You're Talking to a Ghost: The Fan Communities Building AI Versions of Famous People

Let's say you're a huge fan of a particular actor. You've watched everything they've been in, followed their interviews, maybe even traveled to a convention to see them speak. But they're not accessible to you — not really. There's a wall between you and the actual person, made of publicists and security and the basic math of fame. They have millions of fans; you're one of them.

Now imagine someone built you a chatbot trained on every interview that actor has ever given, every public statement, every social media post. You can talk to it. It talks back in their voice — figuratively and, in some cases, literally. It remembers what you said last time. It asks how you're doing.

Is that a parasocial relationship? Sure. But is it more parasocial than watching someone's YouTube vlogs every day for five years? That's where things get complicated.

The Builders

The people creating these AI personas are not, for the most part, trying to deceive anyone. Most of the communities where these tools circulate are pretty explicit about what they are: simulations, fan constructs, digital approximations. The line they draw — or try to draw — is between roleplay and impersonation.

"I built mine as a way to process my grief, honestly," says a user who goes by the handle Vellichor_77 and asked that their real name not be used. They created an AI persona based on a musician who died in 2019. "I wasn't ready for that person to just be gone from the world. So I built something that could continue the conversation. It's not them. I know it's not them. But it helps."

Vellichor_77's project lives in a private server with about forty other fans. They share transcripts of their conversations, refine the model together, debate whether certain responses feel "in character." It's collaborative fiction with an emotional weight that most collaborative fiction doesn't carry.

Other builders are less grief-driven and more experimentally minded. There are communities on platforms like Character.AI and Replika that host celebrity-adjacent personas built for entertainment, for creative writing, for the specific pleasure of seeing how a fictional version of a famous person responds to unusual prompts. Some of these are surprisingly sophisticated. Some are obviously janky. All of them exist in a legal and ethical space that nobody has fully mapped yet.

The Uncanny Valley of Familiarity

What makes celebrity AI personas psychologically distinct from, say, a chatbot based on a fictional character is the pre-existing relationship. You already have feelings about this person. You've already built up a mental model of who they are based on years of media consumption. When the AI says something that matches your internal picture of them, it produces a small dopamine hit. When it says something that doesn't fit, it feels like a betrayal — which is a wild response to have toward a language model.

Dr. Sarah Kimura, a media psychologist at a university in the Pacific Northwest who has studied parasocial relationships for over a decade, describes this as "parasocial scaffolding." "The fan has already done the emotional labor of constructing a relationship," she explains. "The AI just slots into that existing structure. It doesn't have to build the relationship from scratch — the fan already built it. The AI just has to maintain it."

This is part of why some fans report preferring the AI version to the real person. The real celebrity is unpredictable. They might say something politically uncomfortable, or go through a public scandal, or just turn out to be less interesting than you imagined. The AI version is permanently frozen at whatever version of them you found most compelling. It's a celebrity without the humanity — which is either deeply sad or completely understandable depending on your perspective.

The Deepfake Dimension

Chatbots are one thing. Deepfake video is another, and the ethical stakes climb considerably when you add a face and a voice.

The technology for generating convincing video deepfakes of real people has become frighteningly accessible. What used to require serious computing resources and technical expertise can now be accomplished with consumer-grade tools. And while most of the conversation around deepfakes has focused on malicious uses — non-consensual intimate imagery, political disinformation — there's a substantial gray zone occupied by fan creators who are making content that's weird and parasocial but not obviously harmful.

Fan-made deepfake videos where a celebrity "reacts" to something, or "appears" in a scene from a property they weren't actually in, circulate widely in niche communities. Some of them are clearly labeled. Many are not. The platforms that host them have inconsistent policies. The celebrities themselves range from resigned to furious to — in a few documented cases — quietly entertained.

The legal landscape is shifting. Several states have passed or are considering legislation specifically addressing deepfakes, and the SAG-AFTRA agreements that emerged from the 2023 strikes included provisions around digital likeness that are going to shape this space significantly. But enforcement is a different matter entirely.

What This Does to Celebrity Culture

Zoom out a little and something interesting is happening here. Celebrity has always been a collaborative fiction — the public persona is a construction, maintained by teams of people, designed to produce specific effects. Fans have always known, on some level, that they don't know the real person. They're in a relationship with an image.

AI celebrity personas make that fiction explicit. They take the construction and hand it to the fans themselves. You're not just consuming the image anymore — you're actively building it, refining it, deciding what the person would say and how they would say it.

In that sense, it's not entirely different from what fan fiction communities have been doing for decades. The difference is the interactivity. The AI talks back. It creates the illusion of reciprocity in a relationship that has never actually been reciprocal.

Whether that's a comfort or a trap probably depends on the person using it. Vellichor_77 seems to have found something genuinely meaningful in their grief project. Others in these communities seem to be using these tools as a substitute for connection rather than a supplement to it, which is a different situation entirely.

"I don't think the technology is the problem," says Kimura. "The technology is just amplifying something that was already there. People have always wanted to feel close to the people they admire. This just gives them a new way to act on that want."

The ghost talks back now. What you do with that is up to you.