LinkedIn adds a button to report AI-generated ‘slop’

LinkedIn just confessed that its own product strategy helped pollute the feed. The Microsoft-owned network spent months pushing an "enhance your post" button that rewrote users' thoughts into generic corporate cadence. Now it's yanking that feature and replacing it with a proofreader that leaves the voice alone. The reversal is the clearest signal yet that platforms are losing control of the synthetic tide they helped unleash.

The new "seems like AI slop" reporting button sits beside every post. Click it and you train LinkedIn's classifiers to spot the next wave of synthetic filler. Users become unpaid moderators for a platform that monetized the very tools producing the garbage. Hari Srinivasan, the chief product officer, framed the move as caring about "real people" and "real perspectives." He didn't mention that LinkedIn's parent company has invested billions in the models flooding the zone.

Cloudflare reported last week that bot traffic now exceeds human requests across the web. That milestone arrived ahead of schedule. Substack partnered with Pangram to flag AI-written newsletters. Pangram just raised $9 million to fight the flood. Digg shut down its Reddit clone in March because bots overwhelmed it. The infrastructure is buckling. LinkedIn's admission that it blocks "hundreds of thousands" of automated comment attempts daily — millions in recent months — confirms the scale. The platform isn't curating conversation anymore. It's running a perimeter defense.

The private "inauthentic" flag in user dashboards is the most revealing addition. LinkedIn will quietly tell you when your post smells synthetic. The company bets this nudge will improve writing rather than drive users away. That's a gamble. People who lean on AI to sound competent won't welcome the signal that they sound hollow. They'll either strip the assist and post less, or they'll learn to prompt better — producing slop that passes the classifier. The arms race moves to the prompt layer.

Verification tools are expanding. Profile and page badges will signal human provenance. Comment blocking from unverified accounts adds another gate. These are sensible perimeter moves. But they treat authenticity as a binary — verified or not, human or bot — when the real problem is a gradient. A verified executive pasting ChatGPT output is still slop. An unverified engineer writing raw insight is not. The platform's classifiers will struggle with that distinction because the platform's business model still rewards volume and engagement over substance.

Microsoft wants AI everywhere. Copilot in Office, Copilot in Windows, Copilot in GitHub. LinkedIn is the only major Microsoft property publicly retreating from AI generation. That fracture matters. It suggests the consumer tolerance for synthetic content has a hard floor, and LinkedIn hit it. The professional network's value proposition — trusted identity, credible insight — collapses when the feed becomes a content farm. No amount of classifier tuning fixes a business model that incentivizes performative posting.

The "seems like AI slop" button won't stem the tide. It externalizes the cost of detection to users while the platform keeps the engagement metrics. A more honest move would be demoting synthetic posts in the feed algorithm — but that would depress time-on-site. So LinkedIn chooses the theater of a report button and the quiet withdrawal of its own enhancer. The slop keeps coming. The classifiers keep training. The users keep clicking.