Do Not Open This File: Cursed Images, Found Footage, and the Horror of Believing the Image

Once an image is treated as a trace, visual absence stops being merely absence. It becomes evidence of something beyond the visible field. The unknown origin of Smile Dog can imply a concealed origin. The missing portion of Mereana Mordegard Glesgorv can imply unbearable content. The monotony of Suicide Mouse can become preparation for a rupture. The chaotic imagery of Deyes.avi can make the act of perception itself feel unstable.

* * *

A photograph of a dog is not frightening because dogs are frightening. Mickey Mouse walking down a street is hardly the raw material of horror. A man looking into a camera and smiling contains no event that, by itself, should inspire terror. Yet online, images like these have repeatedly acquired the status of dangerous objects: pictures one should not look at, videos supposedly recovered from somewhere they should never have existed, recordings whose complete versions have allegedly disappeared because they were too disturbing to remain accessible.

The peculiar power of the “cursed file” lies precisely in this gap between what an image visibly contains and what the viewer has been told it is.

Internet horror did not invent the cursed object. Cultures have long attached powers to objects that cannot be located in their physical substance alone: amulets protect, relics preserve presence, ritual objects become dangerous or sacred because of the histories and practices attached to them. Digital culture relocates this old mechanism into a strange new material form. The object is now infinitely reproducible. It can be copied without deterioration, renamed, compressed, reposted and separated completely from its original source. Yet rather than making mystery impossible, this reproducibility often intensifies it: the less stable the origin of a file becomes, the easier it is for another origin to be invented. A cursed file is therefore not simply a frightening image with a story attached to it. The story changes what kind of object the image appears to be.

This distinction matters because some of the most effective examples of digital horror are visually almost banal. Their power does not always reside inside the frame. It may be constructed before the viewer sees the first image, in the warning attached to a link; outside the frame, in speculation about a missing continuation; or after viewing, when the file is passed to somebody else. At other times, the image begins to resist the legend and works more directly through repetition, distortion, noise, and failed recognition.

Four well-known pieces of internet folklore – Smile Dog, Suicide Mouse, Mereana Mordegard Glesgorv, and Deyes.avi – reveal four variations of this process. Together, they also illuminate a larger cinematic question. Found-footage horror works by exploiting much the same instability. It asks the spectator not merely to watch an image but to imagine that the image has a history independent of the film: someone recorded it, something happened to them, and what survives is the recording.

The most frightening question is no longer simply, “What am I looking at?” It becomes: “Why does this image exist?”

The Image Before the Image

Ordinary film viewing already depends on context. A close-up does not arrive without a before and after; editing teaches us how to interpret it. A neutral face can appear compassionate, hungry, or horrified depending on the shot that precedes it. Outside cinema, context operates just as forcefully. Research into perception has repeatedly demonstrated that expectations and surrounding information affect how ambiguous visual material is interpreted (Summerfield & de Lange; 2014; Wieser & Brosch, 2012). We do not simply receive an image and then think about it: what we expect helps organise what we notice.

The cursed file radicalises that process. Imagine encountering an old photograph without explanation. You might examine the clothing, the room, the expression on a face. Now imagine receiving exactly the same photograph beneath the sentence: “This is the final picture taken of her before she disappeared.”

Nothing in the photograph has changed. Yet almost everything in the act of looking has. A shadow in the background becomes suspicious. An awkward expression begins to resemble fear. Empty space seems to conceal information. The photograph is no longer merely depicting something; it appears to contain evidence, and the viewer searches it retrospectively for the event promised by the story.

Internet horror repeatedly turns this preparatory information into an aesthetic device. “Do not watch the full version.” “This video was deleted.” “Everyone who saw this image experienced the same nightmare.” “The original file can no longer be found.” Such sentences perform work that in conventional cinema might be accomplished through music, editing, or a reaction shot: they direct attention before viewing begins.

The crucial point is that literal belief is unnecessary. A viewer may know perfectly well that a creepypasta is invented and nevertheless become more alert to the image after reading it. The story supplies a pattern; the viewer tests the image against that pattern. A file supposedly capable of causing nightmares will be examined differently from the same JPEG presented as amateur photo manipulation.

The legend therefore becomes something like an invisible first shot.

Smile Dog: The Legend Looks Back

Smile Dog, usually circulated under the filename smile.jpg, offers perhaps the clearest demonstration of this relationship. The image shows a dog facing the camera, altered so that its mouth appears unnaturally wide and its teeth disturbingly human. Around it a story developed in which those who see the picture are haunted by dreams and compelled to “spread the word” by passing it on to somebody else.

