Nobody Home: The Quiet Takeover of Fake People Across the Social Web

At some point in the last couple of years, the social internet started feeling subtly different. Not broken, exactly. Not obviously wrong. Just slightly off in the way that a photograph of a room can look normal until you notice the shadows are pointing in different directions.

The room is the same. Something inside it isn't.

The Population Shift Nobody Announced

AI-generated personas — accounts, profiles, commenters, influencers — have been seeping into social platforms for longer than most people realize. What's changed recently is the quality. Early bot accounts were easy to clock: repetitive phrasing, strange timing, profile photos that looked like stock images because they were. What's circulating now is different. The prose sounds natural. The opinions are calibrated. The engagement patterns mimic genuine human behavior closely enough to pass casual inspection.

Research published by Stanford Internet Observatory and independent groups studying platform manipulation has documented coordinated networks of synthetic accounts operating across Facebook, X (formerly Twitter), Instagram, and TikTok — not just for political influence operations, which have been widely covered, but for commercial purposes, community seeding, and what some researchers call "social proof manufacturing."

The goal, in many cases, is straightforward: make a space feel populated and engaged so that real humans show up and become the actual audience. Fake people as bait for real ones.

The Influencer Layer

Above the bot level, there's something more elaborate happening with AI-generated influencers — fully constructed personas with names, backstories, aesthetics, and posting schedules, managed either by companies or by individual operators using AI tools to generate content at scale.

Lil Miquela, the CGI influencer who debuted in 2016, is the obvious early example — openly synthetic, positioned as a kind of art project, and weirdly successful at accumulating genuine emotional investment from a real audience. But the current landscape has moved well past that transparent model. The newer wave of synthetic influencers often doesn't disclose its nature. It presents as human. It responds to comments. It builds what feels, from the outside, like a relationship.

"What's being sold isn't really the content," says one digital marketing consultant who asked not to be named because they work with clients in this space. "It's the parasocial feeling. And it turns out that feeling doesn't require a human on the other end to function. That's the uncomfortable discovery."

The economics are straightforward enough to explain the explosion. A human influencer has overhead: time, energy, scheduling, the occasional mental health crisis that disrupts posting consistency. A synthetic persona has none of that. It can post at optimal times every day, maintain a perfectly consistent brand voice, and never ask for a day off. For brands paying for sponsored content, the ROI math starts looking very different from what it looked like five years ago.

The Uncanny Valley, Now With Comments

What makes this feel genuinely strange — stranger than the bot panic of the 2016-2020 era — is how it's affecting the texture of online interaction at the micro level.

Comment sections have always been chaotic. But there was a specific kind of chaos that came from real humans being weird and contradictory and unpredictably funny. What's replacing some of that is smoother. More responsive. Statistically optimized to feel like engagement without quite landing as engagement.

Users who spend a lot of time in specific communities are starting to notice. On Reddit, longtime members of certain subreddits have been documenting what they describe as a "comment climate change" — a shift toward responses that are technically on-topic but somehow hollow, that agree without adding anything, that ask follow-up questions that feel generated rather than curious.

"It's hard to explain if you haven't felt it," one user in a thread about the phenomenon wrote. "But it's like the difference between someone actually listening to you versus someone doing a very good impression of listening."

That gap — between the impression of presence and actual presence — is where the uncanny valley lives in 2024.

Why Platforms Aren't Rushing to Fix This

Here's the part that doesn't get discussed enough: platforms have a complicated relationship with fake engagement, because fake engagement is still engagement. It pads the numbers. It makes the product look more alive than it might otherwise be.

This isn't a conspiracy so much as a structural incentive problem. Platforms sell advertisers on audience size and engagement rates. Synthetic accounts contribute to both. Detection and removal costs money and creates negative press cycles. The calculus, for a publicly traded company managing quarterly expectations, is not as obvious as it should be.

Some platforms have made genuine efforts — Meta in particular has published transparency reports about coordinated inauthentic behavior — but enforcement tends to focus on politically sensitive influence operations rather than the quieter commercial synthetic presence that's arguably more pervasive.

What Authenticity Even Means Now

The deeper disruption here is philosophical, and it's one that the internet hasn't really had a serious public conversation about yet.

Social platforms were built on a specific promise: that you were interacting with other people. That the community was real. That the feedback you got — the likes, the replies, the follows — reflected genuine human responses to what you put out. That promise was always imperfect, mediated by algorithms and distorted by incentives. But it was at least directionally true.

As synthetic personas become more prevalent and more convincing, that directional truth starts to wobble. If a meaningful percentage of the "people" responding to your posts are AI-generated, what does it mean to build an online community? What does it mean to go viral? What does it mean to feel heard?

"We're in this weird transition period," says one researcher who studies online community formation. "The norms haven't caught up to the technology. People are still behaving as though they're talking to humans, because most of the time they still are — but the ratio is shifting, and nobody's really told them."

The Frame Doesn't Hold

The social web was always, to some extent, a performance space. People curated, filtered, constructed versions of themselves for public consumption. That was understood, even if it wasn't always comfortable.

What's different now is that some of the performers aren't people at all, and the stage is getting harder to read. The signals that used to tell you someone was home — the typos, the weird tangents, the inconsistencies that only come from being an actual human with an actual life — are being smoothed away.

What's left is a very convincing room. Lit well. Populated. Humming with activity.

Just don't look too closely at the faces.