AI slop is becoming a distribution risk
For years, businesses could treat generic social content as harmless filler.
A post might say very little, but it kept the company visible. A recycled opinion still filled a slot in the calendar. If the reach was disappointing, the cost appeared limited to the time spent making it.
LinkedIn’s new “Seems like AI slop” feedback option changes that calculation.
Members can now open the three-dot menu on a post or comment and flag material they consider generic, repetitive or lacking substance. LinkedIn says more than one million members used the option in its first two weeks. It also reports that people are seeing 40% fewer views of content the platform classifies as AI slop than they were a few weeks earlier.
Those numbers should not be turned into a simple cause-and-effect claim. LinkedIn uses several signals, not a single report, and says safeguards are intended to prevent individuals from unfairly targeting an author.
The important development is broader: low-substance content is no longer merely easy to ignore. It can become a distribution and reputation risk.
AI use is not the test
LinkedIn’s definition does not treat every use of AI as a problem.
It describes slop as content that may appear polished but lacks a clear point of view, distinctive experience or useful substance. The same weakness can appear in writing produced with no AI at all. A person can recycle familiar advice just as easily as a model can.
The useful distinction is not human versus machine. It is whether anybody exercised judgement before pressing publish.
That explains another change LinkedIn is making. Its “enhance your post” feature is being replaced by tools focused on proofreading and clarity without changing the member’s voice. Assistance remains acceptable. The platform is signalling that authorship still needs to come from the person or business whose name appears above the post.
For brands, this makes “Was AI used?” the wrong internal question. A better one is: “What have we added that a reader could not get from a hundred similar posts?”
Cheap volume creates an expensive feed
Generative tools have made competent language abundant.
They can turn one announcement into a blog post, five social updates, a newsletter introduction and a set of comments in minutes. That efficiency is real. The problem begins when format multiplication is mistaken for insight.
If every company reacts to the same announcement with the same summary, the feed fills with language that is technically correct and practically interchangeable. Readers have to spend more time finding somebody who has tested the idea, noticed a limitation, made a decision or explained what changes in the real world.
That is why generic content can damage more than one post’s reach. It trains an audience to expect little from the next one. A business may continue publishing regularly while slowly reducing the likelihood that customers, recruits or peers will stop and listen.
The production cost has fallen. The cost of wasting attention has not.
A point of view needs raw material
Useful business content usually begins before the writing.
It begins with something observed: a customer question that keeps returning, an experiment that failed, a process that changed, a result that surprised the team, a constraint that forced a decision or a disagreement worth resolving.
An industry announcement can still be the trigger. It should not be the entire substance.
Before commissioning a post, collect the material that gives the business a legitimate reason to comment:
What have we seen in our own work?
What evidence or source supports the claim?
Where does our experience differ from the obvious interpretation?
What decision could a reader make differently after reading this?
Who is prepared to stand behind the conclusion?
These questions are not a demand for a dramatic personal story in every update. A precise technical explanation or carefully sourced analysis can be valuable without becoming autobiographical. The requirement is substance, not theatre.
The article should contain an observation, evidence or decision that survives after the trend name is removed.
Do not make style carry the argument
Much of the discussion about AI writing focuses on stylistic tells: short lines, repeated contrasts, excessive headings, familiar opening phrases or suspicious punctuation.
Editing those patterns can improve readability. It does not create expertise.
A weak post remains weak when its sentences are rearranged to sound more human. Conversely, a structured post is not automatically low quality because it uses a format that AI systems also favour.
The stronger editorial test is harder to game:
Does the post contain a specific claim?
Can the claim be checked?
Is there a meaningful consequence?
Does the author make a choice, recommendation or interpretation?
Would removing the company name leave something any competitor could publish unchanged?
That last question is especially useful. If the same copy could sit comfortably on twenty competitors’ pages, it is probably filling space rather than building authority.
Keep AI in the production process
Businesses do not need to abandon AI-assisted content production.
There is valuable work for it to do: gathering background, tracing claims to primary sources, comparing explanations, finding counterarguments, testing structure, identifying repetition and adapting approved material for different channels.
The boundary is authority.
AI can help organise the evidence. It should not invent the experience, choose a position nobody owns or manufacture confidence where the business is uncertain. A named person should still decide what the company believes, which claims it can defend and whether the piece deserves the reader’s time.
That means approval cannot be reduced to checking grammar and brand tone. Someone close enough to the subject must be able to reject a polished draft because it adds nothing.
Stopping a post is sometimes better content management than publishing it.
Measure what the content earns
LinkedIn’s move is also a reminder to review the metrics used to judge content programmes.
Post count and publishing consistency measure activity. They do not show whether the work strengthened the brand.
Reach and reactions matter, but useful signals may also include qualified replies, saves, direct enquiries, informed disagreement and whether customers refer to the idea later. The right measures depend on the purpose of the content.
The important change is to stop rewarding volume without examining value.
A sensible editorial standard is simple: every piece should contain evidence, judgement and a consequence for the reader. It should sound like a business that knows why it is speaking, not one feeding a calendar.
LinkedIn’s button will not settle every argument about quality. Reader feedback is subjective, and any reporting system needs protection against misuse.
But the direction is clear. Polished language is no longer scarce. Credible experience, careful evidence and accountable judgement are.
The businesses worth following will use AI to support those things, not to imitate them.
Source: LinkedIn, “How LinkedIn is Continuing to Tackle AI Slop”, 15 September 2026.
