Images made with artificial intelligence (AI) are now seen almost everywhere online. They are used regularly on social media, in Google image search and on Pinterest, and even in companies’ advertisements. The spread of poor-quality or excessive AI content has grown so much that a new term for it, “AI slop”, has become popular.
As the technology advances, it is becoming harder to tell real photographs from AI-made images. Modern image generators can produce realistic pictures within seconds. So before taking any picture seen online to be true, it is important to check a few things.
1. Look for oddities in text
One of the easiest ways to identify an AI-made image is to examine the writing in it. If the lettering on a poster, book cover, signboard, T-shirt or advertisement looks distorted, blurred, incomplete or meaningless, the image may have been made with AI.
Although modern AI produces text better than before, small mistakes often remain. Zooming in on the writing in a suspicious picture can therefore help.
2. Watch hands and the body
AI has long had trouble with human hands and fingers. Be wary if a picture shows extra fingers, fingers fused together, distorted nails or unnatural knuckles.
Look not only at hands but also at wrists, elbows, the neck, legs and the proportions of the body. Unnatural eyes, noses and teeth can also point to an AI-made image.
3. Faces that are too perfect
Sometimes the person in a picture looks so perfect that although it seems real, something unnatural is hidden in it. Excessively smooth skin, unnaturally bright eyes, extremely neat hair or unnaturally arranged teeth are common features of such images.
If the person in the picture seems unnaturally flawless compared with real life, it may be AI-made.
4. When everything looks unnaturally perfect
AI is often used in pictures of food, products or advertisements. If a dish looks impossibly perfect, with no marks, spots or natural imperfections anywhere, the picture may have been created artificially.
Real photographs usually have light, shadow, oil or other small imperfections. In AI-made images these often look unnaturally perfect.
5. Too much clutter or odd detail
AI-made images sometimes include far more elements than are needed. Unnatural patterns in the background, shadows at impossible angles, reflections that do not match reality or unnatural lighting can be features of such pictures.
Often the whole picture looks as if it is trying to show many things at once. That kind of excess drama can point to the use of AI.
6. Fine detail goes missing
When AI is used to colourise or enhance old photographs, important fine details are often lost. The brickwork of a building, the distinct features of leaves or the fine pattern of a garment may become blurred.
As a result the picture begins to look like a painting instead of a real photograph. That too points to the possible use of AI.
Technology can help with checking
Besides observation by eye, pictures can be checked with various technical tools. Information about a picture’s source and context can be found with the Circle to Search feature on Android phones or the About this image feature in Google Lens.
An idea of whether a picture is AI-made can also be had by uploading it to Google’s Gemini app. Some tools can also help with pictures carrying Google’s SynthID watermark.
Are detectors fully reliable?
Experts say no AI image detector is 100% accurate. Sometimes an AI-made image is identified as a real photograph, and a real photograph may be wrongly shown as AI-made.
So when checking an important picture, do not rely on a detector’s result alone; examine the picture’s source, context and visual features as well. For pictures connected with news, people or important events in particular, verifying the original source is the safest method.
বাংলায় মূল প্রতিবেদন পড়ুন · Read the original Bengali report