Google SynthID: How to Check an Image for AI Marks

Updated 5 November 2026 · Standard formula, shown below

SynthID can help you check whether an image may contain an invisible mark associated with Google’s generative AI tools. Use Google’s supported image-checking feature, where available, and treat its result as one clue—not proof of who made the image or how it was edited. You’ll need the image file or access to the image in a supported Google service.

What You’ll Need

The image you want to check, preferably the original or a high-quality copy.
A Google account if the checking feature you use requires sign-in.
Access to a Google product or feature that currently offers SynthID image checking in your region. Availability and supported formats can change.
A browser or compatible mobile app with an internet connection.

Step-by-Step

Find the original image if possible. Save the file from its original source rather than using a screenshot, thumbnail, or image copied from a social post. Reposting, resizing, compression, cropping, filters, and edits may make a watermark harder to detect. Keep an untouched copy before trying any changes.
Open a Google feature that explicitly supports SynthID checking. Look for an image verification or “about this image” capability in a Google AI product or other Google service. Follow Google’s current help instructions to confirm that the feature accepts images and checks for SynthID. Do not assume that every Google image search, Lens result, or chatbot response performs a SynthID scan.
Sign in if prompted and review the feature’s terms. Some tools may require an account, have age or regional availability limits, or handle uploaded content under specific privacy terms. Check what is uploaded and how the service says it processes it before submitting private, confidential, or sensitive images.
Choose the image for inspection. Use the upload or image-selection control provided by the feature. Select the original file from your device, or choose an image already available in that service if the option is offered. Make sure you have selected the intended image, especially if several files have similar names.
Submit the image and wait for the result. Let the service complete its analysis; avoid closing the page or app while it is working. The tool may report that it detected a SynthID watermark, did not detect one, or could not determine a result. Read the exact wording rather than reducing it to a simple “AI” or “not AI” label.
Interpret a detection carefully. A positive result means the service found signals it associates with SynthID in the image. It does not necessarily identify the person who generated it, the prompt, the date, or whether the image has since been altered. It also does not establish that every part of a composite image came from AI.
Interpret no detection cautiously. A result that says no SynthID was found is not proof that a human made the image. It may have been created with a different tool, lack a supported mark, or have been transformed enough that detection is difficult. Treat the result as “not detected by this check,” not a universal authenticity verdict.
Save the result and verify the source separately. If the image matters for a report, moderation decision, or fact-check, record the tool and date of the check and preserve the original file. Look for the original publisher, context, metadata, and independent corroboration. Ask the creator or source for provenance when appropriate.

Common Mistakes

Assuming every AI image has SynthID. SynthID is associated with supported Google generative AI systems; it is not a universal watermark embedded in every AI-generated image.
Treating a negative result as proof of human authorship. The checker looks for a particular kind of mark, not every possible sign of AI generation.
Checking a screenshot instead of the image file. Screenshots and repeated downloads can change image data and reduce the chance of detection.
Using a general search result as if it were a watermark scan. A search engine’s image description, source label, or AI summary is not necessarily a SynthID analysis.
Reading more into a positive result than it says. Detection alone does not reveal the full editing history, identity of the creator, or whether the image is truthful.
Uploading sensitive material without checking privacy details. An image can contain faces, documents, location clues, or other private information. Review the service’s current terms and use a suitable alternative if needed.

Tips

Keep an unedited original and check that before trying a compressed or cropped copy.
Use only a feature that specifically says it checks for SynthID; product capabilities and availability can change.
Record the exact result wording. “Not detected” is more accurate than writing “confirmed human-made.”
For important decisions, combine the check with source verification, publication history, and other evidence. No single watermark check settles every authenticity question.
If you are checking a disputed image, avoid altering the only copy. Make a duplicate for any additional tests and note where each version came from.

Troubleshooting

You cannot find an image-checking option: The feature may not be available in your account, country, app, or current product interface. Check Google’s current official help information and try again through a supported browser or service. Do not rely on an unrelated feature as a substitute.
The upload control is missing or the file is rejected: Confirm that the image type and size are accepted by the specific service. If necessary, make a copy in a commonly supported image format without overwriting the original. Avoid unnecessary conversion because it may affect detection.
The check fails or keeps loading: Check your connection, reload the page, and try again. If the issue continues, try a current browser or the service’s official app, and check whether the service is temporarily unavailable.
The result is inconclusive: Try the original file rather than a screenshot or social-media copy. If the original is unavailable, report the outcome as inconclusive and investigate provenance instead of repeatedly editing the image in an attempt to force a result.
The result conflicts with other evidence: Preserve the result and the file, then seek the original source or independent corroboration. A watermark check and other evidence may address different questions; neither should be treated as a complete account of an image’s history.

Frequently Asked Questions

What does a SynthID result actually tell me?
A detection indicates that the checking feature found a signal it associates with SynthID. It is evidence of a supported mark, not a complete record of the image’s creator or editing history.

Does no SynthID detection mean an image is not AI-generated?
No. It means the check did not detect that watermark. The image could come from another AI system, have no watermark, or have been changed in ways that affect detection.

Can SynthID check images made by any AI generator?
No. SynthID checking concerns marks associated with supported Google AI systems. It is not a universal detector for all image-generation tools.

Will checking a cropped or compressed image still work?
It may, but transformations can make a watermark harder to detect. Check the original, highest-quality file available for the best chance of a useful result.

Can SynthID tell me who made the image?
No. A detection does not normally establish the creator’s identity, the prompt, the exact generation date, or the image’s full chain of edits.

Is the image-checking feature available everywhere?
Availability can depend on the product, account, region, and current rollout. Consult Google’s current product help for the feature you plan to use.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.