. Letting AI invent personal experience I do not have. Letting AI ship the comment without me reading every word out loud.

The rule is simple. AI is allowed to help me think. AI is not allowed to think for me. Every comment that goes out under my name must be a comment I would defend in a room with the author sitting opposite me.

That one rule kills ninety percent of the brand damage tools cause, without throwing out the tools.

The Math Of A Manual Comment

The objection I hear most is "I do not have time to comment manually."

Here is the math.

A high-signal manual comment takes about ninety seconds. Read the post. Identify one specific claim. Write three or four sentences that add a frame, a counter, or a piece of data. Ship.

Eight comments per day at ninety seconds is twelve minutes. Five days a week is sixty minutes. One hour. Less than a single cold call block. Less than one internal meeting most executives sit through without speaking.

That hour produces forty high-signal touches per week with the exact decision-makers you want to reach. Each one is a deposit into a relationship account that compounds for years.

Compare that to the blind alternative. Two hundred generic comments per week, zero relationship deposits, negative brand equity, and a sixty-nine dollar monthly invoice.

The Manual Comment Protocol

This is the protocol I run on my own profile. Three rules. Each one designed to make the comment impossible to confuse with a generated one, even if AI helped me think it through.

Rule 1: Add, Do Not Agree

Every comment must contribute something the post did not contain. A counter-data point. A different frame. A specific example from your own work. An adjacent question that complicates the thesis.

If the only thing you can say is "agree," do not comment. Like the post and move on. A like costs nothing and signals nothing dishonest. A redundant comment costs your brand.

weak: "Totally agree, cold email is dead in 2026."

strong: "Cold email is not dead. The deliverability ceiling moved. I send 40 per day per inbox, hand-written, and book at 3.2x last year's rate. The volume model is dead. The medium is not."

Rule 2: Reference, Do Not Restate

If you must reference the post, quote one specific phrase, five words maximum. Then build off it. Restating the whole thesis signals you read the headline and skipped the body.

weak: "Great point that ICP drift kills outbound conversion over time as teams grow."

strong: "On 'ICP drift kills conversion'. The failure mode I see most is account scoring that gets frozen at year one and never re-trained. Quarterly recalibration is the unsexy fix."

Rule 3: One Thought, One Sentence, One Costly Question

The best comments are short. One sharp thought. One sentence of evidence. Optionally one question that costs the author something to answer.

weak: "What are your thoughts on remote AEs?"

strong: "Would you hire a remote AE today knowing they would never meet a customer in person? Not asking rhetorically. Wrestling with this for my next two roles."

The costly question creates obligation. The author has to think before they reply. You are buying a moment of their cognition. That moment is the relationship.

Recovery Protocol

If you read the seven tells above and recognised your own comments in those examples, this section is for you.

Step 1: Audit

Go to your profile. Click Activity. Filter to Comments. Scroll back ninety days. Read every comment as if you were a prospect seeing it for the first time. Be honest about which ones look generated. Most founders find that between forty and sixty percent of their recent comments fail the test.

Step 2: Delete The Worst Offenders

LinkedIn lets you delete your own comments. Do it. The argument against deletion is that it leaves a hole. The hole is invisible. The comment is not. Delete every comment that hits three or more of the seven tells.

Step 3: Turn Off Auto-Send

Not the tool. The auto-send behaviour. If your tool has an option to draft and let you edit before posting, switch to that. If it does not, cancel it and pick one that does. The point is to put a human hand back between the model and the publish button.

Step 4: Thirty Days Of Human-Final

For the next thirty days, every comment, every DM, every post is written or finalised by you in the browser. Use AI to help you think. Do not use AI to ship the final words.

This rebuilds the muscle and re-trains the LinkedIn algorithm, which has been clustering your account with the noise farmers, that you are a different kind of operator.

Step 5: Re-Emerge Slowly

After thirty days, ramp engagement back up. Eight thoughtful comments a day. Two posts a week. Real DMs to real people. The account will recover. The relationships will recover. The pipeline will follow.

I have walked operators through this exact protocol. A SaaS founder. A sales VP at a logistics company. An independent management consultant. All three reported the same thing. Within sixty days, the inbound that had quietly died came back. Not because the algorithm forgave them. Because the humans on the other side of the screen started recognising them as humans again.

The Larger Pattern

Blind AI comments are not really about AI. They are about the belief that the parts of selling that look like work can be outsourced without cost.

They cannot.

The one thing that makes outbound work, at any volume, in any channel, in any industry, is the application of judgment in front of a specific human being. A buyer can tell within seconds whether you applied judgment or whether you delegated it to a machine.

Use AI. Use it heavily. Use it to think faster, read faster, draft rougher first passes, and translate across languages. Just do not use it as the last hand on the keyboard. The economics of trust do not survive that shortcut.

This is the entire thesis of the SENT manifesto. It is why I do not run sequencers, do not buy lead lists, and do not let any tool ship a single character of a message under a client's name without a human reading it first.

If your account has been restricted, throttled, or warning-flagged beyond the comment pattern, the 12-month recovery timeline Dispatch covers what comes next.

Stop shipping AI comments on autopilot. Start using AI the way a senior operator uses any tool. With judgment, on top, every time.

If you want the full manual outbound system behind this, targeting, sequencing, sender infrastructure, see how I build it.

, Arjen