An AI-generated game is still a game. Players encounter controls, images, rules, sound, pacing and decisions—not a production pipeline. GameVibe therefore begins with the experience itself, while reporting clearly on how generative systems shaped it.

We do not award novelty points for a prompt

The use of AI can be newsworthy, but it is not evidence of quality. We ask whether the game is coherent, readable, responsive and worth a player’s time. A conventional production method does not excuse a dull game, and an unusual one does not excuse broken design.

At the same time, the method can affect the work. A game may generate dialogue during play, create levels from a model, assemble art assets before release or use AI only for early prototypes. Those are materially different practices. We avoid collapsing them into a single label.

What we disclose

Where information is available, coverage will identify the systems used, which parts of the game they influenced, whether generation happens during play and how much human selection or revision was involved. We will distinguish claims made by a developer from details we have independently verified.

We also look for the less visible work: dataset choices, licensing, consent, moderation, quality control and the labour required to turn generated material into a finished product. If those facts are unknown, we will say so rather than fill the gap with assumptions.

How we judge the result

Our criticism considers the same fundamentals that matter elsewhere—design, performance, accessibility, expression, originality and value—alongside questions specific to generative systems.

• Does generation create meaningful variety or merely more content?
• Can the player understand and influence the system?
• Does the game remain coherent across repeated sessions?
• Are failures handled safely and gracefully?
• Could the central experience exist without the AI component?

People remain part of the story

“Made by AI” can erase the decisions of the people who scoped, prompted, selected, edited, integrated and shipped the work. We will credit that labour and avoid language that assigns human intent to a model. We will also take concerns from artists, performers, writers, developers and players seriously rather than treating them as obstacles to innovation.

The field is young, its terminology is unstable and its most confident predictions are often wrong. Our job is to document what was built, test what it does and judge the experience with clarity. Neither celebration nor rejection should arrive before the evidence.