URL Scanner · Beta
Scan a whole page for deepfakes.
One URL in. A full report out: is the body text AI-written? Are the images AI-generated? An overall risk score across the whole page.
- · Body-text AI detection (Resemble text API)
- · Per-image deepfake detection (DETECT-3B image head)
- · Unified risk score — one number across the whole page
- · Processed fresh on every scan, nothing stored
Who this is for
Built for people who verify things for a living.
Journalists
Before citing a viral article, scan the URL. Know if the images and copy were AI-generated before it reaches your editor.
Fraud & trust teams
Triage suspected phishing or disinformation pages in seconds. One scan checks every image and the body copy in one pass.
Brand protection
Found a page impersonating your brand? Scan it for synthetic imagery and AI-generated copy to evidence the complaint.
Frequently asked questions
- What does the URL scanner do?
- Paste any web page URL. We fetch the page, extract the main body text and all embedded images, run them through our text-detection and image-detection models in parallel, and give you one consolidated report — overall risk score plus per-asset verdicts.
- What kind of pages work?
- News articles, blog posts, social-media article embeds, and most static pages. Heavy single-page apps may return thinner extractions — for best results on Twitter/X, TikTok, and Instagram, download the media directly and upload to /detect/*.
- Does it scan videos and audio embedded on the page?
- Not yet for embedded <video> and <audio> tags — those typically need direct file access. Paste the direct URL to the video/audio file at /detect/video or /detect/audio instead. Embedded YouTube, Vimeo, and TikTok players are planned in v2.
- Is my scan stored?
- No. The page content is fetched fresh on every scan, text is analyzed in privacy mode (never persisted by the text detector), and we do not retain the report. Every scan is an independent, stateless request.
- Can sites block the scanner?
- Yes — sites can block or rate-limit our crawler. We use a clearly-identified user-agent (DetectDeepfakesBot/1.0) and respect standard rate-limiting signals. If a page returns a block page or requires login, we won't be able to scan it.
- How should I interpret the risk score?
- The score (0–100) is a weighted blend of text and image verdicts — 40% text, 60% image. High risk (70+) means strong evidence of synthetic content across multiple assets. Elevated (40–69) means some signals but not definitive. Use context: a page full of AI-generated stock images is common and not inherently fraudulent.