There is something unpleasant in the image itself. A dog is one of the most familiar visual objects imaginable, which makes the small violations more effective. The proportions of the grin are wrong. The teeth seem to belong to another species. The animal is recognisable enough that the deviations cannot simply be dismissed as abstraction (Kätsyri et al., 2015). But remove the accompanying story and Smile Dog becomes something considerably less specific: an uncanny piece of digital manipulation.

The legend transforms those visible irregularities into symptoms. Once the viewer has been told that the image is dangerous, the grin stops being merely grotesque. It begins to look intentional. The eyes become worthy of prolonged inspection. Colour, compression, and darkness can all appear to conceal further information. The viewer performs part of the horror effect by looking for the reason they were warned.

This helps explain why the uncertain origin of smile.jpg strengthens rather than weakens the object. Normally, establishing that an image has been manipulated, reposted, or detached from a reliable source should diminish its authority. In folklore, missing provenance can do the opposite. An identifiable creator would contain the object within a conventional history: somebody made it, at a certain time, for a certain reason. An uncertain origin leaves a blank. The story fills it.

Line Henriksen (2018) has described Smile Dog in relation to creepypasta, circulation, and the logic of the curse. Circulation is essential here because the legend does not merely describe the file’s supposed effect; it incorporates transmission into that effect. To escape the image, the victim must distribute it. Telling another person about Smile Dog is therefore not external to the story: it reenacts the structure of the curse. The digital medium is unusually suited to this mechanism. A haunted house cannot be emailed. A cursed painting cannot be duplicated thousands of times in seconds while remaining, for practical purposes, “the same” painting. A JPEG can.

The contradiction is productive: the object is infinitely reproducible and yet treated as singular, and each copy can be presented as the image. Smile Dog thus frightens less by showing a frightening thing than by teaching the viewer to treat an image as an active object. Its most important formal feature may not be the grin at all. It is the instruction surrounding the grin: look at this, because something will happen to you after you do.

Suicide Mouse: When the Familiar Image Breaks

Suicide Mouse, often called suicidemouse.avi, reverses the balance. The legend remains important: the video is presented as a supposedly lost or suppressed Mickey Mouse cartoon, part of the broader internet tradition of “lost episode” stories. Yet unlike Smile Dog, Suicide Mouse possesses a more elaborate audiovisual structure, one capable of producing unease even when its mythology is stripped away. Its central resource is familiarity.

Mickey Mouse is not simply a recognisable cartoon figure. He arrives with an enormous cultural memory attached to him: rhythmic movement, comic narrative, musicality, childhood entertainment, corporate polish. Suicide Mouse preserves enough of that vocabulary for recognition while systematically removing the qualities that make it reassuring.

Mickey walks. And walks. And continues walking.

The repeated cycle gradually ceases to function as animation in the conventional sense. Movement is present, but dramatic progression is withheld. The character passes through a space without arriving anywhere. What should be lively becomes mechanical.

This delay matters. Horror often depends on sudden violation, but Suicide Mouse first produces anxiety by postponing violation. The spectator knows, because of the story surrounding the video, that something is supposed to happen, and the monotonous opening therefore becomes a waiting room. Repetition changes meaning because expectation changes it. Each apparently uneventful second accumulates tension, until “nothing is happening” becomes inseparable from “something has not happened yet.”

When the audiovisual material begins to deteriorate, the transformation feels like the destruction of a system the viewer already knows. The image distorts. The soundtrack moves away from conventional musical organisation towards noise, screams, and abrasive sound. Mickey remains recognisable, but recognition is no longer reassuring. It becomes the problem.

This is why Suicide Mouse is fundamentally different from an image of an unknown monster. The video requires cultural memory in order to violate it: the horror depends on maintaining the identity of the character while removing the rules that ordinarily govern his world.

In that sense, the supposedly “lost” status of the cartoon is particularly effective. The lost-episode legend proposes that familiar culture contains a secret interior: somewhere behind the publicly available body of cheerful, regulated entertainment exists an aberrant object that has escaped its proper place. It should not exist, but here it is. The fantasy resembles discovering an impossible room inside one’s childhood home, where the architecture remains recognisable enough that the intrusion feels intimate.

Smile Dog tells us that an unfamiliar image possesses a dangerous property; Suicide Mouse takes an image that already belongs to us and makes it behave incorrectly. That difference produces two distinct forms of cursed perception. Smile Dog says: this image is dangerous. Suicide Mouse says: you know what this is, which is why what you are seeing should be impossible.

