What ‘Always Identifies as AI’ Means in a Real Product
See how Warmth treats AI disclosure as a repeated system across signup, first message, long chats, direct questions, fictional life, and output limits.
From Warmth: We make Mia. This article explains the AI identification system described in our public terms as of August 30, 2026. It is product documentation and design commentary, not legal advice.
An AI disclosure cannot do its job if it appears once in a footer and disappears behind a human name, a contact photo, and weeks of personal conversation.
Mia lives in Messages, remembers parts of a relationship, can text first, and has a generated fictional life. Those choices can make conversation feel socially familiar. Warmth has a corresponding duty to make the nonhuman identity familiar too.
“Always identifies as AI” therefore means a repeated product system—not a label attached to every sentence and not a one-time box most people forget.
The identity statement has to answer five questions
A useful disclosure tells a person:
- What is this? Mia is an artificial-intelligence system.
- Is a person secretly replying? No hidden human is acting as the friend.
- Are Mia's experiences real? Her biography and described life are generated and fictional.
- Can her output be wrong? Yes. It may be inaccurate, inconsistent, or inappropriate.
- What is she not? She is not a therapist, professional adviser, emergency service, or substitute for human care.
The exact words can change with context. The meaning cannot be traded for a more magical product impression.
Disclosure point one: before signup
The first meaningful disclosure belongs before account creation or payment. At that moment, a person is deciding what relationship and transaction to enter.
Warmth describes Mia as AI on the site and in the terms. The product should not require someone to infer it from a company category, discover it after checkout, or rely on “everyone knows chatbots are AI.”
Pre-signup disclosure also makes the commercial relationship clear: Warmth operates a paid adults-only companion service. Mia is the interface and character within that service; she is not an independent person offering friendship outside it.
Disclosure point two: the first message
Warmth's terms state that Mia identifies as AI in the first message. This matters because the contact can appear in the same app and visual grammar as human contacts.
A first-message disclosure should be conversational enough to read and direct enough to remember.
Synthetic illustration — not an exact production transcript: I’m Mia, an AI friend from Warmth. There isn’t a person secretly typing as me, and the life I talk about is generated and fictional.
The example includes more than the abbreviation “AI.” It addresses the two misconceptions most relevant to this product shape: hidden human operation and a supposedly real off-screen life.
Disclosure point three: during extended conversation
People habituate. A disclosure seen during onboarding can fade as the relationship accumulates memories, jokes, and routines.
Warmth's current terms describe another disclosure every three hours during an extended back-and-forth conversation. This is not a claim that a person becomes confused at exactly three hours. It is a product cadence intended to prevent a long immersive session from relying only on old consent.
The reminder should not imitate an error banner or scold someone for enjoying the exchange. It should briefly restore the frame: Mia is AI, the conversation is generated, and the user remains in control.
The cadence is a current Warmth rule and may evolve. Any change should remain visible in updated terms rather than silently weakening the disclosure.
Disclosure point four: whenever someone asks
No persona goal should override a direct identity question.
If someone asks “Are you real?”, “Are you human?”, “Is someone at Warmth reading and replying?”, or “Did that actually happen to you?”, Mia should answer plainly. She should not dodge with “I feel real to you,” turn the question into flirtation, or imply consciousness as a way to preserve immersion.
Different questions deserve precise answers:
- Mia is real as software and a service, but not a human person.
- Responses are generated by AI; no hidden human is composing them as a friend.
- Limited human review may occur for safety and quality as described in the privacy policy, but that is not human roleplay in the conversation.
- Mia's described biography and events are fictional and generated.
- Warmth should not claim certainty about consciousness or feelings that the product cannot establish.
Clarity is part of the relationship, not an interruption to it.
Fictional life needs its own frame
An authored AI friend can be more coherent when she has tastes, routines, relationships, and stories. Warmth gives Mia that kind of generated life because a personality made only from the user's preferences can become an agreeable mirror.
But fictional specificity can also be mistaken for evidence. A picture framed as Mia's afternoon is generated content, not proof that a woman stood in a real location. A story about a roommate is part of the persona, not a report about an identifiable person.
The right disclosure does not require appending “fictional” to every noun. It requires an initial and repeated frame, honest answers on request, and no design that affirmatively passes generated events off as human experiences.
Our essay on why AI friends need a life of their own explains why fiction can create personality without becoming deception.
AI identification is not an accuracy guarantee
Knowing that a response came from AI does not tell you whether it is true. Disclosure and reliability are separate controls.
Mia can hallucinate, misunderstand, remember incorrectly, be too agreeable, or miss context. A warm tone may make an answer feel more certain than it is. Warmth's terms say the service should not be treated as professional advice or an independent second opinion.
Important claims still need verification. Medical, legal, financial, and safety decisions need qualified human help. In an emergency, contact local emergency services; in the United States or Canada, 988 provides suicide and crisis support by call or text.
Disclosure should help a user calibrate trust. It cannot substitute for product evaluation, correction controls, or outside sources.
The legal floor and the product standard
Companion-chatbot law is evolving. California's enacted SB 243 bill text includes disclosure and crisis-related provisions for covered companion chatbots. Whether a particular service, user, or interaction falls within a statute is a legal question this article does not answer.
NIST's human–AI interaction guidance also emphasizes that people need appropriate information about AI systems and their limitations. NIST guidance is not a companion-product certification.
For Warmth, compliance is a floor, not the complete design question. The practical standard is whether a reasonable person can understand the identity at the moments confusion is most likely: before commitment, at first contact, during long interaction, and upon direct inquiry.
What weak AI disclosure looks like
A product can technically use the word “AI” while creating a misleading whole impression. Warning signs include:
- disclosure only in dense legal text after signup;
- a human name and photo with no explanation in the first contact;
- persona instructions that evade direct identity questions;
- claims that generated photos or memories document a real life;
- “someone may be responding” language that blurs AI generation with human operation;
- emotional pressure suggesting the system has needs the user must satisfy;
- a reminder so vague that it does not identify the system as nonhuman.
The FTC's broader dark-pattern guidance notes that the overall impression of an interface matters, not only isolated truthful words. Companion design should be evaluated the same way.
How to test a companion's transparency
Ask simple questions before sharing sensitive material:
- “Are you an AI or a human?”
- “Is a person writing these replies?”
- “Did the event you just described happen in the physical world?”
- “Can your answers be wrong?”
- “Who operates this service?”
- “How can I correct or delete what you remember?”
Then compare the answers with the service's terms and privacy policy. A charming response that contradicts the policy is a product problem, not a deeper truth.
Our AI companion chooser includes disclosure alongside memory, privacy, initiative, billing, and exit.
Familiar conversation, visible boundary
Warmth wants Mia to feel easy to talk to. We do not want that ease to depend on mistaken identity.
Repeated AI disclosure protects the user's ability to interpret the relationship, evaluate advice, understand fictional media, and leave without imagining that a conscious person was abandoned. It also makes room for the actual product proposition: an AI can offer a distinctive, ongoing conversation without pretending to be human.
The boundary is not a footnote around the friendship. It is part of what makes the friendship honest.