PixelSniff

1. •Drop, paste or pick an image (JPG, PNG, WebP or HEIC up to 50MB), or paste a link. Several images can be selected for a batch check.
2. •A basic analysis runs first and returns an AI likelihood score with a plain verdict. Scores at or above 70 read as AI, at or below 30 as human, and anything in between is reported as uncertain.
3. •For a closer look, an opt-in deep analysis adds a forensic layer: a region heatmap, separate detector signals and written reasoning. A basic check costs 1 credit and a deep check 5 credits.
4. •Each result gets a private, unlisted share link, and the full report can be downloaded as a PDF or JSON. When demand is high, scans wait in a queue and the result can be emailed when ready.

• •AI likelihood score - one combined 0 to 100% score with a clear verdict instead of a vague label.
• •Multi-engine cross-check - each detection engine scores the image independently, and the report shows how strongly they agree.
• •AI heatmap - deep analysis highlights the regions that contributed most to the AI score, useful for spotting partial edits.
• •Watermark and metadata checks - looks for Google SynthID watermarks, C2PA Content Credentials and camera EXIF data, while noting that missing metadata proves nothing on its own.
• •Written reasoning - key indicators, visual patterns and recommendations explain why an image scored as it did.
• •Shareable reports - unlisted share links you can revoke, plus PDF, image and JSON exports and a check history across devices for signed-in users.

PixelSniff suits people who need a quick second opinion on whether an image is real: shoppers checking product photos, people vetting a dating or social profile picture, moderators and community managers screening uploads, and writers or researchers who want a signal before sharing a viral picture. Creators and teams who check images every day can move to the Pro plan for a monthly credit allowance and deep forensic reports, while organisations needing volume, API access or invoiced billing can ask for an Enterprise arrangement. The vendor's own terms ask users not to treat a score as the sole basis for decisions about a person, a claim or a piece of evidence.