AI Post Spotter

Marks likely AI-generated social media posts as you scroll, using platform labels, image metadata and on-device text checks.

AI Post Spotter marks posts that look AI-generated while you scroll X (Twitter), Reddit, Instagram, Facebook and LinkedIn. You can turn it on for other sites yourself. It never says "definitely AI". Each post gets a score and a plain-language list of reasons, labeled "Likely AI" or "Possibly AI". Hover the marker to see why. WHAT IT LOOKS AT • Platform labels: AI-generated content tags the site itself shows. • Image provenance: Content Credentials and other AI-generator tags embedded in image files, read directly on your device. • Hidden characters: invisible Unicode (zero-width characters, TAG characters, variation-selector tricks, look-alike letters). Context rules ignore legitimate uses such as emoji sequences, French spacing and right-to-left text. • Writing style: cautious stylometric checks (emoji-headed bullet lists, stock chatbot phrases, template structure). One tell alone never flags a post. • Optional remote detector: off by default. If you want it, connect your own API key for a third-party AI-image detection service. WHAT YOU CAN DO • Choose how AI posts are shown: a neon-green slime splat, a subtle badge, collapsed to a one-line bar, or hidden. • Collapse or hide ads and sponsored posts (collapsed by default). • Adjust sensitivity from Conservative to Aggressive, switch sites on or off, and "wipe off" a marker you disagree with. • Optional synthesized splat sound (off by default). PRIVACY Everything runs locally in your browser. Post text and images are analyzed on your device. Nothing is collected by, sold to or sent to the developer. There are no analytics, accounts or tracking. The optional remote detector only runs if you turn it on with your own key, and it then sends image links to the provider you picked. Full policy: https://gist.github.com/jleboube/6d6bc16614b1ef9dba7a6ad44520dd30 HONEST LIMITS: PLEASE READ • No tool can reliably tell whether a post was made with AI. Treat the score as a hint, not proof. It will miss AI content and will sometimes flag posts written by people. • Most platforms strip image metadata on upload, so provenance checks often find nothing. A clean result does not mean human-made. • Hidden-character checks can be fooled, and invisible characters also come from copy-paste and some keyboards. • Writing-style checks can flag people who simply write in a "LinkedIn voice". Short posts are not style-checked. • Statistical AI watermarks cannot be detected without the AI vendor's own detector, and this extension does not claim to. • Social sites change their layouts often. Some posts may be skipped until the extension is updated.