What These Studies Cannot Tell Us
AI-mediated appearance images raise a real body-image question, but the evidence does not answer every question that concern can generate. This chapter audits the difference between an observed result, a plausible cause, and a clinical or developmental claim. The reader's task is practical: when someone says that a filter or generated self-image is harmful, ask what was measured, who was studied, what alternative explanations remain, and what future evidence could change the judgment.
That audit is not a way to dismiss the findings. It is how a useful signal is protected from exaggeration. The adult experiment provides a short-term randomized comparison. The child study provides cross-sectional associations and a richer set of contextual measures. Their limitations do not make them worthless; they define the range of conclusions that can responsibly be drawn from them.
Start with the observed outcome
The most direct outcomes in the available evidence are body satisfaction and related body-image or self-image measures. In the adult experiment, body satisfaction decreased after participants viewed ten AI-generated images, regardless of whether they were informed about the role of AI. In the child study, greater reported G-AI exposure was associated with lower body satisfaction, and higher parental mediation was associated with higher body satisfaction.
Those are meaningful outcomes, but they are not interchangeable with every form of psychological harm. Body satisfaction is a person's evaluation of their body at the time and under the conditions measured. Body image is broader and can include perceptions, thoughts, and feelings about appearance. Self-image and identification with a digital self add further dimensions. None of these labels, by themselves, establishes an eating disorder, body dysmorphic disorder, persistent anxiety, or a lasting change in self-worth.
The same distinction applies to “appearance self-esteem.” It can be a useful phrase for the human concern at stake, but the child study's reported scales focused on physical appearance, body-image satisfaction, and identification with digital representations. The adult abstract reports body-image measures and trust in technology. The source material does not provide a direct long-term measure of appearance self-esteem, and it does not directly measure whether social norms about normal appearance changed.
Self-discrepancy belongs in the same category. The child article proposes that a stylized digital self could widen the gap between an ordinary appearance and a desired appearance. That is a coherent mechanism and a useful hypothesis. It is not a directly established outcome here. Neither available study reports a dedicated self-discrepancy measure or a longitudinal test showing that discrepancy mediates the effect of exposure on body satisfaction.
The diagnostic rule is simple: keep the claim at the level of the instrument and design that produced it. A lower score is evidence of a lower score on that measure. It becomes evidence about a mechanism only when the mechanism is measured or the design can identify the relevant pathway. It becomes evidence about clinical harm only when clinical outcomes are assessed with appropriate methods.
Association, short-term effect, and developmental claim
Three forms of statement often appear next to one another but should not be treated as equivalent.
An association says that two measured variables moved together. The child study's negative correlation between G-AI exposure and body satisfaction describes such a relationship.