There is no single test that proves an image is AI-generated, but a few checks together give a reliable answer most of the time. Start with the image’s source and any built-in labels, then look for hidden watermarks and content credentials, and only then study the picture itself for visual clues. This guide explains how to tell if an image is AI-generated, which free tools help and where those tools fall short.
Where did the image come from?
The fastest check has nothing to do with pixels. Ask who posted the image first and whether a trustworthy source stands behind it. A news photo from a known agency or an official account is very different from a dramatic picture shared by an anonymous account with no original link.
Run a reverse image search with Google Lens, Bing or TinEye. If the image first appeared on an AI art community or in a creator’s portfolio of AI work, you have your answer. If it only appears in recent reposts with no original source, be cautious.
How can you check for an AI watermark?
Some AI tools add an invisible watermark when they create an image. Google’s is called SynthID. According to Google, you can upload an image to the Gemini app and ask “Was this created with Google AI?” or “Is this AI-generated?”, and Gemini will check for the SynthID watermark and explain what it finds.
The limit is important: this check only detects images made or edited with Google’s AI tools. An image without a SynthID watermark may still come from another AI generator.
What are Content Credentials (C2PA)?
Content Credentials are a standard from the Coalition for Content Provenance and Authenticity (C2PA) that attaches a signed record of how an image was made and edited. Google says images generated with its Nano Banana Pro model in the Gemini app, Vertex AI and Google Ads carry C2PA metadata, and many cameras, editing apps and AI tools from other companies support the standard too.
You can inspect these credentials with a C2PA viewer such as the Content Credentials verify site. Keep in mind that metadata can be stripped when an image is screenshotted or re-uploaded, so missing credentials do not prove an image is real.
Which visual clues still give AI images away?
- Text: signs, labels and book covers with garbled, misspelled or nonsense letters.
- Hands and small details: extra or merged fingers, jewelry that melts into skin, mismatched earrings.
- Repeating patterns: faces in a crowd that look alike, or bricks and tiles that change pattern halfway.
- Physics and light: shadows pointing different ways, reflections that do not match, objects floating or blending together.
- Too-perfect skin and backgrounds: a glossy, airbrushed look with dreamy, blurred backgrounds.
These clues are getting harder to spot as image models improve, so treat them as hints, not proof.
Can AI detector websites be trusted?
Online “AI image detectors” can be a useful extra signal, but they make mistakes in both directions. They can flag real photos as AI and miss AI images, especially after compression or cropping. Never rely on a single detector score to accuse someone or to share a claim publicly.
What about AI-generated text and video?
The same approach applies. For video, look for unnatural blinking, mouths that do not match the audio, flickering edges and objects that change between frames, and check the original source. For text, detectors are even less reliable than for images, so judge the source and look for verifiable facts instead.
A quick checklist before you share an image
- Find the original source with a reverse image search.
- Check for platform labels such as “AI info” or “Made with AI”.
- Look for Content Credentials or ask Gemini about a SynthID watermark.
- Zoom in on text, hands, reflections and shadows.
- If you still are not sure, do not share it as real.
To learn which AI tools create images and how to use them responsibly, read our AI tools guide.
Curious which tools make these images? See our comparison of the best AI image generators.
Posting AI images yourself? Learn when Instagram, TikTok and YouTube require it in how to label AI-generated content on social media.
Scammers also use fake messages with AI images. Learn how to spot a phishing email or text.
Frequently asked questions
Can Gemini detect images made by other AI tools?
Not with SynthID. Google says the check only covers images made or edited with Google AI. Google has said it plans to support C2PA credentials from other sources in the future.
Does a missing watermark mean an image is real?
No. Many AI tools do not add a watermark, and metadata is often removed when images are re-uploaded.
Is it legal to post AI-generated images?
Generally yes, but many platforms ask you to label realistic AI images, and using real people’s faces can raise legal and ethical issues. Check each platform’s rules.
What is the safest rule to follow?
Trust the source more than the pixels. Use watermark and credential checks when they are available, treat visual clues as hints and pause before sharing anything that is surprising, emotional and unsourced. For more ways to protect yourself online, see our guide on AI chatbot privacy settings.
Sources
- Google: How we are bringing AI image verification to the Gemini app
- C2PA: Coalition for Content Provenance and Authenticity
- Content Credentials: Verify tool
This article is based on the official documentation and reporting listed in the sources above, checked on October 17, 2026. Features and prices change, so confirm details on the official pages before you buy. ShokokoTech does not accept payment for coverage. Read our editorial policy.
