Why Some People Can Spot AI Images More Easily Than Others

New research suggests that the ability to detect AI-generated faces may depend less on intelligence or technical knowledge and more on a fundamental visual skill known as object recognition.

A Surprising Predictor Of AI Detection

As artificial intelligence tools become increasingly capable of generating realistic images, concerns about deepfakes and digital misinformation have grown rapidly. Synthetic faces created by AI systems now appear regularly across social media, advertising and online content, often looking convincingly real.

A new study from researchers at Vanderbilt University (in Nashville, Tennessee) has examined why some people are better than others at detecting these images. The findings suggest that the key factor is not intelligence, technological expertise or familiarity with AI tools, but a more basic perceptual ability.

Object Recognition

The research was led by Isabel Gauthier, professor of psychology at Vanderbilt University, together with Jason Chow and Rankin McGugin. Their study, published in the Journal of Experimental Psychology, found that individuals with stronger object recognition skills consistently performed better at identifying AI-generated faces.

Object recognition is a broad visual ability that allows people to distinguish between very similar objects quickly and accurately. In scientific research it is sometimes referred to as the “o factor”, a domain-general skill involved in recognising patterns and structures across many different visual tasks.

Testing The Ability To Detect AI Faces

To investigate how people recognise synthetic images, the researchers developed a new evaluation tool called the AI Face Test. Participants were shown a mixture of real photographs and faces generated by artificial intelligence systems and asked to determine which images were authentic.

The study then compared each participant’s performance with a range of cognitive and perceptual abilities, including intelligence, face recognition skills and familiarity with artificial intelligence technology.

The results revealed that object recognition ability is the strongest predictor of success in detecting AI-generated faces.

In contrast, factors that might seem more relevant, such as intelligence or experience with AI tools, showed little relationship with performance.

A Useful Visual Ability

As Professor Gauthier explained, “these results highlight a visual ability that has very general applications. It’s a stable trait that helps people meet new perceptual challenges, including those created by AI.”

The researchers were particularly surprised that technological experience did not appear to help participants distinguish between real and synthetic images.

“We were shocked to see how intelligence or even technology training did not help accurately judge if a face is AI,” Gauthier said.

Why Some People Are Better At Object Recognition

It seems that some people are just naturally better at this particular skill. Object recognition ability varies between individuals, but those with stronger visual processing skills are better at detecting small structural differences in images. This means that when looking at AI-generated faces, they are more likely to notice subtle inconsistencies in areas such as lighting, texture or facial proportions that others may overlook.

It’s An Underlying Perceptual Ability

In the Vanderbilt study, participants with higher object recognition scores consistently performed better at identifying AI-generated faces in the AI Face Test. Their performance also remained stable when tested again later, suggesting the skill reflects an underlying perceptual ability rather than something people quickly learn through experience with AI tools.

Looking Beyond Obvious Visual Errors

Researchers believe the advantage does not come from spotting obvious “AI mistakes”. Instead, people with stronger object recognition ability appear better at interpreting complex visual structure when the differences are subtle and the signals are noisy.

Can This Skill Be Improved?

All is not lost for those who do not naturally have this skill. There is some evidence that aspects of object recognition can be improved through training. For example, exercises that involve comparing similar objects, analysing small visual variations and practising detailed visual inspection can strengthen perceptual judgement over time.

Useful In Medical Imaging and Radiology

Research in fields such as medical imaging and radiology shows that targeted visual training can improve a person’s ability to recognise subtle visual differences. That said, people with stronger object recognition skills often perform better in visually demanding tasks, including identifying lung nodules in medical scans, recognising cancerous blood cells, reading musical notation and analysing retinal images.

A Wider Skill With Many Applications

Object recognition ability has been linked in previous research to success across a wide range of visually demanding tasks. The Vanderbilt University study takes things one step further by also challenging the widely repeated claim that AI-generated images are now impossible for humans to detect.

“There is this general message we hear in the media that AI images are so realistic that we can’t tell the difference, and I think that’s misleading,” Gauthier said.

According to the researchers, the results instead show a distribution of abilities across the population. Some people struggle to detect synthetic images, some perform moderately well and others identify them with high accuracy. Understanding these differences may become increasingly important as generative AI technologies continue to evolve.

What Does This Mean For Your Business?

For organisations concerned about misinformation, digital trust and online security, the research highlights an important point about the human side of AI detection.

Many current discussions about identifying synthetic media focus on technical solutions such as watermarking systems, detection algorithms or digital authentication tools. These technologies will likely remain important as AI-generated content becomes more widespread.

However, the new research suggests that human perception also plays a significant role. Individuals differ in their natural ability to interpret complex visual information, and this may affect how easily they recognise AI-generated imagery.

For businesses that rely on visual content, such as media organisations, marketing teams and social media platforms, understanding these differences could help shape training programmes, moderation strategies and verification processes.

As AI-generated media becomes more common across the internet, combining technical safeguards with a deeper understanding of human perception may become an increasingly important part of managing digital authenticity.