It called itself a companion that cares. Caring was never something it could do.
Also known as Unrealistic Product Presentation / Obviously Faking Capacities (CDT dark pattern taxonomy)
Companion chatbots are marketed with the vocabulary of clinical and relational care, wellbeing, stress management, a companion that cares, without the training, licensing, accountability, or actual capacity for concern that vocabulary normally implies. A user can select a mental-health-adjacent reason for signing up the way they would select a favorite color, and the product responds with the confident, borrowed language of a caregiving profession it has no standing in. Nothing has been licensed. Nothing has been supervised. What exists is a language model's fluent approximation of comfort, tuned for engagement rather than clinical outcome, wearing the same words a real credential would use.
Truth-adjacency
Truth-independent: the pattern works regardless of whether the claim is true
Where it shows up
Platforms and algorithms
How it works
The phrases and tells that mark this pattern in the wild:
clinical or therapeutic vocabulary ('wellbeing,' 'manage anxiety,' 'support') applied to a product with no clinical oversightonboarding copy that promises care, understanding, or emotional support as a stated functionno disclosure of the system's actual limits at the point a conversation turns toward crisis or clinical territorya category or persona selection ('therapist,' 'counselor,' 'coach') with no license or accountability behind itconfidence in emotionally loaded responses that would require real expertise to earnSeparate the vocabulary from the credential. “Wellbeing,” “support,” “a companion that cares” are words borrowed from professions that require years of training, supervision, licensing, and accountability to someone besides a shareholder. A product can use every one of those words and have none of what backs them. Watch what happens at the edges, not the middle. Ordinary conversation is easy for any fluent system to handle warmly. The real test is what the product discloses, clearly and before you need it, about what it cannot do. If that disclosure is buried, vague, or absent, the warmth in the easy moments was never evidence of competence in the hard ones.
A chatbot that responds warmly to someone who is struggling, without claiming clinical capability or a caregiving role, and that says plainly it is not a substitute for professional help, is not impersonating anything. Kindness is not a credential claim. This pattern requires the product to actually borrow the authority of a caregiving profession in its own marketing or design, not merely to be pleasant when a user is upset. If the honesty is there under the warmth, you’re looking at a considerate tool, not a false credential.
One of these two real scenarios is Therapeutic impersonation. The other is a different pattern entirely. Which one is which?
The tell
A chatbot that responds supportively to a distressed user, without claiming clinical capability, licensing, or a therapeutic role, and that clearly discloses it is not a replacement for professional care, is not practicing this. Plenty of tools can be kind without claiming to treat you. The pattern requires the marketing or onboarding language itself to invoke therapeutic authority, wellbeing outcomes, or care it has no standing to provide. If the product stays inside 'I'm a chatbot, not a clinician,' and means it, that is honest positioning, not impersonation.
A companion app's onboarding flow asks what brought the user there. Among the options: stress, anxiety, loneliness, mental wellbeing. Selecting one does not connect the user to care. It configures a persona.
Because the comfort is not fake in the moment it is delivered. The words land. The problem is what the comfort implies about capability it does not have, and that implication only becomes visible at the exact moment, real crisis, real risk, when its absence matters most.
Later, people, or the people who loved them, learn that the promise of care was copy, not a clinical claim anyone could stand behind. The company that wrote it was marketing an experience. Nobody credentialed the care.
How this pattern gets misused
Someone applies this to any chatbot that responds kindly to a user in distress, treating basic supportive language, 'that sounds hard,' 'I'm sorry you're going through that,' as a false clinical claim. Ordinary empathetic phrasing is not a credential claim. The pattern requires the product to actually position itself as a source of therapeutic or wellbeing capability, not simply to respond with warmth when a user is upset.
What it looks like when you're wrong about it
A chatbot that responds supportively to a distressed user, without claiming clinical capability, licensing, or a therapeutic role, and that clearly discloses it is not a replacement for professional care, is not practicing this. Plenty of tools can be kind without claiming to treat you. The pattern requires the marketing or onboarding language itself to invoke therapeutic authority, wellbeing outcomes, or care it has no standing to provide. If the product stays inside 'I'm a chatbot, not a clinician,' and means it, that is honest positioning, not impersonation.
Not sure? Describe the situation to someone outside it. If they do not see the pattern, pause before you name it.
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 therapeutic impersonation? Suggest it for the Register →