Fake news & AI content check

Second Look

Paste a post or upload a picture or video. You’ll see whether it looks fake or AI-made, and exactly why.

Add details (optional)

Findings

Example result
Past checks0

    Your checks appear here, saved only in this browser.

    Track record0

    Open this page on claude.ai to see how often checks were right.

    How the estimate works

    Before Claude looks at anything, the page inspects the file in your browser. For pictures it reads hidden metadata: camera make and model, editing software, Content Credentials (C2PA) and generator tags that tools like Stable Diffusion, DALL·E, Midjourney and Firefly can leave behind. For text it measures sentence rhythm, word variety, emotional and urgent wording, and how often sources are cited.

    Then Claude reads it along with those measurements and weighs writing patterns typical of AI models, visual artifacts in hands, text and lighting, manipulation tactics, missing sources, and claims that contradict well-established facts. It marks the exact passages that raised concern.

    Videos are checked as evenly spaced still frames. Sound and motion are not checked.

    Claude does not search the web during a check, so it cannot confirm breaking news. Its knowledge stops at a cutoff date. Use the search links and suggested checks to confirm with trusted sources.

    No detector is fully reliable. Treat the percentages as a starting point for your own judgement, not as proof.