Watermarks and their scope
Nano Banana is the popular name for Google's Gemini image generation and editing models. Google says media made with its AI tools carries SynthID, an invisible watermark embedded in the content itself rather than stored as an EXIF field. In the Gemini app you can upload an image and ask “Is this made with AI?” to check for that signal; Google reported in May 2026 that the check had been used 50 million times and was expanding to Search and Chrome. A filename containing the nickname does not establish which model or export route produced a file.
Source: Google: identifying AI-generated media (opens in a new tab)
Preserve the edit history
Many Nano Banana questions concern edits to an existing photo rather than a wholly generated scene. Ask for both the original capture and the edited output if the owner can share them. Compare the changed region in context: an object added to a photograph is a different claim from inventing an entire scene. A single overall score cannot explain how much of an image is synthetic or certify each region.
Use the best available file
Check the original JPG, PNG, WebP or HEIC export when possible. A compressed social preview usually carries fewer useful signals than the file Gemini produced, and resizing, cropping or screenshotting can make any result harder to interpret. Note those changes as limitations of the evidence. They are not instructions for removing marks or bypassing checks, which Image Evidence does not support.
Interpret a watermark result
A positive SynthID result from Google's tool suggests that Google AI was involved somewhere in the file's history; it does not say how much was changed or whether the caption is accurate. A negative result means no Google watermark was detected within that tool's coverage. The image could come from another generator, an older workflow, or a copy changed enough to weaken the signal. Image Evidence has not measured a Nano Banana-specific detection rate and does not claim coverage of every model version.
Verify a claim responsibly
Use Google's official route for watermark questions, then review the result against the claimed scene. A watermarked image can be legitimate creative work, and an unwatermarked photograph can still carry a misleading caption. Preserve the creator's explanation and ask permission before checking private material. A watermark check is informational, not an authenticity certificate, and should not decide a consequential question about a person on its own.
How do I check if an image was made with Nano Banana?
Upload the original file to the Gemini app and ask whether it was made with AI; Gemini looks for Google's SynthID watermark. Google has also added C2PA Content Credentials verification to Gemini, which can show whether a file records its editing tools. If the image is part of a dispute, also check its metadata and ask the creator for the unedited source.
Does Nano Banana add a watermark?
Google says content generated with its AI tools includes an invisible SynthID watermark, and that covers images created or edited with Gemini's image models. The watermark is not visible and is not stored as ordinary metadata, so a metadata viewer will not show it. Only a SynthID-aware tool, such as the check in the Gemini app, can look for it.
Can a Nano Banana edit be detected if only part of the photo changed?
Possibly, but less predictably. A watermark check reports whether a signal was found and may not show precisely which region changed, and a general detector may score a lightly edited photo as real. The most reliable evidence is often the unedited original next to the edited version, together with the editor's own description of what was changed.