This article contains the use of generative AI for journalistic purposes.
As my sister and I drove home from watching “Backrooms” in the theatre, we discussed what, outside of the traditional horror aspects, made the film so eerie. We became stuck on the concept that the backrooms seem to remember things, with more details getting lost or warped with each iteration of the memory.
The movie — released in May of this year — follows lonely furniture store owner Clark as he discovers an endless series of familiar yet uncanny places known as the backrooms. The backrooms are reminiscent of an office setting yet contain several strange elements: objects melded into the floor, illogical layouts, and people whose faces appear distorted and fragmented. Clark becomes obsessed. Worried about him, his therapist, Dr. Mary Kline, enters the backrooms to find him and is confronted with far more than what she bargained for. Barely escaping with her life, she discovers that the backrooms take scenes from the outside world and replicate them, corrupting them into something inhuman.
For my sister and me, the idea of replication that “Backrooms” proposed draws a lot of parallels to the functions of generative AI. While “Backrooms” may not be about generative AI, its central concept offers an unexpectedly accurate metaphor for the age of AI-generated content, revealing how originality and human-derived ideas can become muddled in a world increasingly occupied by AI.
The concept for “Backrooms” began with an image posted in 2019 on social media website 4chan that contained an eerie, yellow-lit room reminiscent of an empty office. Social media became filled with images of vacant office spaces, large poolrooms with slides leading to unknown places, and endless suburbs with no one present, contributing to the seemingly limitless environment of the backrooms.
In 2022, YouTube creator Kane Parsons, under the name Kane Pixels, published a short film series titled “The Backrooms,” created using Blender, a free 3D computer graphics software. It became an overnight sensation, with talks about creating a movie occurring just a month after the release of the first video in the series. After choosing A24, an adaptation of Kane Parsons’ “The Backrooms” was announced in February 2023. It is now the studio’s highest grossing movie ever.
The backrooms evoke a sense of familiarity tainted by wrongness. At first glance, they appear to be part of our world and mostly normal. In the movie, when Clark first enters the backrooms, he stumbles into what appears to be just an office space, albeit with a large pile of furniture in the middle. But once Clark explores more, he realizes that the backrooms seem to go on forever and contain a multitude of nonsensical architectural designs: three-knob doors, arbitrarily protruding and receding walls, and incredibly distorted proportions — failed attempts at recreating the normal world.
Due to a similar process of creation, AI-generated content also often exhibits a sense of wrongness. During training, AI learns patterns found in existing media, which it then uses to produce work that appears assembled instead of reasoned through with purpose. In written work, this can present itself in the form of overused sentence structures and hollow statements. Take this sentence, which I prompted ChatGPT-5.5 to generate: “AI is not just transforming the way we create content; it’s also redefining the very fabric of human creativity, opening the door to a future where innovation knows no bounds.” The cliche statements and the “It’s not X, it’s Y” sentence structure show that generative AI follows patterns instead of writing with originality.
Furthermore, the backrooms get stranger with each remembrance. There is a scene in the film where Mary’s childhood home replicates itself, slightly changing with each descending floor, until it becomes something unrecognizable and backrooms-like. First, the natural light disappears, and then the room begins to empty. In the end, her home has fully gained the uncanny quality of the backrooms.
This same effect can be seen in generative AI. When AI is asked to recreate the same image over and over again, each iteration tends to become more distorted than the last. This is because AI can mistake small errors for intentional details and amplify them over time, leading to warped features, oversaturated colors, and even a completely different visual style than the original input.
Finally, the backrooms lack originality: they are “stealing” from the real world. They exist as a manifestation of deformed memories rather than an authentic creation.
Similarly, AI is taking its toll on originality. Many artists report seeing AI versions of their work on sites like X and TikTok, and Molly Crabapple of the Guardian has coined the phenomenon as “the greatest art heist in history.” Furthermore, Alvin de Rooij of Tilburg University and Micheal Biskjaer of Aaarhus University found that AI use tends to have a homogenization effect, making online writing and other works more similar across users. This effect is even bleeding over into human speech. Everyday use of some of ChatGPT’s favorite words, such as “delve”, “showcase”, and “meticulous”, have all increased since its launch.
While my sister and I found “Backrooms” to be refreshingly unique, the backrooms themselves do not create anything new. Everything contained in the seemingly endless expanse of the backrooms is a distorted remembrance of the real world. Similarly, generative AI produces never-ending content by leeching off and replicating what humans have already created. As AI becomes increasingly prevalent in social media, education, and other aspects of daily life, the concept that “Backrooms” presents feels almost perfectly suited to the world we find ourselves in. Contrary to the human vibrance and uniqueness we are accustomed to, the rise of generative AI risks submerging us in a world like “Backrooms”: sanitized and homogenized beyond recognition.
