Ghost in the Machine: The Invisible Censors Deciding Which Darkness You're Allowed to Experience

You've never met the entity that decides whether your horror content gets seen. It doesn't have a name, a face, or a conscience. It's a content moderation algorithm, and it's quietly reshaping the entire horror creative landscape — one shadowban at a time.

We went looking for the creators on the other side of the kill switch.

The Invisible Hand on the Throat of Dark Art

Let's start with something concrete. Marcus Webb — not his real name, he asked us to change it because he's still active on the platforms in question — runs a horror fiction newsletter and maintains a following across Instagram, TikTok, and YouTube. He writes dark psychological horror, the kind that sits in the space between literary fiction and genuine nightmare fuel. Nothing gratuitous. No explicit gore. Lots of existential dread, moral ambiguity, characters doing terrible things for understandable reasons.

About eighteen months ago, his reach fell off a cliff.

"It wasn't a ban. It wasn't a strike. Nothing happened that I could point to," he told us. "My posts just... stopped reaching people. Content that used to get a few thousand views was getting two hundred. I'd post something and my own followers weren't seeing it. I asked around and other horror creators were saying the same thing."

What Marcus experienced has a name, even if the platforms themselves rarely use it: shadowbanning. Or, in the more clinical language the tech companies prefer, "reduced distribution" or "visibility filtering." The content stays up. The account stays active. But the algorithmic amplification that makes any content discoverable gets quietly throttled.

The question is: why? And who — or what — is making that call?

Inside the Black Box

Content moderation at scale is genuinely hard. That's worth acknowledging before we get into the criticism, because the alternative — completely unmoderated platforms — has its own catastrophic failure modes that we've seen play out in real time. The platforms aren't wrong to want systems that catch harmful content before it spreads.

But the systems they've built are blunt instruments operating in a space that requires surgical precision.

Most major platforms now rely on layered moderation systems: automated AI scanning happens first, flagging content based on keyword detection, image recognition, and behavioral pattern matching. Human reviewers handle escalated cases, but at platforms with hundreds of millions of posts per day, the vast majority of moderation decisions are made by machine learning models that nobody outside the company fully understands — including, often, the engineers who built them.

These systems are trained on labeled data. Content that was previously removed gets used to teach the model what to flag. Which means the biases, errors, and contextual failures of past moderation decisions get baked into future ones, compounding over time.

For horror creators, this creates a specific and maddening problem: the vocabulary of horror overlaps substantially with the vocabulary of content that platforms genuinely want to suppress. Violence. Death. Psychological manipulation. Themes of self-harm. Predatory behavior. Darkness, in all its forms. A horror short story exploring suicidal ideation as a literary theme looks, to a keyword-matching algorithm, almost identical to content that's actually promoting self-harm. The machine doesn't understand the difference between depicting something and endorsing it.

The Creators Caught in the Crossfire

Sarah Chen runs a Patreon-supported horror fiction community and has been creating dark content online for nearly a decade. She's watched the moderation environment tighten in real time.

"There are words I can't use anymore. Not because I got told I can't use them — I just know from experience that certain terminology will tank a post's reach or get it demonetized immediately," she said. "So I've started writing around things. Using euphemisms. Implying rather than stating. And honestly? Sometimes that makes the work better. But sometimes it makes it worse. Sometimes the horror is in the directness, and when you sand that off, you lose something real."

That process of self-censorship — anticipating algorithmic punishment and preemptively softening your work — is something almost every horror creator we talked to described. It's a form of creative pressure that doesn't show up in any platform's official policy documentation, because it's not a policy. It's a behavioral adaptation to an opaque system.

Jordan Pruitt, who produces a horror fiction podcast and has a substantial YouTube presence, puts it bluntly: "The algorithm has its own aesthetic. It prefers certain kinds of content. Horror that's safe enough to advertise against, dark enough to feel edgy but not dark enough to actually disturb anyone. And if you want reach, you learn to make that kind of horror. Which is fine, I guess. But it's not the horror I want to make."

The Asymmetry of Enforcement

Here's where it gets genuinely complicated: the moderation isn't applied evenly, and that unevenness follows some predictable patterns.

Independent creators — the ones without industry relationships, legal teams, or platform partnerships — report significantly more moderation friction than established studios and production companies producing similar content. A major streaming platform can post a trailer for a film depicting graphic violence and it gets amplified as premium content. An independent horror filmmaker posting a clip from their self-funded short film with similar imagery gets flagged.

The platforms would argue this reflects trust and verification systems — established partners have agreed to stricter content guidelines and have accountability structures in place. Critics would argue it reflects something simpler: money. Advertising revenue flows from premium content partnerships, and that revenue creates a kind of implicit immunity.

What this means in practice is that the horror getting amplified — the horror that the algorithm rewards — tends to be horror that's already been through a commercial development process that smoothed off its edges. The weird, the transgressive, the genuinely challenging stuff gets pushed to the margins.

What's Actually Getting Suppressed

The content that gets hit hardest tends to cluster around a few categories: horror that explores mental illness with any degree of complexity, horror with LGBTQ+ themes (particularly anything involving queer identity and horror tropes), horror that deals with race and violence in ways that feel politically charged, and horror that depicts real-world extremism or ideological violence as a subject of artistic examination rather than celebration.

The irony is that some of the most important horror ever made falls squarely into those categories. Get Out is fundamentally about racial violence. The Babadook became a queer icon precisely because of how it handled grief and repressed identity. Hereditary depicts mental illness and grief in ways that would almost certainly trigger automated flags if it were being distributed as independent digital content rather than through A24's distribution infrastructure.

Is There a Way Out?

Some creators have moved toward subscription-based models — Patreon, Substack, direct newsletter distribution — that bypass platform algorithms entirely. It's a real solution, but it comes with a real cost: discoverability. The algorithm, for all its flaws, does surface content to new audiences. Without it, you're dependent on your existing community to spread your work, which creates its own ceiling.

Others are advocating for more transparency from platforms about how their moderation systems work — not full disclosure of proprietary models, but clearer documentation of what gets flagged, why, and what the appeals process looks like. Right now, most creators are navigating a system whose rules they can only infer from its outputs.

What nobody seems to have a clean answer for is the deeper question: in a media environment where algorithmic amplification determines what gets seen, and where those algorithms are trained on commercial and political pressures that have nothing to do with artistic value — what happens to the horror that most needs to exist?

The darkness that challenges us, disturbs us, forces us to look at things we'd rather not look at — that's the darkness that has always mattered most. And right now, it's the darkness that's most at risk of being quietly, efficiently, invisibly switched off.

Something to think about the next time a piece of horror content doesn't show up in your feed.