For a decade the advice never changed: post more. More Reels, more carousels, more comments, more of everything. AI made that advice effortless to follow, and a working social media strategy in 2026 is now close to the opposite of it.
On 30 July, LinkedIn added a button to every post in the feed reading “Seems like AI slop”. Within three weeks, people had pressed it more than a million times.
That is not a complaints feature. It is a training set, and your audience is building it.
Platforms have always filtered content. What changed on 30 July is who holds the switch.
The report button sits in the same menu as every other post action, and it takes one tap. LinkedIn’s chief product officer Hari Srinivasan told TechCrunch the company is “introducing new classifiers to identify if a post is AI slop or other low-quality content”. Every press teaches those classifiers what your industry considers slop.
Two details matter more than the button itself. LinkedIn now blocks hundreds of thousands of automated comment attempts a day, so the comment-pod tactic is already dying. And it has started showing private flags in your own analytics when your audience is reading your posts as machine-written. You will be told, quietly, before your reach tells you.
Srinivasan was careful about how the signal is used: “No single piece of feedback is used to adjust how widely a post travels.” One annoyed reader cannot bury you. A pattern can.
Here is the number that should reorganise your week. Accounts that copy and paste AI-written content into LinkedIn are seeing around 40 percent fewer views than they did before the button launched.
Not banned. Not labelled. Just quietly seen by fewer people.
Read that alongside the other thing LinkedIn did: it killed its own “enhance your post” AI writer and replaced it with a proofreader. The platform that was selling you AI drafting a year ago has decided drafting was the wrong job for the machine, and tidying is the right one. When a platform removes its own feature, it is telling you what its data says.
The practical reading is not “stop using AI”. It is that the judgment has to be yours and the polish can be borrowed. Draft it yourself, badly and fast, then let the tool tighten it. That order survives the classifier. The reverse order is what the button was built to catch.
The counterpart to all this policing is quieter, and most people have missed it.
Instagram shipped Replace Audio on 21 July. You can now change the music on a published feed post, video or carousel, as many times as you like, and your likes, comments, shares and reach carry over untouched. The old fix for a bad audio choice was delete and repost, which meant throwing away every bit of engagement the post had earned.
One caveat worth knowing before you get excited: replacing the track does not put the post back into distribution. It is a preservation tool, not a promotion tool. Nobody new sees it because you swapped the song.
What it does give you is a reason to treat published posts as assets you can maintain rather than things you shipped and forgot. A carousel that performed well with a track that has since become tired can be refreshed for the people who find it through search and your profile grid for months afterwards.
Instagram has also started extending creator bonuses to photo and carousel posts, not just Reels. That is a small change with a large signal inside it.
Reels reward production speed, which is exactly what generative tools are best at. Carousels reward sequencing, argument and a point of view, which is exactly what they are worst at. Paying for the second category is a bet on the thing that is harder to mass-produce.
TikTok’s own research points the same way. Its marketers rated AI tools highly for production speed and noticeably lower for original copywriting. The industry has quietly worked out that the bottleneck was never making more; it was having something worth making.
No. The penalty pattern is around copy-pasted AI output, not AI assistance. LinkedIn’s own replacement for its writing feature is a proofreader, which is the company signalling that editing with AI is fine and generating with it is not.
No. Srinivasan was explicit that no single piece of feedback changes how far a post travels, and that distribution is set by a range of signals. The risk is a consistent pattern across many posts, not one reader having a bad morning.
No. Engagement and reach are preserved, but the post is not re-entered into distribution. Treat it as maintenance on something that still gets found, not as a way to relaunch it.
For written copy, current industry guidance does not require a label, though the rules differ by format and jurisdiction. We covered what actually needs disclosure and what does not in a separate piece.
Every change here points one way. The platforms spent three years watching what unlimited content production did to their feeds, and they are now building the machinery to slow it down, with your audience holding one of the levers.
That is good news if you were never going to win on volume anyway. The accounts that suffer are the ones whose entire advantage was speed. The ones that gain are the ones with a series worth following and something to say inside it.
Post less. Finish what you post. That is the whole social media strategy for the rest of this year.
Not sure which half of your content is carrying the account? Book a free consultation with Social Lady and we will audit your posting mix and rebuild it around what the platforms are actually rewarding.
