I scrolled past a LinkedIn post and felt the same soft horror you do when the words read like a corporate echo. You paused, realized the voice wasn’t human, and wanted a button to punish the offender. That button is arriving.
A recruiter told me their feed had become a loop before I even finished my coffee — here’s what LinkedIn is adding
LinkedIn product chief Hari Srinivasan said the company is adding a “Seems like AI slop” option to the three-dot menu on posts and comments. You tap it; the platform logs your feedback and the poster sees a private analytics note. Srinivasan framed this as a human signal to tune models and improve feeds.
The network has already blocked billions of AI-generated posts and comments in recent months, Srinivasan says, and the new button is one more lever to push quality upward.
How does LinkedIn’s “Seems like AI slop” button work?
When you flag a post, LinkedIn treats that action as a human-labeled example for its classifiers. The platform will fold that signal into new models while also retiring the old AI-driven “enhance your post” feature in favor of a proofreading tool that promises to preserve your voice. Expect fewer generic one-size-fits-all posts and more attempts to separate low-quality automation from authentic human material.
I watched a Substack thread explode — the industry fights back
Substack CEO Chris Best publicly announced integration of the AI detector Pangram after criticizing platforms that reward “fakeness.” Pangram’s report, which Best cited, argues that up to 40% of long-form LinkedIn text is generated.
Pangram’s analysis claimed LinkedIn produced a third of scanned items yet accounted for 62% of flagged AI content. That triggered a public back-and-forth: LinkedIn pushed human feedback; Substack pressed detection tools. The Atlantic has reported skepticism about detection accuracy, reminding us that automated detectors can be wrong.
Will reporting remove AI-generated posts?
Not instantly. Flagging contributes to a dataset that improves automated classifiers and moderation decisions over time. LinkedIn will combine your flags with algorithmic signals and manual review to reduce slop, but human and machine errors remain part of the process.
A colleague who learned English later told me they fear the system — and they aren’t alone
Human reviewers and machine detectors both misfire. Many of the traits labeled “AI-like” are simply good grammar or stylistic preferences—em dashes, formal cadence, tidy paragraphs—that can unfairly penalize non-native English writers. Studies have shown human detection of AI prose is nearly a coin toss, and toolmakers themselves often rely on models whose explanations are opaque.
That creates a double risk: innocent creators could be flagged, and sophisticated generators could slip through. LinkedIn says it will roll out improved classifiers and lean on human feedback to reduce false positives, but the balance is delicate.
Can AI detection harm non-native English writers?
Yes. Detection methods often favor a narrow set of stylistic cues, so writers who write clearly and formally may be mistaken for automation. If platforms use these flags to downrank or penalize content, non-native speakers could lose visibility. That’s why LinkedIn is pairing user flags with model adjustments rather than automatic bans.
One manager told me they now treat the feed like a crowded conference — what you need to do
You can be a gatekeeper. Start clicking the new button when a post feels hollow, robotic, or repetitive. Your clicks won’t erase slop overnight, but they teach LinkedIn which posts to demote and which to surface.
The move is also an industry mirror. TikTok and Substack are wrestling with similar noise; platforms are experimenting with detection tools, moderation policies, and user signals to fight a flood of AI content. Some companies favor third-party detectors like Pangram; others lean into human review and improved classifiers.
The feed has become a conference where everyone recites the same slide deck, and LinkedIn is handing users a broom to sweep the attic.
If you care about preserving real voices on your feed, will you start pressing the button when the slop shows up?