Five clear examples

Coaching

A boss-fight strategy coach

The player shares a screenshot and current equipment. The AI identifies the encounter, explains the likely attack pattern, and proposes two strategies. It cannot control the character and asks before using any additional screen context.

Context
Screenshot, loadout, player question
Authority
Advice only
Value
Current encounter
In-game

An adaptive party member

A role-playing game includes an AI-controlled companion that remembers recent quests, comments on player decisions, and changes dialogue according to the current location and party state.

Context
Game state and dialogue history
Authority
Character behavior
Value
Inside one game world
Agentic

A commandable co-op teammate

The player asks an AI teammate to defend a location, collect a named resource, and retreat if health falls below a set threshold. The player can interrupt or revoke the command.

Context
Game state, action space, player commands
Authority
Direct actions
Value
Shared objective
Social

A gaming conversation companion

The system talks with the player about games, reacts to milestones, and maintains a consistent personality. It does not need screen access or the ability to act in a game.

Context
Conversation and preferences
Authority
Conversation only
Value
Social presence
Continuity

A long-RPG restore point

Before taking a break, the player saves the active character, current objective, major decisions, unresolved side quests, and intended next move. Weeks later, the AI restores that specific run instead of mixing it with another character.

Context
Intentional saved state
Authority
Store and restore
Value
Across sessions and chats

Hybrid examples

Coach + continuity

A strategy companion remembers the player’s preferred build and previous failed attempts, then uses that history during the next session. It advises now and preserves learning across time.

In-game + agentic

An adaptive party member talks as a character and also performs combat or exploration actions. The same system inhabits and acts.

Social + continuity

A conversational companion remembers which games and campaigns the player is discussing. The social layer and game-state layer should remain separately controllable.

Useful non-examples

SystemWhy it is not automatically an AI game companion
Static game wikiIt provides information but does not adapt through AI.
Ordinary cloud saveIt stores files but does not interpret or restore meaningful conversational context.
Scripted NPCIt may accompany the player, but its behavior is not necessarily AI-adaptive.
Generic chatbotIt needs a meaningful gaming role or gaming context to fit the category.
Screen recorderCapture alone is not interpretation or support.

How to evaluate a real product example

Before assigning a type, verify the product through current official documentation. Identify what information it can access, what actions it can perform, whether memory persists, what the player can delete or correct, and whether the advertised capability is actually available.

A category example proves that an architecture is conceptually possible. It does not prove that a named product implements the architecture safely or reliably.

Frequently asked questions

Why are these examples not brand rankings?

The purpose is to clarify architecture. Product capabilities change and should be verified through current primary sources.

Can a normal game character become an AI game companion?

Yes, when adaptive AI behavior gives it a meaningful supporting role rather than only a fixed script.

What makes the continuity example different from a transcript?

The state is intentional, named, scoped to the correct run, and designed to be restored later.