Do not index
How do I get taken seriously on LinkedIn when everyone assumes a machine wrote my posts? That question shows up on almost every call I take with agency owners now, usually about ten minutes in, right after they admit their reply rate fell off a cliff this year. Here is the answer. LinkedIn becoming the internet's punchline for AI slop is the best thing that has happened to the people still writing their own posts. When the baseline is generic machine output, a specific point of view stops competing against a million other posts and starts competing against noise. That is a far easier fight to win.
Straight Arrow News covered the shift in late July 2026, framing authenticity as Silicon Valley's new operating buzzword and describing platforms moving moderation away from what gets said and toward who or what is saying it. Substack CEO Chris Best, quoted in that piece, put it plainly: "We're sick of slop and we don't want substack to turn into LinkedIn." The same article notes that Pangram research crowned LinkedIn "the most AI-saturated platform" on the web. If you write on LinkedIn for a living, that is your operating environment. Your work now gets graded against a default assumption of fake.
Most people read that and panic. The correct read is the opposite. A reputation problem at the platform level is a distribution problem for everyone posting generic content and a distribution advantage for anyone posting something a model could not have produced. Reputation collapse does not lower the ceiling. It lowers the floor, which raises the gap between the floor and anything with a human fingerprint on it.
This applies most directly to agency owners running between $200k and $2M in revenue, ghostwriters charging $5k to $30k per month for founder content, and founders writing their own posts to build pipeline rather than applause. You sell judgment. Your buyers are now primed to assume judgment got outsourced to a model somewhere between the brief and the publish button. That assumption is the actual threat to your business, not the volume of AI content in the feed.
Skip this if you are running a volume play. If your model depends on shipping 200 posts a month across 30 accounts at $300 a client, none of this changes your economics and you should keep optimizing for cost per post. This is also not for people who believe the fix is a better prompt. If you are still looking for the model that finally sounds human, this article will not change how you operate. The premise here is that the sound was never the problem. The absence of a position was.
The Slop Contrast Test
Here is what I would actually do. Before anything publishes, run what I call the Slop Contrast Test on every paragraph. The question is simple. Could a competent model, handed the same brief and the same public information, have produced this exact paragraph? If the answer is yes, the paragraph is doing nothing for you. It is not neutral. It is actively confirming the reader's suspicion.
Passing the test requires one of three things in the paragraph. A specific only the writer could know, meaning a number from inside the business, a client situation, a decision that was made on a particular Tuesday for a particular reason. A position that could plausibly cost the writer money, meaning a take that some segment of the audience will disagree with hard enough to unfollow. Or a piece of reasoning that reverses the obvious conclusion, meaning the writer looked at the same input as everyone else and came out somewhere different. Paragraphs with none of those get cut, not rewritten.
The reason this works is mechanical. A model generates the statistical center of everything written about a topic. That center is exactly where the slop lives, because the slop is also the statistical center. Anything specific, contrarian, or costly sits away from that center by definition. You are not trying to sound human. You are trying to say things the center cannot generate.
In practice this cuts a lot. On a typical 900 word founder post, the test removes 40 to 60 percent of the first draft, and what survives is shorter and considerably more pointed. That is the trade. Clients who came up in the era of posting five times a week find this uncomfortable, because the test makes it obvious that three of those five posts were filler. The ones who adjust end up posting less and hearing from more qualified buyers, which is the only metric that pays retainers.
What Changes When the Feed Assumes You Are a Bot
The second order effect is that proof of humanity becomes a positioning asset rather than a hygiene factor. For years the differentiator on LinkedIn was consistency. Show up, post on schedule, compound. Consistency is now table stakes and, worse, it is trivially faked, so it no longer signals anything about the person behind the account. What signals is specificity that would be expensive or embarrassing to fabricate.
That reorders how you build a presence. Instead of a content calendar built around topics, you build it around the handful of things the founder actually knows that nobody else can claim. This is the same logic behind treating LinkedIn strategy as a positioning exercise rather than a publishing exercise, which is worth reading if you are rebuilding a client program from scratch: how to build a LinkedIn content strategy that holds up.
For agency owners the operational implication is that voice extraction stops being a nice differentiator and becomes the entire product. If your process is a questionnaire and a monthly check in, a model can approximate your output and your client will eventually notice. If your process pulls things out of a founder that they would not have written down on their own, no model competes with that, because the raw material never existed in the training data.
The platform's reputation is not going to recover on its own timeline, and waiting for it to is a strategy that hands the next two years to whoever is willing to say something specific in the meantime. The accounts that come out of this period with real pipeline will not be the ones that posted the most. They will be the ones whose readers could tell, paragraph by paragraph, that a person with something to lose wrote it. That is a durable position, and it gets more durable every time another competitor decides the answer is more volume.
