Do not index
Why is my LinkedIn reach dropping when nothing has changed about my content?
That is the question I get on every discovery call right now, and the people asking it have already been told it is the algorithm. It is not the algorithm. The algorithm started reading your drafts and noticing patterns you did not realize you were writing.
The blunt answer is that LinkedIn has named specific phrasing structures it now treats as AI tells. According to Engadget's reporting on the new crackdown, the platform is targeting "posts and comments that have obvious signs of AI construction like 'it's not X, it's Y' phrasing." That construction, the one you have seen in roughly 40 percent of LinkedIn posts for the last two years, is the exact thing getting flagged. It is the most visible tell. There are about a dozen others sitting in your drafts right now.
This is for agency owners between $200k and $2M in revenue who run a content operation for 5 to 15 clients, and for ghostwriters charging $5k to $30k per month for premium personal-brand work. If you are a founder paying a writer in that range and your reach has dropped 25 to 50 percent in the last 60 days with no obvious change in topics or cadence, the issue is almost certainly buried in your phrasing patterns rather than your posting frequency.
This is not for hobby creators who post twice a month from inspiration. Skip this if you do not run a structured content process. If you are still in the phase of writing every post from scratch with no template or framework, this article will not change your model because you do not have a repeatable phrasing pattern to audit.
The AI Phrase Audit
Call it what I call the AI Phrase Audit. It is a 20-minute pass through your last 30 posts looking for the specific construction patterns LinkedIn has learned to flag. The audit is not about cutting AI out of your workflow. It is about cutting out the phrases that betray AI, regardless of whether AI wrote them, because the platform does not know who typed the words. It knows what shapes the words make.
Five patterns make up the bulk of the damage. The first is the contrast flip, the it is not X, it is Y construction LinkedIn called out by name. The second is the rule of three closer, where the post wraps with three parallel sentences of similar length that hit the same beat. The third is the rhetorical hook that opens with a question, answers it with a single counterintuitive word, and then explains. The fourth is the here is what most people miss mid-post pivot. The fifth is the closing that starts with the takeaway or the lesson or what this means and delivers a generic principle. None of these are bad writing on their own. The problem is that AI tools default to all five at once because they all scored well in pre-2026 engagement data.
The audit pass is mechanical. Pull your last 30 posts into a single document. Search for each of the five patterns. Highlight every instance. If a post has three or more, it is a likely candidate for suppression and should be your first rewrite. If a post has all five, it is almost certainly already getting throttled and you can probably see the impressions to confirm. The reason this works is that LinkedIn is not catching one phrase. It is catching density. The more flagged patterns stack inside a single post, the higher the confidence score the system applies.
What replaces the patterns without breaking the writing
The replacement is not a different template. The replacement is the personality layer that templates cannot fake. A specific dollar figure from a client engagement. A name of a specific tool you stopped using and why. A sentence in the language a customer actually used during a sales call. A number that surprised you in your own pipeline. None of that lives in a prompt. It lives in your week. The writers and operators winning right now are the ones whose intake process captures that raw material weekly and feeds it into drafts.
This is also why humanizing tools have stopped working. They run the same five patterns through a slightly different shuffle, which is what the detection model was trained to catch. The fix is not different phrasing. The fix is different source material. If you do not know what you actually believe about a topic in concrete terms, no rewriter is going to manufacture that.
For agencies, the audit becomes a weekly hygiene step before client posts go live. For founders, it becomes the first thing you check when reach starts to slip. The metric that matters here is not impressions or comments. It is the gap between your stated reach and the reach the analytics dashboard is actually showing, which is exactly the territory I wrote about in how to measure LinkedIn success when the dashboard is not telling you the whole story. If you do not know which posts are running at full reach and which are getting held back to your direct connections, you are flying without instruments while LinkedIn quietly recalibrates the runway.
The strategic implication is that the next 12 months will reward operators who treat their own writing as the asset most worth auditing. The platforms will keep adding pattern recognition. The detection will keep getting sharper. The agencies and founders who build a recurring audit habit will hold their distribution. Everyone else will assume the algorithm is against them, hire another consultant, try another tool, and watch their reach erode in five-percent slices until they are no longer in the conversation.
