LinkedIn's New “Seems Like AI Slop” Button Turns User Reports Into Detection Training Data
LinkedIn is letting members flag AI-generated low-quality posts while simultaneously replacing some of its own AI-writing features with a proofreading tool.
What matters
- LinkedIn added a “Seems like AI slop” reporting option that feeds user flags into its AI-detection training data.
- The company is replacing its own AI writing feature with a proofreading tool, per TechCrunch.
- The explicit “AI slop” label is unusual — most platforms group AI content under generic spam categories.
- It is unclear how false positives, human review, and enforcement thresholds are handled.
- The move highlights the tension between platforms building generative AI tools and fighting the low-quality content those tools produce.
Launch facts
- Availability:
- Testing; availability may be limited
- Platforms:
What happened
LinkedIn has added a new reporting option that lets users flag posts they suspect are low-quality, AI-generated content. The button, labeled “Seems like AI slop,” appears within the standard post-reporting flow and is designed to help LinkedIn train and improve its automated AI-detection tools, according to CNET.
Engadget corroborated the reporting tool and added that LinkedIn is also removing some of its own AI-writing features. TechCrunch provided a more specific detail: LinkedIn is replacing its AI writing feature with a proofreading tool, signaling a shift away from fully generative post drafting toward lighter-touch assistance.
The explicit “AI slop” label is unusual. Most social platforms group AI-generated content under generic spam or low-quality categories rather than calling it out by name. LinkedIn's choice of language suggests it is treating AI-generated junk as a distinct content-quality problem, not just a moderation nuisance.
Why it matters
The move highlights a growing tension for platforms that build generative AI tools: the same companies enabling AI content creation are now spending resources to detect and suppress its lowest-quality outputs. LinkedIn's approach is notable because it turns a user-facing report button into labeled training data for detection models. Every user flag potentially becomes a supervised learning signal for automated classifiers.
That design pattern could be significant. Crowdsourced labels can reduce reliance on purely manual moderation and help platforms build proprietary datasets for AI-content detection. If LinkedIn's detection models improve from this feedback loop, the resulting data could become a competitive asset for feed quality and trust.
However, key questions remain unanswered by the available reporting. It is unclear how LinkedIn handles false positives, whether reports trigger automatic action or human review, and what enforcement thresholds apply before a flagged post is removed or downranked. The feature may also be limited in availability; Engadget described it as being tested, suggesting it may not be rolled out to all users or regions yet.
The simultaneous pullback from LinkedIn's own AI-writing features is worth noting. Replacing a generative writing tool with a proofreading tool suggests LinkedIn is recalibrating how much AI assistance it wants to provide directly, even as it asks users to police AI-generated content from others.
What to watch
Watch for whether LinkedIn discloses how the reported data is used in model training and whether it publishes any metrics on detection accuracy or enforcement outcomes. Also monitor whether other major platforms adopt similarly explicit AI-content reporting categories rather than burying them under generic spam labels. The replacement of LinkedIn's AI-writing feature with a proofreading tool may indicate a broader industry shift toward assistive rather than fully generative content tools.
What to do next
Developers
Study how LinkedIn frames user reports as labeled training data for detection models; consider whether your own content platforms could use explicit AI-content flags as supervised learning signals.
The design pattern of turning a user-facing report button into model-training data is directly relevant to anyone building moderation or detection systems.
Founders
Evaluate whether your product needs an explicit AI-content reporting or labeling surface, and whether user signals can bootstrap automated detection.
AI-generated spam is a cross-platform problem; LinkedIn's approach shows how crowdsourced labels can reduce reliance on purely manual moderation.
PMs
Audit your content-reporting taxonomy and consider adding a dedicated, plainly named AI-content category rather than burying it under generic “spam.”
Clear, user-recognizable labels can increase report quality and give your detection teams cleaner training data.
Investors
Track platforms that are building proprietary AI-content detection datasets from user signals, as that data moat could differentiate feed-quality and trust products.
LinkedIn's move suggests user-reported AI-slop data is becoming a strategic asset for platform integrity.
Operators
If your team manages community or content moderation, review how AI-generated content is currently classified and whether a dedicated reporting path would improve triage.
Separating AI-slop reports from general spam can sharpen moderation workflows and surface emerging abuse patterns faster.
How to test
- 1Open LinkedIn and find a post you believe is AI-generated low-quality content.
- 2Use the post's reporting menu (typically the three-dot menu on the post).
- 3Look for a “Seems like AI slop” option within the reporting flow.
- 4Submit the report and note any confirmation or follow-up messaging from LinkedIn.
Caveats
- The feature may not be rolled out to all users or regions yet; Engadget describes it as being tested.
- It is unclear whether reports trigger automatic action or human review.
- False positives are possible; use the button for genuinely suspected AI-generated low-quality content.