SlopScan quietly scores every post and comment for AI-likelihood while you scroll. Hover any badge to see what it spotted. Blocklist the worst offenders. And, when you're ready, find the real humans worth following.
LinkedIn used to be where people posted real wins, real flops, and the occasional cringe. Now half the timeline reads like it was written by a model at 4am, 3 coffees deep.
"I fired my best employee. Here's why it changed my life." Every post is a six-line hook, a fake confession, and a question to farm comments.
Every model trained on the same corpus reaches for the same words. You can almost smell the temperature setting.
Identical structure. Identical em dashes. Identical closing question. Different headshot. The signal-to-noise ratio of your feed is collapsing.
SlopScan runs locally in your browser. It reads the post, scores it, adds a small badge. That's it.
One click from the Chrome Web Store. No account needed on the Free tier; SlopScan doesn't ask.
Every post and top-level comment gets a small badge: green for likely human, amber for mixed, red for likely AI. Hover for the breakdown.
One-click blocklist for serial offenders. SlopScan then surfaces creators in your industry who consistently score human. The ones worth following.
There's no magic AI-detector and we don't pretend otherwise. What we do: blend seven well-studied signals, show you exactly which ones fired, and let you decide. The score is just the start of the conversation.
Words and short phrases that show up far more in LLM output than in real writing. You know the ones: "delve", "tapestry", "navigate the complexities", "a testament to".
Em-dash density, Oxford-comma rate, sentence-final hedges. Models have stylistic fingerprints that show up across thousands of posts.
Humans vary sentence length wildly. Models tend toward uniform medium-length sentences. We measure variance.
The "hook → 3-bullet list → rhetorical question" template is the single biggest tell on LinkedIn. We score adherence to it.
"It's worth noting", "while it's true that", "however, it's important to remember". The soft-edged scaffolding that models can't help but produce.
Posts that end with a tidy, generalising summary or "What are your thoughts?" almost never come from a hurried human typing on their phone.
The giveaway shapes that LLMs reach for over and over: "It's not X, it's Y", "Not just X, but Y", "The catch?", and the rule of three. Catches even thesaurus-laundered posts.
The core detector is free and always will be. Premium is in private beta. It adds a hosted model with higher accuracy plus a few quality-of-life features. Join the waitlist to get early access (and a launch discount).
£0 / forever
Everything you need to clean up your feed today.
Pricing TBD
For people who live on LinkedIn and want a sharper filter. Launching later this year.
Founding members get a launch discount and early access. No spam, just one email when it ships.
No detector is. SlopScan is a heuristic. It blends seven independent signals and hands you a probability. Treat the score the way you'd treat a spam filter: useful, opinionated, occasionally wrong. We show the breakdown for every post so you can make your own call.
On the Free tier, no. Detection runs entirely in your browser, so the post text never leaves your device. On Premium, posts you hover are sent to our hosted model for a higher-accuracy score; we don't store them, and your own posts are never analysed.
SlopScan is a read-only DOM annotator. It doesn't automate clicks, scrape data into a server, send messages, or interact with LinkedIn's API. From LinkedIn's perspective, it's invisible. We've designed it specifically to stay on the right side of their terms.
Honestly, this is the part that worries us most. A few lexical tells overlap with second-language English, which is why no single signal can flag a post on its own. To turn the badge red, a post has to score high on several independent signals at once: structure, burstiness, closure, punctuation, sentence templates. We're still tuning, and we'd rather miss real AI than mislabel a real person.
Yes. One click in settings, no support ticket required. Your blocklist stays with you on the Free tier.
By default, yes. Your blocklist lives in local browser storage. On Premium it's synced through our servers (encrypted at rest) so you can use SlopScan on multiple machines. We never share or sell it.