The experiment behind the claim

A study published on arXiv (2409.08895) examined how exposure to AI-generated images affects people's recollections. Participants were shown a mix of real photographs and AI-generated images and later asked which images they remembered seeing. The study's published figures include examples of the stimuli used and the experimental flow.

The research found that participants sometimes reported remembering AI-generated images they had never actually seen — a phenomenon consistent with established findings that suggestive or fabricated visual material can distort human memory. The feeling captured in the claim — "That's strange, I'm sure I saw this photo" — mirrors the subjective experience participants described when mistaking fabricated images for genuine memories.

Why it matters

As generative AI images spread across social media and news feeds, the risk is not only deception in the moment but lasting memory distortion: an image seen once can be recalled later as a genuine experience. The researchers' materials suggest practical implications for eyewitness testimony, misinformation, and how media literacy efforts should account for AI-generated content.

Limits of the claim

The viral framing should be read with some caution. The study demonstrates memory distortion under experimental conditions with specific stimuli and participant groups; it does not show that AI images universally implant false memories in everyone. The claim's tone of an "eerie experiment" is broadly accurate to the research, though individual results vary.

The claim that AI can create unsettling false memories of images people are sure they saw is Mostly True.