The extension
Recheck with model (Pro)
Ask a vision-language model for a second opinion on a post. The answer is cached and shown to everyone who meets that post.
Updated
Recheck with model appears in the card footer for Pro accounts. It sends the post's text, or the image URL, to our API, which asks a model running on Cloudflare Workers AI to score it against a fixed rubric and explain itself.
What you get back
The card gains a Model recheck row with the model's probability and its notes, for example about hands, text in the image, textures, lighting, a stock-photo look or uncanny symmetry. The combined score is updated: when a classifier result also exists the two are blended (60% classifier, 40% recheck); otherwise the recheck stands on its own. A trusted provenance signal still outranks both.
Limits
- Pro accounts get 300 rechecks per rolling 24 hours. Past that the button returns a rate-limit message until the window moves on.
- Rechecks are off by default and only run when you ask.
- On YouTube and other video players only the thumbnail or poster is rechecked, never the video frames. On the free tier the extension never uploads image bytes; a Pro recheck can capture a frame from a same-origin video, but never from cross-origin YouTube, Vimeo, TikTok, Twitch or Facebook players.
- A recheck is a second opinion, not a verdict. Models are confidently wrong sometimes; the card shows the reasoning so you can weigh it.
Shared with everyone
The result is written to the verdict cache against the content hash, so the next person to see that post, on any tier, sees the recheck without spending anything. The cache entry never records who requested it. See Community reports and the verdict cache.