TL;DR: This glossary covers the ten terms that come up most often when comparing AI companion apps — from persona memory and character cards to content filters and token-based pricing. Each entry has a short definition plus one line of our own editorial context on why it matters when you're evaluating an app.
- Persona Memory
- The system that stores facts about you and your character across sessions so the app can recall them in later conversations, rather than starting over each time.
- Roleplay / Story Mode
- A chat mode built around ongoing narrative and scene-setting rather than direct back-and-forth Q&A, often supporting scene descriptions, actions in asterisks, and multi-part storylines.
- Character Card
- A structured profile (name, personality traits, backstory, speaking style) that defines how a character behaves, often shareable or importable between compatible apps.
- Character Gallery
- A browsable library of pre-made characters within an app, usually organized by category, popularity or tags, that users can start chatting with immediately.
- Content Filter (SFW Toggle)
- An in-app setting that controls whether generated text or images can include mature content, typically switchable between a safe-for-work mode and an unrestricted mode.
- In-Chat Image Gen
- A feature that generates an image of the character (or a described scene) directly inside the conversation, usually based on the character's card and the current context.
- LLM (Large Language Model)
- The underlying AI system that generates a character's replies. Apps may use a general-purpose LLM, a fine-tuned version, or a custom-trained model.
- Token / Gem Currency
- An in-app virtual currency, purchased with real money, that's spent on specific actions like sending messages, generating images or unlocking voice replies, as an alternative to a flat subscription.
- Companion vs. Chatbot
- A general chatbot is typically built for one-off task assistance without a persistent character. An AI companion app centers on a consistent character personality intended for repeated, ongoing interaction.
- Age Gate
- The verification step (ranging from a simple self-declared birthdate to ID-based checks) that an app uses to control who can create an account or access mature content.
Our take: this is the single feature most likely to determine whether an app feels like a continuing relationship or a series of disconnected chats — see our full memory guide.
Our take: apps differ a lot in how well story mode holds together over long sessions — short test chats often don't reveal this, so we test across longer conversations specifically.
Our take: the quality of the character-card system affects how consistent a character feels — a thin card produces generic responses regardless of the underlying model.
Our take: gallery size is often marketed heavily, but we weigh gallery quality and moderation over raw character count.
Our take: this is a separate control from age verification — see our privacy and age-verification guide for why that distinction matters.
Our take: consistency of the character's appearance across multiple generated images is one of the more variable and testable quality signals between apps.
Our take: the specific model in use often isn't disclosed in detail by the app, so we describe how the responses actually behaved rather than speculating about the exact model.
Our take: worth comparing against a flat-rate plan based on your actual usage pattern — token systems can be cheaper for light users and more expensive for heavy chatters.
Our take: this distinction is why we evaluate companion apps on continuity and character depth, not just answer accuracy.
Our take: age-gate strength varies widely between apps and isn't standardized across the category — check the specific app's method rather than assuming a baseline.
Related on CompanionRank
What Is Persona / Long-Term Memory? Privacy & Age Verification, Explained Companion Comparison — feature tablesFrequently asked questions
What's the difference between a companion app and a chatbot?
A general chatbot is typically built for one-off task help and doesn't maintain a consistent character or ongoing relationship. An AI companion app is designed around a persistent character with a personality, and often memory, meant to be interacted with over repeated sessions.
What does "LLM" mean in this context?
LLM stands for large language model — the underlying AI system that generates the character's replies. Different apps may use different LLMs, or custom-tuned versions of one, which is part of why the same character concept can feel different across platforms.
Why do some apps use tokens or gems instead of a flat subscription?
Token or gem systems let apps charge per action rather than a flat monthly fee. This can be cheaper for light users and more expensive for heavy users, so it's worth comparing against a flat-rate plan based on your expected usage.