Our photo of canola fields, or Gemini's? Only the label told them apart

Two pictures of canola fields: one we took ourselves, one Google Gemini generated. We checked both, and a screenshot of the Gemini image.

• Our photo: taken on a phone.
• Generated by Google Gemini: downloaded as the original file.

What changed

• The label. The Gemini file carries Content Credentials signed by Google, declaring it AI-generated. The report verified the signature.
• The estimate. It couldn't tell them apart: 15% for our photo, and 21% for a screenshot of the Gemini image, which the report then called "likely authentic". The screenshot had lost the label.
• The watermark. None found in any of the three. Google says Gemini images also carry a SynthID watermark that only its own tools can read.

What the report said: Authenticity F for the Gemini original, certified AI-generated. Our photo: 15%, "likely authentic".

A low AI estimate on its own was never an all-clear.

Updated 15 October 2026

We checked all three files again. The findings above are what the report said when we first ran this.

• The Gemini original reads F "AI-Generated (Certified)", from its signed label.
• Our photo reads C "No AI-Generation Evidence Found": a 1% estimate, and the pixel analysis found no signs. It grades Authenticity B.
• The screenshot also reads C, with an 18% estimate, and the pixel analysis found no signs.

The label is still the only thing that told them apart. The pixel analysis found no signs in the screenshot or in the Gemini original. What each card and reading means: How to read your forensic report.

Go deeper

• How to check one yourself: Is this photo real, fake or AI-generated?
• Why the label matters: What C2PA Content Credentials prove, by Kenneth Springer
• Try it on your own photo