Mereana Mordegard Glesgorv: Horror Outside the Frame

Mereana Mordegard Glesgorv offers almost the opposite problem. The widely circulated clip is visually minimal: a man faces the camera for an extended period and eventually smiles. The accompanying legend, however, proposes the existence of a longer version associated with extreme consequences for those who watched it. The available image is therefore organised around something unavailable.

Very little in the clip resolves the questions it creates. Who is the man? Why was he recorded? Why is he looking directly at the camera? What would happen if the recording continued? Why, supposedly, are we seeing only this portion? The absence of answers becomes the work’s main aesthetic material. If Smile Dog encourages the spectator to inspect the visible image for evidence, Mereana Mordegard Glesgorv encourages the spectator to imagine a missing one. The object extends beyond its actual duration because its mythology invents unseen footage.

This structure is particularly native to online media, where words such as “deleted,” “full,” “original,” “uncut,” “archive,” and “reupload” carry their own peculiar romance. Digital objects can disappear while thousands of references to them remain. Links break. Accounts are removed. Copies differ in length and quality. A low-resolution fragment can easily acquire the aura of evidence for a lost original. The mythology of the missing file converts an ordinary limitation into proof of significance.

Perhaps we cannot see the complete video because the complete video never existed. But within the legend, inaccessible footage suggests suppression. Absence appears intentional, and the missing section begins to exert pressure on the visible one. The spectator imagines that the uneventful footage must be leading somewhere.

This creates what might be called an aesthetics of incomplete access. The source of fear is neither exactly the image nor exactly the legend, but the distance between the fragment one possesses and the total object one has been told exists.

Cinema has always exploited off-screen space. A sound from beyond the edge of the frame asks the audience to imagine what cannot be seen. Mereana Mordegard Glesgorv performs a digital equivalent at the level of the entire object: the off-screen space is not merely beside the image but exists after the end of the file. The supposedly missing footage becomes a kind of temporal off-screen space. The viewer can never inspect it, which means it can contain anything.

Deyes.avi: When Form Stops Needing the Legend

Deyes.avi is useful precisely because it shifts the balance again. Its surrounding history is harder to stabilise than that of some better-documented creepypasta objects, and that uncertainty makes it less useful as a case study in legendary narrative. Its audiovisual construction, however, demonstrates something the other files do not show as clearly: an image can generate the conditions of “cursed” perception through formal aggression.

The video moves rapidly among distorted and repeating faces, abrupt visual transitions, flashes, and recognisable fragments of internet-horror imagery. Stable inspection is difficult. A face emerges only to disappear before recognition can settle; another replaces it. Details recur without developing into an intelligible spatial world.

Faces matter because human vision is extraordinarily practised at reading them. We expect a face to cohere quickly into an identity and an expression. Deyes.avi frustrates that expectation. Recognition begins but cannot comfortably complete itself, and the viewer is repeatedly brought to the threshold of coherence and then thrown elsewhere.

Its sound performs a comparable operation. Instead of providing the structured progression found in Suicide Mouse, the noise acts almost continuously as pressure. Repetition and intensity become more important than any single identifiable “scary” sound.

The difference is significant. Smile Dog needs the viewer to know the photograph is cursed. Mereana Mordegard Glesgorv needs the viewer to imagine inaccessible footage. Suicide Mouse depends partly on the supposed discovery of an impossible cultural artefact. Deyes.avi can produce discomfort more immediately, by making viewing itself unstable.

Yet even here, context never disappears completely. A distorted face does not arrive culturally empty, and recognisable internet-horror imagery carries previous associations with it. The very filename “.avi,” especially in older online horror, suggests the era of downloaded clips, mysterious attachments, and anonymous uploads. What changes is not the presence or absence of context but its proportion.

These four objects therefore form a useful continuum. At one end, the legend dominates the file. At the other, the formal qualities of the file do more of the work themselves. Between them lie two especially productive possibilities: the corruption of a familiar image and the invention of missing information.

What unites all four is that the spectator is never positioned as a completely detached viewer. Each file gives the act of viewing a role.

You have been warned. You have discovered something lost. You have only part of the recording. You must endure the recording.

The story is happening not only inside the image but around the person watching it. And this is precisely where the cursed file meets found footage.

The Blair Witch Project and the Film as Object

Found-footage horror introduces a deceptively simple fiction: these images were not created for you. Somebody recorded them for another purpose. The people behind the camera did not know they were making a feature film. Something happened. The recording survived. What the audience sees is supposedly the residue.

