Generative AI: 6 tips for identifying synthetic images

A flawless photograph does not always depict a moment that actually happened. As generative artificial intelligence (AI) tools advance, synthetic images are becoming increasingly convincing, and relying on visual perception alone may no longer be enough to establish what is real.

For Igor Baliberdin, a strategic designer and founder of LOOOP with more than 20 years’ experience at the intersection of design, business and technology, the key shift is recognising that there is no longer one single “AI mistake” that exposes an image. “Technology has advanced significantly. Today, looking only for an odd hand or a misspelt word may work in some cases, but it is not a reliable strategy. The most important thing is to develop an investigative eye and combine different signals,” he explains.

According to the expert, a few precautions can help people assess an image before believing or sharing it. See the tips below.

1. Examine hands, eyes, teeth and fine details

Body parts remain useful areas to inspect, especially when they appear in complex situations. Hands with disproportionate fingers, accessories that seem to blend into skin, teeth that are excessively uniform, or eyes with implausible reflections may point to artificial generation.

“That does not mean every unusual detail is proof of AI. A real photograph can also contain distortions, blur or lighting issues. The point is to check whether several small inconsistencies are occurring at the same time,” says Igor Baliberdin.

2. Enlarge the image and look for inconsistencies

An image may look entirely natural during a quick glance on a mobile phone, yet expose issues once enlarged. Pay close attention to skin textures, strands of hair, object edges, patterns, lettering and elements in the background.

“It is worth looking at what we normally ignore. AI may get the main subject very right while still showing minor inconsistencies in the relationships between objects. The problem is often in the detail, rather than at the centre of the image,” the designer explains.

3. Assess light, shadows and reflections

Lighting can also reveal that an image is synthetic. Check whether the direction of the light is logical, whether shadows follow the objects, and whether reflections in mirrors, glass or metallic surfaces match what is happening in the scene.

“A photograph is the outcome of a series of physical relationships: light, distance, perspective, depth and reflection. When one of those relationships does not add up, that is a sign to investigate further,” the expert advises.

4. Be wary of text and objects that look “almost right”

Signs, packaging, screens, documents and notices are other important areas to examine. Even when AI creates visually convincing phrases, small details may contain unusual characters, meaningless words or information that does not fit the context.

“It is the kind of image that appears correct at first glance, but when you read what is written, you realise something does not make sense. Visual information needs to be assessed alongside the information it is intended to convey,” says Igor Baliberdin.

5. Investigate where the image came from

This may be one of the most important steps. Rather than asking only, “does it look like AI?”, it is worth asking: who posted it? When? In what context? Does the original image exist? Have other reputable outlets or profiles published the same scene? Reverse-image searches can also help uncover the photograph’s source and establish when it began circulating.

For Igor Baliberdin, this change in behaviour is essential. “The source of an image often says more than the image itself. A photograph without context, received in a group or reposted by an unknown profile, deserves a higher level of verification,” he adds.

6. Do not rely solely on your own judgement

Even an image that seems convincing should be examined carefully. Where it could have an impact, especially in news, health, safety or allegations, visual observation should be combined with source research and confirmation from trustworthy sources.

“The more convincing the image is, the less sense it makes to trust intuition alone. The question should not be ‘can I tell that this is AI?’, but ‘what evidence do I have that this image is genuine?’”, says Igor Baliberdin.

New literacy means learning to be sceptical without stopping seeing

The development of generative AI is changing not only how images are produced, but also people’s relationship with what they see. A Gartner survey released in June 2026 found that 49% of American consumers believe generative AI has worsened the quality of available content, a figure rising to 57% among Gen Z and millennials.

For Igor Baliberdin, the answer is not to assume that every image might be fake, but to build a more critical attitude towards digital content. “We are entering a reality in which seeing will no longer be enough to believe. This does not mean distrusting everything, but understanding that images need to be put into context. The perspective of the future is not simply that of someone who sees a good photograph, but of someone who can question the origin, coherence and intention behind it,” the designer concludes.

By Maria Rossi