Do not index
Should I stop using AI to write my LinkedIn posts? That question is landing in my inbox every day this week, and the answer is no. You have to move it. AI does not belong in the generation step. It belongs in every step around the generation step. LinkedIn just made that call at the product level, which is the loudest signal the platform has ever sent about what it will and will not reward.
What actually happened is worth reading twice. LinkedIn removed its own enhance your post AI writer and replaced it with a proofreader that fixes grammar without changing your voice. It shipped on the same day the seems like AI slop report button passed one million flags, and two of the company's help pages for AI post writing were retitled no longer available. LinkedIn CPO Hari Srinivasan said content the company classifies as AI slop is now getting 40% fewer views than it did a few weeks earlier, according to IBTimes UK. Pangram found about 41% of long-form LinkedIn posts were fully AI-generated, the worst rate among the major platforms it studied, with Reddit at roughly 4%.
The platform that spent five years telling everyone to post more just deleted the button that made posting more effortless. That is not a moral position. It is a supply problem. When generation becomes free, the scarce thing stops being words and starts being the judgment behind them.
This matters most if you are a founder running your own personal-brand content while also running the company, or an agency owner between $200k and $2M in revenue whose retainers are priced on voice. Ghostwriters charging $5k to $30k per month are selling exactly one thing, and it is not typing speed. It is the extraction and defense of a specific human point of view, at volume, on someone else's behalf. If that is your business, the platform repriced your competition downward and your product upward on the same afternoon.
This is not for everyone. Skip it if your content exists to fill a calendar slot and nobody internally can tell you what the company actually believes. Skip it if you are running a 3 person shop that took on 12 clients at $800 a month and needs generated drafts to stay solvent, because the fix there is your pricing, not your prompt. And if you are still measuring the work by posts shipped per week rather than by which specific buyers replied, this article will not change your model. The 40% view penalty is a distribution problem, but the thing underneath it is a positioning problem, and distribution fixes never solve positioning problems.
The Judgment Line
Every content task falls on one side of a line. Below the line sits mechanics: transcription, formatting, research retrieval, headline variants, spelling, scheduling, cutting a long post into three shorter ones, checking whether you already made this argument back in March. Above the line sits judgment: what you believe, what you are willing to be wrong about in public, which client story you are allowed to tell, which take will cost you a deal and is worth publishing anyway. AI goes below the line without limit. It never crosses above it. That is what I call the Judgment Line, and it is the only rule you need for this.
Most operations run it backwards. They use a model to produce the opinion and a human to fix the commas, then wonder why 200 posts produced no inbound. Reverse it. Pull the point of view out of the founder in a 30 minute conversation, then use the model for everything mechanical between that conversation and the published post. The thinking stays human and expensive. The production gets cheap and fast. LinkedIn just endorsed that exact split with a product decision, because a proofreader is a below-the-line tool by definition and an enhance button is not.
What the AI slop penalty means for client work
If you sell content, you now have a conversation to run with every client on the roster. The honest version is short. Their reach may already be down, and it is not the algorithm being unfair, it is a classifier doing the job it was built for. What protects a retainer through that conversation is a documented review step that catches generic drafts before they ship, which is the same quality control system that prevents client churn before your retainer ends. Agencies that already had one are having an easy month. Agencies that did not are discovering that their process was never a process, it was a prompt.
There is a client education piece too. Founders will ask whether touching AI at all gets them flagged. It does not. Detection is looking at generated text patterns, not at whether a tool was involved in the file. A post you dictated, restructured with a model, and rewrote in your own words is not slop. A post where the model chose the argument is, and readers were catching it by hand a million times over before any classifier got involved.
The trajectory question is what your business looks like in 18 months if generation stays free and judgment stays scarce. Every operator selling volume is now competing with a tool that costs almost nothing and just got demoted by the platform it publishes to. Every operator selling extracted point of view is competing in a shrinking pool, because that skill does not compound with compute. LinkedIn deleting its own writer is not the story. The story is that the platform ran the numbers on generated content, watched a million people flag it manually, and concluded that the cheap version was destroying the thing that made the feed worth opening. Your clients will reach the same conclusion on their own timeline. The ones building around a human point of view now will not need to reprice when they do.