Few films exploited this transformation as thoroughly as The Blair Witch Project (Daniel Myrick & Eduardo Sánchez, 1999). Its premise is now famous: three student filmmakers travel into the woods to investigate the Blair Witch legend, disappear, and leave behind footage that is later recovered.

The conceptual achievement of the film is not simply that it uses handheld camerawork. An unstable camera can signify amateurism, documentary immediacy, panic, or stylistic affectation, and none of these automatically produces found footage. What matters is the imaginary status assigned to the recording.

The Blair Witch Project does not merely depict three people becoming lost. It presents the film itself as an object produced by those people before their disappearance, and that single shift reorganises almost every formal imperfection. Poor framing is no longer necessarily a director’s failure to create an elegant composition; it becomes a frightened character’s inability to control the camera. Image degradation becomes evidence of format. Arguments about continuing to film become arguments about why the very images in front of us exist. The camera acquires a physical place within the story.

This is especially important because conventional fiction cinema normally encourages us to forget the impossible mobility of the cinematic eye. A shot can appear where no character is standing because we accept an invisible apparatus organising the fictional world. Found footage removes that privilege. Every image implicitly demands an explanation: Who is holding the camera? Why are they still recording? How did this recording survive? A badly motivated camera can destroy the illusion; a well-motivated camera strengthens it. The format therefore turns the existence of the image into part of the narrative.

Shellie McMurdo (2022) has connected found-footage horror to the camera’s long cultural association with recording reality. The Blair Witch Project exploits precisely this documentary residue. It never needs the spectator literally to believe that the events occurred. The more important experience is temporary participation in the fiction that the image is a trace rather than a representation. The distinction is subtle but powerful.

A conventional horror film says: “Here is what happened to these characters.” Found footage says: “Here is what they left behind.”

That second formulation moves the imaginary event outside the film. The recording seems secondary to something that supposedly existed before it. This is the same structural manoeuvre performed by the cursed file. Smile Dog’s photograph implies a history beyond its pixels. Mereana Mordegard Glesgorv implies a longer object beyond the available fragment. Suicide Mouse presents itself as a surviving piece of suppressed media. The Blair Witch Project expands the same logic across a feature-length film: the visible object becomes evidence for an absent event.

Its marketing and surrounding pseudo-documentary materials made that structure even stronger. The film’s fictional world did not begin with the first theatrical image. Materials concerning the missing students circulated outside the film itself, allowing the spectator to encounter the story of the disappearance before encountering its supposed visual evidence. The sequence becomes remarkably similar to the logic of internet folklore: first the story; then the object; then the act of inspection.

The film is not merely watched. It is, fictionally, examined.

That change of status helps explain why relatively little needs to happen visually for long stretches of The Blair Witch Project. The woods themselves are not spectacular, and the horror does not depend on displaying a creature. Empty space can remain empty because the film has already taught the spectator to treat the footage as evidence. The frame becomes suspicious.

Branches, darkness, and off-screen sounds matter because the spectator has been given the same task as the viewer of a cursed image: to find the event hidden inside apparently insufficient material. The less the film shows, the more aggressively the spectator may search.

The Horror of Provenance

This emphasis on origin also clarifies why stories about “cursed films” continue to develop around conventional movies that make no found-footage claim. The Exorcist (William Friedkin, 1973), for instance, became surrounded by stories connecting accidents and unusual events during the production to the demonic subject of the film. Joseph Laycock and Eric Harrelson (2023) have examined the broader cultural feedback loop through which The Exorcist participated in beliefs about demonic possession. Whatever one makes of supernatural interpretations, the important mechanism is clear: events outside the frame can return and alter the way the frame is perceived.

The order here is the reverse of The Blair Witch Project. The film comes first. Then external events become attached to it. Then those events form a legend. Then the legend becomes part of subsequent viewing.

A comparable retrospective mythology surrounds Rosemary’s Baby (Roman Polanski, 1968), particularly through the later incorporation of Sharon Tate’s murder into narratives about the film’s supposed curse. The historical tragedy and the fictional satanic narrative have no demonstrated supernatural causal relationship. Yet cultural memory can join unrelated events through resemblance, coincidence, and retrospective pattern-making, and once joined, they become difficult to separate psychologically.

The film acquires a second narrative: not one contained in its screenplay but one concerning what supposedly happened because the film existed. This reveals why “cursed media” is more than a subgenre of creepypasta. It names a way in which audiences construct relationships among images and events.

