LinkedIn built a button to report your founder's thought leadership

On 30 July, LinkedIn added an option to the three-dot menu on every post: "Seems like AI slop." Tap it and the post vanishes from your feed, and LinkedIn treats the signal roughly the way it treats "not interested" - the post's reach narrows from there (TechCrunch, Fortune).

The agencies exposed by this are not the ones using AI. They are the ones who bought a founder voice on a two-posts-a-week retainer and have never read a month of it in one sitting.

The button is not a classifier

Most of the commentary went to detection. Jennifer Stack made the fair version of the case in The Drum last week: automated slop-spotting punishes careful writers, because a model cannot tell the difference between a founder who drafted something and tidied it up, and a bot that paraphrased a competitor's post. That criticism holds against LinkedIn's May update, which leaned on in-house detection - Laura Lorenzetti's editorial team claimed 94% accuracy at tagging generic content, and tagged posts stay inside the author's own network rather than travelling.

The button is a different instrument. It is not a model guessing. It is your first-degree network - the client you pitched in March, the ex-colleague who ran your planning team, the journalist you want to know better - deciding whether the post sounds like you.

That is a harder test to pass and a worse one to fail.

A detector can only find patterns. The people who know you can hear a voice that isn't yours.

What agencies actually bought

Around 41% of long-form posts on LinkedIn are now estimated to be machine-written. Some of that is individuals saving twenty minutes. A meaningful share of it is our industry: founders who signed a content retainer, approved a monthly grid, and have been publishing under their own name ever since.

The failure pattern is consistent. The posts are competent. They are on-topic. They contain nothing first-hand - no client meeting, no number the founder would only know from inside the business, no position they would have to defend at an industry dinner. They read like someone summarising a category rather than working in it.

That was survivable while the feed rewarded volume. It stops being survivable when the reader has a button.

LinkedIn told on itself

The detail worth noticing is not the button. It is that LinkedIn removed its own "enhance your post" writing tool from the compose box in the same week, and replaced it with a proofreader built to correct writing without changing the voice.

LinkedIn spent two years encouraging users to generate posts, measured what generation did to the feed, and took the feature out. That is a platform conceding that the thing it sold as productivity was the thing degrading the product.

Every agency running a founder-content retainer is now running the experiment LinkedIn has just ended.

What it costs, and when

Nothing dramatic happens on Monday. Reach softens. A post that used to reach second-degree connections stays home. Nobody tells you, because nobody can see it happening from the inside.

The expensive version arrives months later, in a pitch. A client team looks your founder up before the chemistry meeting, reads a quarter of posts in one sitting, and forms a private view about whether there is a mind in there. That view never appears on a scoring sheet. It sets the temperature of the room before anyone has spoken.

We wrote in May about how often agency founders should publish. The cadence argument has largely settled itself this week. Two posts you would defend out loud beat five you have never read.

The honest test takes an hour: open your founder's profile, read the last three months end to end, and ask whether a stranger would finish it knowing what your agency believes. If the answer is no, the problem started at positioning rather than publishing, which is where our method tends to start too.