To check a file for Google’s SynthID in Gemini, sign in, attach one image, video, or audio file, and ask whether Google AI created or edited it. Gemini’s help page describes this file-by-file workflow and the limits on its uploads.

Check a file in Gemini

You need a signed-in personal Google account or an eligible Google Workspace account. If you use the Gemini app, Google says to use its latest version. Then:

  1. Sign in to Gemini on a computer or in the app. This gives you access to its verification feature.
  2. Attach one image, video, or audio file. You can upload from your device, Google Photos, or Google Drive. For Drive uploads, Keep Activity must be on and Google Workspace must be connected to Gemini Apps.
  3. Ask a focused question, such as whether Google AI created or edited the file. Gemini checks for a SynthID watermark, a signal embedded in supported media.

Gemini and the public Detector have different scopes

Google announced on October 7, 2026, that public access to SynthID Detector had expanded worldwide in English. The announcement described checks for images, video, and audio from Google and participating partners, naming OpenAI, NVIDIA, and Kakao. Google described Apple as a future partner in that announcement. NeoTeo covered the public-access announcement in Spanish.

Gemini’s verification feature is separate. Google’s help page says Gemini currently recognizes SynthID in content created by Google AI tools. So the partner list in the public Detector announcement should not be read as Gemini’s recognition scope.

OpenAI separately says that supported images made with ChatGPT, Codex, or the OpenAI API use SynthID, and that it extended the watermark to supported OpenAI-generated audio. That statement applies to supported outputs, not every response or file. OpenAI’s provenance announcement describes its use of the watermark.

Read a result as a watermark signal

A positive Gemini result means that all or part of an image or video was created or edited with Google AI. It does not establish that the entire file was generated from scratch, who made it, or whether its claims are true. For video, Gemini may identify a portion where it found SynthID; that does not mean every frame in the highlighted portion carries the watermark.

A negative result means Gemini did not detect a Google SynthID watermark. It does not rule out content from another AI system. Google also says repeated alterations can make a watermark harder to detect. An unclear result can occur when a file has too little detail or an edit is too small to leave a detectable mark.

A creator-reported test illustrates why a single result needs a narrow reading. In a video published October 15, 2026, AI with Surya said its test detected six of eight files described as watermarked. The creator also reported detections after WhatsApp-style compression and an Instagram-style filter, but misses for an AI image inside a phone screenshot and a vertically reposted Veo clip. These are observations from that creator’s test, not a general accuracy rate. Watch the creator’s report.

Gemini file limits and daily quotas

Google’s limits apply to Gemini Apps. Each check accepts one file, up to 100 MB. Videos must be under 90 seconds, and audio files must be under one hour.

Google lists approximate quotas in any rolling 24-hour period:

  • Images: about 10 checks.
  • Video: about 10 checks, with five minutes of video total.
  • Audio: about 10 checks, with three hours of audio total.

These limits can shape how you prepare a file for Gemini: choose the specific image, clip, or audio recording you want to check, and keep it within the relevant size and duration limits.

SynthID and Content Credentials

SynthID is an embedded watermark. Content Credentials are compatible provenance metadata that can describe a file’s composition, editing history, and whether AI was used. Google says those credentials can apply to AI-made and non-AI content, so their presence alone does not establish that a file was generated with AI.

How the Detector reached public access

Google says SynthID technology began in 2023. The SynthID Detector portal followed in 2025, when Google introduced it for early testers. On October 7, 2026, Google announced expanded public access worldwide in English. Google DeepMind’s SynthID page and its 2025 announcement for early testers distinguish the portal’s history from the watermarking technology’s earlier start.