Fact check: How do I spot AI-generated images?

In today’s digital age, anyone with an internet connection and an AI tool can generate photorealistic images within seconds, often for free. These AI-generated images are spreading rapidly across social media, making it increasingly challenging to distinguish between real and manipulated photographs. The quality of AI tools has improved significantly, reducing the number of noticeable errors, or "AI hallucinations," that can tip off users to the image's origins.

One challenge is that some images are entirely AI-generated, while others are real photos that have been subtly altered using AI. For instance, a broadcast image from the 2026 World Cup showing a fan replaced by a Hitler lookalike was created using an OpenAI-powered tool. To spot AI-generated images, it’s crucial to take your time and carefully examine the details.

Common AI errors might be hidden in the background, such as distorted objects or nonsensical text. However, if you don’t have the ability to zoom in sufficiently, online tools like Forensically, a free photo forensics platform, can assist in closely analyzing the image.

One verification technique involves looking beyond the pixels and questioning whether the depicted event makes sense. For example, the Hitler lookalike would have likely been noticed by broadcasters, stadium officials, and other spectators, yet no reputable media organizations reported this incident, and no fans in the stadium shared posts featuring the lookalike.

Checking whether the event has been covered by trustworthy news organizations can help verify its authenticity. If an image appears to depict an unusual or controversial incident with no evidence in credible reporting, it should raise suspicions.

Another technique is to examine who is posting and sharing the image. If the account often shares AI-generated material or lacks a connection to the event, the image should be treated with caution. Conducting a reverse image search can also be helpful. Tools like Google Images or TinEye allow you to upload an image and uncover its source.

While some AI-generated images may be entirely new and yield few or no results, manipulated images are often based on older, authentic pictures. In such cases, a reverse search can reveal the original image, providing important context.

If an image has already been fact-checked, it’s essential to determine if the organization is certified by the International Fact Checking Network (IFCN). The IFCN assesses fact-checkers based on standards such as fairness, transparency, impartiality, and open corrections. A list of IFCN-certified outlets, including DW Fact check, is available online.

AI detectors can also aid in assessing whether an image was created or modified using AI. These tools search for invisible watermarks and digital traces associated with AI-generated content. Google’s SynthID and OpenAI’s verification tool are examples of such systems.

However, it’s important to remember that AI detectors are not foolproof and should be considered as one piece of the verification puzzle rather than definitive proof. Cross-checking results using multiple tools can help reduce the risk of drawing incorrect conclusions. Even after following these steps, uncertainty may remain. In such cases, it’s advisable not to share the image, especially on social media, and instead privately warn friends and family that the image’s authenticity has not been confirmed.

If still unsure, you can send the image to a fact-checking service, such as factchecking@dw.com, for further assistance.