LinkedIn adds a “seems like AI slop” report button as AI-generated posts flood the feed

The professional network has a professional problem: a feed filling up with machine-written inspiration, and it is now crowdsourcing the cleanup. LinkedIn has introduced a button that lets users flag posts as “seems like AI slop”, one of the first times a major social platform has handed users a direct tool against AI-generated content. The data behind the move explains why they had to.

The slop numbers behind the button

LinkedIn’s chief product officer Hari Srinivasan has called the problem a top priority, and the Pangram data shows why. More than 40 percent of long-form posts on the platform were flagged as fully AI-generated, and LinkedIn carries nearly two-thirds of all AI-generated long-form social media posts, making it the most AI-saturated major platform. Anyone who has scrolled the feed lately did not need the study: the formulaic hooks, the odd cadence, the comments that read like the posts.

The platform’s incentives built this. LinkedIn rewards consistent posting with reach, and AI made consistent posting free. The result was predictable: an arms race of automated authenticity, thought leadership at industrial scale, and a user base quietly tuning out.

How the new reporting works

The mechanics are deliberately gentle. Flag a post and it disappears from your feed; the poster is not banned or publicly shamed but receives a private note in their account dashboard that their content came across as inauthentic. Srinivasan framed the goal as giving creators feedback from real humans instead of relying solely on automated detectors, which is a tactful way of saying the platform knows AI detectors alone are not trusted enough to act on.

Why human flags beat machine detectors

The choice to lean on user reports rather than pure detection is the technically honest one. AI text detectors remain unreliable, with false positive rates that would be a lawsuit generator on a platform full of professionals. Humans are excellent slop detectors in the aggregate, not because they can prove authorship, but because they can smell content written for nobody. The flag measures reception, not origin, which sidesteps the detection problem entirely: it does not matter whether a model wrote it if it reads like one did.

What actually counts as slop

The term is doing more work than it appears to. Slop is not simply “written with AI”; plenty of professionals use AI to draft, and plenty of humans write vapid content unassisted. What users recognize as slop is the pattern: the generic hook, the structured listicle cadence, the inspirational pivot, the concluding question designed to farm comments. It is content produced to satisfy a posting schedule rather than a reader, and AI made it infinitely cheap to produce. The report button effectively asks users to flag that pattern, wherever it came from, which is a more honest target than authorship.

This is also why the gentle feedback loop matters. A platform that publicly punishes suspected AI use would trigger disputes it cannot adjudicate. A platform that quietly tells posters “this read as inauthentic to real people” applies social pressure without needing proof. It is moderation by vibe, scaled, and on a platform drowning in formulaic sincerity, vibes are arguably the correct instrument.

The incentives that have to change

The button treats the symptom. The cause is an algorithm that rewarded daily posting with reach, which made automation rational. If LinkedIn genuinely wants less slop, the feed ranking itself has to stop paying for volume, and there are early signs it knows that: reducing the distribution of flagged content is ranking change by another name. The real metric to watch is not how many posts get flagged but whether the feed’s baseline quality recovers enough that users stop reaching for the button.

What it means for the rest of social media

Every platform is watching this experiment. If LinkedIn’s crowdsourced approach measurably cleans up the feed without false-positive chaos, expect versions of it everywhere. The deeper shift is philosophical: platforms spent two years racing to add AI writing tools, and are now building tools to suppress the output of those tools. That contradiction is not hypocrisy so much as the market correcting itself. Generation was the easy part; making content worth reading remains unsolved.

The provenance side of this fight is advancing in parallel, with approaches like Anthropic’s model-level text watermarks aiming to make AI content self-identifying. Between source watermarking and destination flagging, the era of frictionless AI slop is getting squeezed from both ends.

For creators, the calculus now changes in a concrete way. Using AI as an editor or brainstorm partner was always defensible; shipping raw model output as daily thought leadership now carries a visible cost, because enough flags quietly kill your distribution. The rational response is not abandoning AI but raising the floor: adding the personal detail, the specific number, the actual opinion that a model cannot supply. The professionals who treated AI as a drafting tool rather than a ghostwriter were always going to win this transition. The button just makes it official.

There is a final irony the platform surely appreciates: LinkedIn itself has spent two years shipping AI writing assistance to its users, including tools that help draft exactly the kind of posts now being flagged. That is not a contradiction so much as the entire industry’s position in miniature. Everyone is selling the water and bailing the boat simultaneously, and the platforms that thrive will be the ones that figure out how to keep the assistance while draining the flood.

LinkedIn’s product updates are posted on LinkedIn’s official site, with the detection data from Pangram.

The bottom line

Should every social platform add an AI slop flag? Tell the tech desk.