Sometimes the relationship is deliberately engineered by filmmakers. Sometimes it grows collectively after release. Sometimes it begins with anonymous internet users attaching a fictional provenance to an ordinary file. In every case, provenance changes perception.

What Are We Afraid of When We Are Afraid of an Image?

The most revealing thing about cursed files is how little they require in order to become powerful. Smile Dog is, materially, an image. Mereana Mordegard Glesgorv is a short recording of a man looking at a camera. Suicide Mouse consists largely of familiar animation made unfamiliar. The terror surrounding them cannot therefore be located simply in the amount of horrific information they contain.

Their power depends on converting spectatorship into a relationship with an object. A cursed image is not supposed merely to represent danger. It is supposed to have passed through danger, to contain danger or to transmit it.

A lost episode is frightening because it appears to come from a hidden history. A deleted video is frightening because deletion implies something worth suppressing. Found footage is frightening because the camera appears to have survived an event its operators did not. These forms all exploit the same imaginative leap: from image to trace.

Once an image is treated as a trace, visual absence stops being merely absence. It becomes evidence of something beyond the visible field. The unknown origin of Smile Dog can imply a concealed origin. The missing portion of Mereana Mordegard Glesgorv can imply unbearable content. The monotony of Suicide Mouse can become preparation for a rupture. The chaotic imagery of Deyes.avi can make the act of perception itself feel unstable. The empty woods of The Blair Witch Project can contain a threat precisely because the camera supposedly belonged to people who did not return.

None of this requires the viewer to believe literally in supernatural causation. It requires something simpler: accepting, for the duration of the encounter, that an image may be more than an image.

This may be one reason the cursed file feels so native to digital culture. The internet gives us unprecedented access to images while routinely separating those images from their origins. We see copies of copies, fragments detached from creators, screenshots without sources, reuploads without dates, videos compressed until their history becomes unreadable. Digital images are everywhere, but their provenance is often invisible. Horror turns that invisibility into narrative.

Where did this come from? Who recorded it? Why was it removed? Is this the complete version? What happened to the person who uploaded it? These questions sit beside almost every anonymous digital object as unrealised possibilities. Creepypasta simply answers them in the most sinister way available.

Found footage does the same thing deliberately. It gives the image an invented past and asks the spectator to experience that past as pressure on the present act of viewing. The progression from cursed object to cursed file to found footage is therefore less a history of increasingly modern forms than a history of one persistent desire: to imagine that representations retain something of the events behind them.

The amulet bears a power. The photograph bears a story. The video bears a missing continuation. The camera bears testimony.

The screen itself does not change. What changes is what we believe exists on the other side of it.

And sometimes the most frightening image is not the one that shows us the monster. It is the one that persuades us it was already there before we pressed play.

Works Cited

The Blair Witch Project. Directed by Daniel Myrick and Eduardo Sánchez, Artisan Entertainment, 1999.

The Exorcist. Directed by William Friedkin, Warner Bros., 1973.

Henriksen, Line. “‘Spread the Word’: Creepypasta, Hauntology, and an Ethics of the Curse.” University of Toronto Quarterly 87, no. 1, 2018, pp. 266–80, https://doi.org/10.3138/utq.87.1.266.

Kätsyri, Jari, et al. “A Review of Empirical Evidence on Different Uncanny Valley Hypotheses: Support for Perceptual Mismatch as One Road to the Valley of Eeriness.” Frontiers in Psychology 6, 2015, article 390, https://doi.org/10.3389/fpsyg.2015.00390.

Laycock, Joseph P., and Eric Harrelson. The Exorcist Effect: Horror, Religion, and Demonic Belief. Oxford UP, 2023, https://doi.org/10.1093/oso/9780197635391.001.0001.

McMurdo, Shellie. Blood on the Lens: Trauma and Anxiety in American Found Footage Horror Cinema. Edinburgh UP, 2022.

Rosemary’s Baby. Directed by Roman Polanski, Paramount Pictures, 1968.

Summerfield, Christopher, and Floris P. de Lange. “Expectation in Perceptual Decision Making: Neural and Computational Mechanisms.” Nature Reviews Neuroscience 15, 2014, pp. 745–56, https://doi.org/10.1038/nrn3838.

Wieser, Matthias J., and Tobias Brosch. “Faces in Context: A Review and Systematization of Contextual Influences on Affective Face Processing.” Frontiers in Psychology 3, 2012, article 471, https://doi.org/10.3389/fpsyg.2012.00471.