Do not index
"Is my client's content going to get flagged?" That question has been landing in my inbox every day since LinkedIn shipped the report button. Here is the answer. Ghostwriting is not the risk. Generic is the risk, and it always was. LinkedIn is not trying to detect whether a machine typed the words. It is detecting whether the words carry information that only one specific person could have supplied. Clear that bar and a flag costs you almost nothing. Fail it and you were already invisible, and the button simply made the invisibility official.
The context is worth stating plainly. LinkedIn's product chief called the AI slop crackdown "a top priority for all of us," according to Forbes reporting from August 2026, and that same reporting puts over 40% of long-form LinkedIn posts as fully AI-generated. Any user can now flag a post as slop, and a flag quietly reduces distribution. No warning, no appeal queue you will ever see, just a post that lands softer than it should have.
What everyone will tell you in response to this is to write like a human. That advice is useless because it describes an outcome, not a mechanism. The useful question is what the classifier and the annoyed reader are both responding to, and the answer in both cases is the same. They are responding to whether the post contains anything that could only come from having done the work.
What LinkedIn's AI detection is actually reading
Structure is the loudest tell. A three-word opener on its own line, a rhetorical question, three parallel clauses of increasing length, a one-line reversal, and a closing question inviting comments. That shape is not evidence of a machine so much as evidence of a template, and templates were spreading through LinkedIn long before generative tools arrived. What has changed is that the shape can now be produced at effectively zero cost, so it has flooded every feed and become the visual signature of low-information content.
Cadence is the second tell. Uniform sentence length across an entire post, no fragments, no asides, every paragraph the same weight. Real people write unevenly because they care unevenly. They rush past setup and slow down on the part they actually have an opinion about.
But the deepest tell is informational. A generated post can produce the claim that discovery calls should qualify budget early. It cannot produce the observation that on calls with agencies between $200k and $2M in revenue, the budget question lands cleanly at minute nine and defensively at minute two, because that requires having sat through several hundred of them. Specificity of that kind is not a style. It is residue from experience, and it is the one thing that cannot be synthesized from a prompt.
The Specificity Floor
The standard I hold every piece of client work to is what I call the Specificity Floor. A post clears the floor when at least one sentence in it could not have been written by someone who had not personally done the thing. Not a sharper phrasing of common knowledge. A fact, a number, a failure, a client reaction, a decision and its cost. One is enough. Two is comfortable. Zero means the post gets rewritten no matter how polished it reads.
Applied to a $10k per month retainer where you are shipping twelve to sixteen posts, the floor changes the shape of the engagement immediately. You stop being able to write from a content calendar and a vague brief, because a calendar cannot produce residue. You have to actually extract it, which in practice means a recorded conversation every two weeks where you push the founder past their prepared positions until they say something they would not have said on a podcast. That call is the deliverable. The writing is downstream of it.
This is not for everyone. Skip this if you sell content by volume at $1,500 per month and your margin depends on never getting the client on the phone. The Specificity Floor is incompatible with that model, and pretending otherwise wastes your time. If your differentiator is turnaround speed rather than judgment, this article will not change your business. It will just describe the wall you are walking toward. The agencies that build extraction into their process are the ones with a defensible service, which is the same argument underneath the quality control system that prevents client churn before a retainer runs out.
Here is where this leads. The report button is not really a moderation feature. It is a repricing event. For three years the market treated writing as the scarce input and paid for words per month, which is why a $5k retainer and a $30k retainer could produce output that looked broadly similar in the feed. The flag makes generic output actively expensive rather than merely ineffective, and once output has a downside, the scarce input stops being writing and becomes access to the thinking worth writing down. If your service is built around producing text, the floor under your pricing is about to move. If it is built around getting something out of a founder that nobody else can get, this is the best thing that has happened to your business in a while.
