One label, several different questions
An “AI game” might mean a conventional game built with generative tools, an experience that generates content during play, or an interactive demonstration without stable rules or goals. Those distinctions change what a reader needs to know.
A game about artificial intelligence is another category altogether. Its subject does not tell us anything about its production process.
Production and runtime are separate
For AI-assisted production, ask what was used during development: code assistance, image generation, audio, or another workflow. A developer can use generated assets in a game that has no generative component during play.
For runtime generation, ask what changes while the player interacts and whether that behavior is central to the experience. This is an editorial distinction for organizing our coverage, not a claim that every product fits neatly into one box.
A showcase is evidence of a relationship
Rosebud’s official showcase links its platform to named projects. That is useful evidence of association. It does not reveal every development step or establish a percentage of generated work.
We therefore record the source of a claim separately from the experience of playing the game. “Official case” and “hands-on tested” answer different questions.
A classification you can actually use
| What is observed | Useful description | Evidence still needed |
|---|---|---|
| Generated images imported into a conventional project | AI-assisted art production | Which resources, how they were used, and their provenance |
| An assistant creates or revises gameplay code | AI-assisted game development | The retained implementation and subsequent edits |
| Dialogue or environments change through generation during play | Runtime generative interaction | What is generated, what persists, and what happens on failure |
| A navigable scene without a documented objective or stable rules | Interactive-world experiment | The actual interaction and its limits |
| A story features robots or machine intelligence | AI-themed game | Separate evidence of generative production or runtime use |
These descriptions can overlap. They organize questions rather than assign a quality score or certify a production process.
Consider a hypothetical puzzle with generated backgrounds and fixed, hand-written rules. Its art pipeline uses AI; that does not make the puzzle’s rules dynamically generated. If a later version adds model-generated hints, record that runtime function separately. One broad label would hide the change that matters to a player.
Keep three kinds of evidence apart
An official showcase establishes that a platform presents a project as its example. A creator’s production account may identify particular steps. Project files or a version history can support more detailed implementation claims. None automatically establishes the others.
Likewise, playing a working build can establish that a restart button works on a particular device and date. It cannot establish who wrote the underlying code. When reading an AI-game profile, look for a concrete claim, a source that can support it, and a clear account of what the editor actually observed.
What readers can expect here
Our profiles identify the available evidence, the claimed use of AI, and the limits of our own assessment. When we do not know the production process, we say so. A screenshot or a listing is not a substitute for a play session, and a play session is not an audit of how the game was made.