Google Launches New Website called Detect AI Fakes, a free online tool designed to help users identify whether images or videos were generated by artificial intelligence. The site accepts uploads and returns a confidence score indicating the likelihood of AI manipulation. But the company warns the system is not infallible, and early testing shows even Google's own AI can make mistakes.

What You Need to Know

Detect AI Fakes uses machine learning to analyze visual artifacts common in AI-generated media. The tool is part of a broader industry push to combat deepfakes, but experts caution that no detection system is perfect. Google's own AI, Gemini, misidentified a real photograph as manipulated during internal tests. Users should treat results as one signal among many when verifying content.

How Detect AI Fakes Works

The tool examines uploaded files for telltale signs of synthetic generation, such as inconsistent lighting, pixel-level anomalies and unnatural texture patterns. Google claims the system can detect outputs from popular generative models including its own Imagen and third-party tools like DALL-E and Stable Diffusion. The process takes seconds and does not require users to log in or share personal data.

Known Limitations and Early Tests

Google acknowledges that Detect AI Fakes is not a perfect solution. In a demonstration, the company fed the tool a real photo of Texas Attorney General Ken Paxton standing with former President Donald Trump. Gemini, Google's flagship large language model, falsely flagged the image as containing a manipulated element — a Grimace character that was not actually present. The incident underscores a critical challenge: even advanced AI can produce false positives, especially when trained to look for specific patterns.

The service is no Perfect Gemini when it comes to accuracy. Users must assume it can make errors. Key limitations include:

  • False positives: Real images may be incorrectly flagged as AI-generated.
  • Evasion techniques: Advanced adversaries can tweak outputs to bypass detection.
  • Evolving models: New generators emerge faster than detectors can be updated.

Why This Matters

The launch of Detect AI Fakes arrives at a moment when synthetic media is flooding social platforms and news feeds. Misinformation campaigns increasingly rely on realistic deepfakes to sway public opinion or damage reputations. For journalists, fact-checkers and average users, having a free, accessible detection tool could reduce the spread of harmful content. But the Gemini test reveals a deeper problem: reliance on imperfect AI detectors can create a false sense of certainty. The real-world impact will depend on how transparent Google remains about the tool's error rates and how quickly it adapts to new generation techniques. Policymakers and tech companies alike face pressure to build verification systems that are not only accurate but also trustworthy enough to influence legal and editorial decisions.

The stakes are high. Without robust safeguards, bad actors could exploit detection gaps to discredit genuine evidence, or conversely, use false positives to dismiss real photos as fakes. Google's effort is a step forward, but as the company itself notes, it is not a final answer.