HumanSignal

LinkedIn has never been noisier. Thoughtful takes and real experiences are mixed in with posts that were clearly drafted in 30 seconds with a prompt. It's hard to tell the difference at a glance — and it's exhausting to try. HumanSignal quietly labels every post and comment as you scroll. No popups, no sign-in, no settings to figure out. Just a small badge that tells you whether what you're reading feels like a real person wrote it. How it works A fast rules engine scores each post the moment it loads. An on-device AI model (RoBERTa, running entirely in your browser) refines the result within a few seconds. The badge updates — from "Possibly AI" to "Feels Human" or "Likely AI" — as confidence improves. Everything stays on your device The AI model runs locally. No text is sent anywhere. HumanSignal doesn't have a server, doesn't collect analytics, and doesn't know you exist. What the labels mean — Feels Human: specific, personal, grounded in experience — Possibly AI: generic or templated, but could go either way — Likely AI: strong structural signals of AI-generated text — Almost Certainly AI: high-confidence detection It's not perfect. No detector is. But it's honest about that — labels come with confidence levels, and you can always dismiss a badge you disagree with.