The image no longer needs a photographer, a place, or a fleeting moment to escape the light. It is now born from a sentence. We write: "A child looking out of a train window on a winter morning," and we are presented with a complete image of steam on the glass, wide eyes, dim light, and a longing that no one has ever experienced... There was no child or train, nor was there any morning... Yet, we may feel something real.

This is the new paradox: The image might be fake, but the effect can be genuine. In an article titled "Designing Emotions: How AI-Generated Images Can Evoke Human Connection," the idea is proposed that new imaging tools are not just used to create beautiful images, but also to evoke a specific feeling: joy, calmness, curiosity, empowerment, or belonging. Instead of designers searching through thousands of ready-made images for a "close" match to their feelings, they can now ask the machine for an image designed to evoke the same emotion: A girl drawing at dawn, warm colors, a sense of hope, and tired features that do not diminish optimism. Here, the image is no longer a document of what happened but rather a design of what we want to happen within the recipient.

Perhaps this is the most dangerous and beautiful transformation simultaneously. Old photographs used to say: "This is what I saw." AI-generated images, on the other hand, say: "This is how you should feel." They do not start from reality but from intention, from the psychological state we want to achieve, from that subtle thread between the eye and the heart.

The image, originally, has never been innocent. The advertisement knew that for decades. The background color, the person’s look, the light angle, the arrangement of hands, the distance between faces—all these elements appeal not only to the mind but also to the body. A family picture around a table doesn't just sell food; it sells warmth. A woman running at dawn doesn't just sell shoes; it sells a stronger version of oneself. A child laughing doesn't sell a product; it awakens in us the desire to protect.

But artificial intelligence has made this process faster, more accurate, and more customizable. We are no longer designing one image for a large audience; we can design hundreds of images for hundreds of psychological states: an image that gives reassurance to a patient, another that ignites ambition in a student, a third that sells well-being to an elite audience, and a fourth that gives a sense of belonging to a lonely person in front of the screen. As if emotions were customizable.

Grand View Research estimates that the market for AI image generators reached about $349.6 million in 2023, and is expected to reach about $1.08 billion by 2030, with a compound annual growth rate of 17.7%.

Other estimates from Emergen Research place the market at $2.5 billion in 2025, and expect it to reach $12 billion by 2035. Even if the numbers differ according to each institution's methodology, the trend is clear: artificial images are no longer a game; they are an industry.

In the announcement, specifically, the race seems clearer, with a Research and Markets report estimating that the generative AI market in advertising grew from $2.72 billion in 2024 to $3.39 billion in 2025, at an annual growth rate of 24.6%.

Media reports also indicate that more than half of the 1,000 marketers surveyed in 2025 used some form of AI in creative campaigns, and a significant proportion of ads may include AI by 2026, meaning that the image we see in an ad, post, or charity campaign might not be captured but designed to evoke a specific feeling.

Science confirms that emotional images can remain in memory. Studies in psychology and advertising show that ads that trigger stronger emotional responses are often associated with higher recall and engagement. In a study published in Frontiers in Psychology on the emotional effectiveness of advertisements, researchers treated an ad not just as an informational message but as a system of attention, engagement, and emotion; the recipient does not always remember what was said to them, but they remember how something made them feel.

That’s why AI-generated images seem powerful: because they don’t need to be real to be impactful. The brain doesn’t always ask about the source of an image before being affected by it. It might see a sad face, soft light, or a scene that resembles memory and respond. We do not interact with the image as a digital file but as an emotional signal. If the signal is crafted skillfully, it can connect.

Recent studies have directly tested this point, with one study on the alignment of human and generative models’ emotions using models such as “DALL-E 2,” “DALL-E 3,” and “Stable Diffusion” to generate 240 images expressing 10 positive and negative emotions. Participants then rated how well the image matched the intended emotion. The researchers concluded that the models were able to produce emotional expressions that humans read as aligned with the intended emotions, with varying quality depending on the model and type of emotion. In simpler terms: machines don’t feel, but they are starting to learn what feeling looks like.

Here we enter gray territory. If an artificial image can evoke empathy in us, is that good or dangerous? It could be good when used by an artist to create a new visual world, or by a charity to bring distant suffering closer, or by a visual therapist to help someone express a feeling they don’t know how to say. But the danger starts when the emotion itself becomes an engineered commodity and when images are designed not to tell the truth but to bypass our defenses.

Take humanitarian campaigns: an image of a tired child may prompt people to donate. But if the child is entirely generated, who are we helping? Are we serving a real cause symbolically or exploiting genuine empathy for an image without an owner? In advertising, an image of a happy woman in an ideal home might sell a product, but it could also sell an illusion about life, the body, family, and success. Artificial intelligence doesn’t invent this danger, but it accelerates it, making it cheaper and more abundant.

The problem is not that synthetic images are always false. Sometimes the real image is more misleading than a clearly generated one. The problem lies in transparency. Does the recipient know that what they see did not happen? Do they understand that this face has not lived through the sadness it displays? Do they realize that the shot was designed to touch their fear, hope, or loneliness? When such knowledge is absent, synthetic images can sneak into emotions through beauty.

Perhaps we should treat generated images as we do with rhetorical language. Not every metaphor is a lie, but it becomes dangerous if presented as a report. A synthetic image can be art, an advertisement, therapy, education, or manipulation. The difference lies not only in the tool itself, but also in intention, context, disclosure, and who benefits from emotional impact. And while synthetic images do not have a heart, they know how to reach our hearts; this is exactly where responsibility begins.