How to detect AI-generated images

AI-generated images – Advantages, disadvantages, and risks.

Images generated by artificial intelligence (AI) are becoming increasingly common on social media. However, they always leave behind distinctive traces, from strange details like extra fingers to nonsensical text. With the proliferation of fake content, detecting them is more important than ever.

AI-generated fake images are spreading at a dizzying pace, creating a new form of 'technological junk' designed to attract clicks. Although GenAI (Generation AI) only emerged in 2022, more than 15 billion AI-generated images are now circulating on the internet.

According to the latest data from OpenAI, more than 2 million images are created daily using DALL-E 2. In 15 months, this tool alone has generated 916 million images. The majority of these are entertaining and unrealistic, such as images of Jesus made from shrimp, animals with human-like features, etc.

Although many AI images still have a humorous or surreal quality, the technology is rapidly evolving and producing increasingly realistic images. This has led to many controversial situations.

In 2024, Republican political activist Amy Kramer shared an AI-generated image of a girl hugging a puppy after Hurricane Helene, using it as evidence to criticize President Joe Biden. More recently, a French woman was scammed out of €830,000 using a sophisticated AI-generated image mimicking actor Brad Pitt.

Despite the ever-improving quality of AI images, there are still ways to identify artificial content. Nghe An Newspaper provides guidance on 7 ways to detect AI-generated images.

1. The sign lies in the human form.

One of the most easily detectable flaws lies in the human-like shape. AI often struggles to recreate hands, resulting in fingers that are too long, too short, or the wrong number (sometimes only four fingers, sometimes six or seven).

Additionally, AI can create unnatural body proportions, disproportionate limbs, or asymmetrical faces. Eyes are another common weakness, as they can appear misaligned, reflect light incorrectly, or look lifeless even though the overall image may seem realistic.

2. Meaningless text

One of the telltale signs of AI-generated images is the presence of meaningless or illegible text. AI models often struggle to accurately reproduce letters, resulting in the creation of nonsensical words.

For example, in AI-generated images, signs, billboards, and product labels often contain jumbled or nonsensical characters. A prime example is the infamous Willy Wonka Experience in Glasgow, where AI-generated posters featured bizarre phrases like "encherining" (possibly "enchanting") or "Cargo Tunnels" (probably "catchy tunes").

3. Shadows and reflections look unnatural.

Light and shadow are key elements that distinguish real images from AI-generated ones. In natural photographs, light always follows the laws of physics, but AI sometimes creates shadows and reflections that don't accurately reflect the actual light source. This makes the photo look strange and unnatural.

Some AI-generated images have overly uniform lighting, making them look plastic or too glossy, giving them a Disney cartoon-like feel rather than real photographs. Additionally, AI images often lack the natural imperfections inherent in real-life photography.

4. Unusual perfection

Another characteristic of AI images is their unusually perfect quality. In real-world photography, especially in low-light conditions, digital noise is common. However, AI either completely eliminates this noise or creates artificial noise uniformly, making the image look unrealistic.

AI also tends to over-smooth textures, making human skin, object surfaces, or fabrics unnaturally smooth. In some cases, AI portraits look like they've been excessively airbrushed, obscuring details like pores or wrinkles that were present on real faces.

5. Unnatural-looking repeating patterns

Another hallmark of AI imagery is the repetition of details. AI often struggles to create completely unique elements, leading to subtle duplication or repetition of details.

For example, in an image containing a crowd, AI can create people with almost identical faces. Or in the background of an image, patterns might appear unusually many times, instead of showing the natural differences seen in a real photograph.

6. Check the context of the image.

Even if an AI-generated image looks convincing, its context may still contain inconsistencies.

AI-generated images often exhibit inconsistencies, such as traffic signs with incorrect colors, architecture that doesn't match the actual location, or light reflections that contradict reality. These can be "red flags" for identifying fake images.

7. Verify authenticity using an image search tool.

If an image looks suspicious, users can use AI image detection tools. While these tools aren't perfect, they can still help check if an image was created by AI.

Another way to verify is to performimage searchThis can be done by searching the internet. This helps determine if the image has appeared elsewhere before, thereby verifying its authenticity and origin.

AI-generated images are becoming increasingly sophisticated, but they still possess characteristics that help identify them. By paying attention to text details, lighting, texture, repeating patterns, and context, we can avoid being fooled by fake content on social media.

Quoc Duan