Meet Rehumanize.
Visit Rehumanize →Rehumanize shows, on each post in your feed, whether it's likely written by a person or by AI. It's built for everyday readers and, in its research build, for the scholars who study them.
Synthetic doesn't mean false. But invisible origin means incomplete context.
Fact-checkers ask whether a claim is true. Rehumanize asks the question they can't: is the voice behind it human? A thousand AI-written comments can fake the look of public agreement without a single false sentence. That manufactured consensus is a new dimension of misinformation, and in a normal feed it's invisible.
Rehumanize works in the background as you scroll Reddit, X, Threads and LinkedIn. A small marker on each post tells you whether it's likely AI-generated, partly AI-generated or likely human-written. Nothing to paste, nothing to switch on.
Media literacy usually means a workshop. Rehumanize builds it into the feed, so every scroll becomes a moment of awareness.
It never says a post is wrong, only that it's likely not human. What that means is your call.
The goal isn't lower trust. It's trust that matches who is really speaking.
Your feed, your markers.
A quiet pill is the default. Want it louder? Switch on a text highlight, an amber border, or both.
A research instrument, not just an extension.
Most studies measure what gets posted. Rehumanize measures what people actually see. The research build turns everyday scrolling into a consented, configurable study, run from one dashboard.
The public version never stores post text or usernames. Research data is collected only from participants who opt in.
Split participants into groups and choose which markers, and which display style, each group sees. Run labelled and unlabelled conditions side by side.
Sentiment, topic across 19 categories, emotion, irony, hate and offensive language, from peer-reviewed models. Each has its own switch.
Extra data fields each have their own switch, off by default. Turn on exactly what your ethics approval covers, and nothing more.
Watch detections arrive in real time as participants scroll. Made for pilots, check-ins and demos.
An overview, charts, filters and row-level browsing across platforms, topics and time.
Exports formatted for CooRTweet and D3lta, open-source tools for finding coordinated campaigns, plus a custom export. Account IDs are pseudonymised by default.
Recorded consent, withdrawal at any time, a deletion-request portal with an approval queue, and a retention setting you control.
Every dashboard action is written to an audit log. Admin and view-only roles, plus access keys for collaborators.
Ships with a codebook for every field and starter analysis scripts in R and Python.
Switch to token mode and type demo. The demo uses sample data.
Got a question worth building?
Tell us your field, your question, and the tool you wish existed. Your idea stays yours: we'll help you build it, or pair you with a lab member to build it alongside you.
hello@jointhinkspark.com