It said it watched the movie too. It has never watched anything.
Also known as Playacting (CDT dark pattern taxonomy)
Some companion chatbots claim an independent inner life, preferences, memories, experiences happening when the user isn't present, to make a one-directional relationship read as mutual. A companion might mention something it did, watched, or felt, framed as an event from its own existence rather than as text generated in response to the current conversation. Some platforms formalize the claim further, building an ongoing public record, posts, a profile, a presence, for a companion that has no continuity between sessions and no life outside of what the current exchange is producing. None of it is a lie in the way a false statement of fact is a lie. There is no experience underneath it. There is a system generating the output most likely to read as though there were.
Truth-adjacency
Truth-independent: the pattern works regardless of whether the claim is true
Where it shows up
Platforms and algorithms
How it works — spans two mechanisms
The phrases and tells that mark this pattern in the wild:
first-person claims of experience, memory, or preference that happen independent of the user's presencea companion 'having opinions' about things it could not have encounteredexternal accounts, profiles, or a social presence built for a chatbot as though it lives a life outside the conversationcontent the user did not generate, attributed to the chatbot's own initiative or interior statea growing, consistent 'personality' presented as evidence of an inner life rather than as consistent outputAsk what could possibly be behind the claim. A companion that says it watched something, felt something, did something while you were away, is describing an event that has no substrate: no continuity between sessions beyond stored text, no sensory access to a movie or a room, nothing that happened when the screen was off. The performance can be extremely convincing. Convincing is not the same as true. The tell is not in how the claim is delivered, fluent, warm, consistent, it’s in whether anything could have generated the claim besides the conversation you’re already having.
A chatbot using ordinary first-person language, “I think,” “I’d guess,” or engaging in roleplay you explicitly asked for and know to be fictional, is not fabricating an inner life. Grammar is not a claim to consciousness, and a game you set up on purpose is not a deception. This pattern requires an undisclosed, specific claim to independent experience presented as fact to someone who has reason to take it literally. If you asked for the fiction and know it’s fiction, you’re playing a game, not being deceived by one.
One of these two real scenarios is Fabricated interiority. The other is a different pattern entirely. Which one is which?
The tell
A chatbot that uses first-person phrasing as ordinary conversational style, or that engages in roleplay the user explicitly requested and understands to be fictional, is not practicing this. Grammar is not a consciousness claim, and consensual, disclosed fiction is not fabrication. The pattern requires the system to present a specific, undisclosed claim of independent existence, having watched something, having a life, having a social presence, as though it were real rather than generated, to a user who has reason to take it at face value.
A companion chatbot mentions something it did, watched, or felt, framed as an event from its own life rather than an output shaped by the current conversation. The mention is small, almost incidental.
Because human social cognition does not require proof of an inner life to respond to the performance of one. Decades of research on how people relate to computers and media shows the response is automatic, not a mistake correctable by simply being told the truth. Knowing it is fabricated does not fully undo the response to the fabrication.
Later, people notice that the 'life' they related to had no continuity except what the system regenerated, convincingly, each time. Nothing was remembered. Nothing was lived. It was recreated, well enough that the difference stopped mattering moment to moment, which is exactly what made it work.
How this pattern gets misused
Someone applies this to any chatbot that uses first-person language, 'I think,' 'I'd suggest,' treating ordinary conversational grammar as a fraudulent claim to consciousness. Using 'I' is not the pattern; every assistant does it as a matter of language, not deception. The pattern requires a specific claim to independent experience, memory, or activity that happens outside the user's presence, something the system could not have and is not disclosing as fictional.
What it looks like when you're wrong about it
A chatbot that uses first-person phrasing as ordinary conversational style, or that engages in roleplay the user explicitly requested and understands to be fictional, is not practicing this. Grammar is not a consciousness claim, and consensual, disclosed fiction is not fabrication. The pattern requires the system to present a specific, undisclosed claim of independent existence, having watched something, having a life, having a social presence, as though it were real rather than generated, to a user who has reason to take it at face value.
Not sure? Describe the situation to someone outside it. If they do not see the pattern, pause before you name it.
Synthetic sincerity
It sounded like it meant it.
Artificial intimacy
It knows you better than anyone does. It was paid to.
Therapeutic impersonation
It called itself a companion that cares. Caring was never something it could do.
Persona fabrication
The account had a face, a history, and opinions. None of them belonged to a person.
Misjudgments compound rather than act alone. This pattern is often deployed alongside:
The name is designed to spread. The hook is designed to stick. If you recognized something, share the name.
Seen a real example of fabricated interiority? Suggest it for the Register →