14.07.2026 - Artificial intelligence has long been part of everyday life and is increasingly being used in scams too. Deceptively realistic emails, professionally designed websites and AI-generated images can now be created in just a few minutes. In this year's summer series, the NCSC will explain how scammers use AI-generated images and how to recognise fakes.
Artificial intelligence (AI) is already an integral part of everyday life in many areas, and has become a useful tool in the workplace to varying degrees. This development has not stopped at scams. Scam emails can now be written convincingly and error-free in many different languages. Perfectly designed websites, fake videos and images, and even conference calls in which people are impersonated can be created in next to no time with the help of AI. This year's NCSC summer series looks at how AI-generated content is used in scams and how fakes can be recognised. This first instalment focuses on AI-generated images and the general developments seen in recent years. The second part looks at text and websites, while the third part covers current scams involving fake images. The summer series concludes with a look to the future.
It is important to distinguish generative AI from more traditional forms of AI. Spam filters in email programs, for example, have used AI for many years. But generative AI only became widely known to the general public with the launch of ChatGPT in 2022. What makes it different is that it does not simply analyse existing data; it can also create new content itself, including text, images and music.
In recent months, the NCSC has received numerous reports of cases in which AI-generated images have been used for scams, many of them on fraudulent investment websites. Fake Swiss Post receipts and other AI-generated images are also being used in fake job adverts and phishing attempts designed to steal login details.
The difficulty lies in the fact that AI models are constantly evolving. This means that a valid identifying feature today may not be valid tomorrow. While many of the warning signs described below are becoming less reliable as AI models improve, they can still provide useful clues when assessing whether an image may have been generated by AI.
Hands
Hands long posed one of the biggest challenges for AI. Irregularities could be seen, for example, in the number or length of fingers, but also in the texture of the skin, which often appeared plastic-like. However, as AI models improve, this indicator is becoming less important. The latest models from this year contain hardly any obvious errors, though they may still include illogical details.
Faces / people
Faces have also posed problems for AI in the past, with telltale errors often appearing in details such as the eyes, skin, hair and teeth. Here too, however, AI has made significant progress. While earlier models often struggled to render faces convincingly, the latest models produce much more natural-looking results. Even so, small inconsistencies can still appear, such as mismatched reflections in earrings, jewellery that disappears into the neck, or ears that are unusually long.
Landscapes
AI-generated landscape images are often used in scams to make them appear more trustworthy, for example by giving a website or offer a familiar Swiss backdrop. Here too, certain features may indicate the use of artificial intelligence. Sunrise and sunset scenes often have exaggerated colour contrasts, creating effects that appear almost impossible. AI-generated images also tend to be repetitive, with objects such as flags or houses appearing multiple times. In reality, houses in a village or town rarely all look identical. This could be an indication that AI has been used.
Backgrounds and groups of people
One well-known problem with AI is its faulty logic. Backgrounds very often appear blurred, as would normally only be the case in photos taken in portrait mode. AI also has shortcomings when generating larger groups. Objects may either be missing or appear in excessive numbers, for example too few plates and too many glasses for the number of people shown. As described above, inconsistencies can also appear when hands and limbs are generated, resulting in two left hands or distorted fingers. Complex scenes also continue to cause problems for AI and can result in people, especially in larger groups, having distorted or blurred faces. Backgrounds may also appear totally out of focus in otherwise normal situations. Objects can merge into one another. For example, chairs may turn into a table, or cutlery may disappear into people's hands. Perspective and spatial logic can also provide clues, for instance if walls, windows or other elements are not straight or do not visually align. Finally, shadows that do not fall in the same direction can be a further indication that AI has been used.
Text in images
Even today, AI still has difficulty inserting realistic-looking text into images. While earlier models had major problems generating individual letters, text in newer models often looks as though it has been digitally overlaid rather than naturally integrated into the image. Distorted letters and spelling mistakes can still be an indication that AI has been used.
Other features
Other clues for detecting AI-generated images can be found in subtle inconsistencies. Light, shadow and perspective may not quite add up, for example if shadows fall in different directions, lines are distorted or spatial structures do not fit together logically. Repetitive patterns can also be revealing, with elements such as windows, people or textures looking strikingly similar. The same applies to small details, such as the number or positioning of buttons on clothing. Finally, there is often a general sense of artificiality: at first glance, the image may look convincing, but upon closer inspection something doesn't seem quite right.
Conclusion
Although there are still many features that can help identify AI-generated images, spotting them will become increasingly difficult. As generative models improve rapidly, many of the classic warning signs are becoming harder to detect, or even disappearing altogether. Rather than relying on obvious errors, it will become more important to consider whether the image makes sense as a whole. The key will no longer be to look for isolated, obvious mistakes, but to notice subtle inconsistencies in perspective, context, proportions or internal logic. In future, people will need to look more carefully at images that seem convincing at first glance, but do not quite add up on closer inspection.
Current statistics
Last week's reports by category:
Last modification 14.07.2026
