In September 2026, researchers Mattea Sim, Yael Eiger, and Tadayoshi Kohno published a preliminary study titled "AI-Powered Human Memory Manipulation." In it, they tested whether misleading summaries generated by AI affect the recall of an event that participants watched in a clip about a traffic accident. The experiment showed that those who read a misleading summary were less able to recall the event accurately compared to those who read a correct summary. What makes the result significant is that the misinformation was not necessarily a deliberate campaign; errors or omitted details in an automated summary may be enough to affect the recall of the event. However, this research remains preliminary and should not be treated as a final verdict on all uses of artificial intelligence.
In a study published in September 2026 in the journal *Computers in Human Behavior*, Maria Soto-Sanfiel and Gina Junhan Fu investigated how the public receives historical narratives presented through characters revived by deepfake technology, compared to written texts. The study found differences in memory and truthfulness assessment between the two media; memory and truth judgments for the fake clips declined at a faster rate than for the texts. This is a result that calls for caution against assuming that the realism of an image alone makes it more convincing or more deeply rooted in memory. This confirms that remembering an event is influenced by what reaches a person after it occurs, and by the way they are asked to recall it, which confuses perception and weakens trust in news and institutions.
The risks extend beyond individual memory to collective memory. A fake clip about a national event, or a recording attributed to a public figure, might spread and then be recirculated until it becomes familiar. Some people may confuse what they saw the first time with what they read in a subsequent explanation. In media, courts, and education, the result may be a dispute over evidence and the source of information, not just a passing difference of opinion.
In Jordan, calling for the detection of fake images is not enough; we need a culture of digital verification that begins with questioning the original source, the date of publication, the context, and the existence of independent sources confirming the incident. Educational and media institutions can teach the public to distinguish between evidence and interpretation, and platforms should clarify synthetic content as much as possible. As for the user, they must retain sources and not consider familiarity or personal trust as proof of the accuracy of information.
The 2026 research does not prove that artificial intelligence will make humans unable to distinguish between reality and memory, but it provides serious reasons to study this possibility and deal with it responsibly. The problem is not just that the machine can fabricate a narrative, but that its narrative may overlap with our own narrative about ourselves. In an age where it is easy to manufacture evidence, protecting memory becomes a shared responsibility among researchers, media, schools, platforms, and individuals; because truth is not measured by how convincing a story is, but by our ability to verify the evidence that supports it before it turns into a collective certainty.
