October 8, 2026
1 user reviews
Indexed September 7, 2026
View on Steam

Disclosure analysis

The game uses live AI while you play

The game uses a local language model to generate narration, dialogue, and lore during gameplay, with player input and filtered output.

Process
AI-generated and AI-assisted
Live AI during play
Yes
Evidence & analysis details
Evidence used
“Some 2D illustrations and the pre-recorded narration were produced with AI tools”
“A language model bundled with the game runs entirely on the player's device”
Model-reported confidence
98% A model estimate, not a measured accuracy rate or a human review score.

Auto-classified from the public disclosure. The raw disclosure below remains the source of truth. How classification works.

AI Usage Disclosure

Pre-generated: Some 2D illustrations and the pre-recorded narration were produced with AI tools and reviewed by the developers before inclusion. Live-generated: A language model bundled with the game runs entirely on the player's device, with no internet connection. During play it writes narration, dialogue, and the names and lore of items, relics and skills; it selects among game-defined options for run content such as synergies, oaths and subclasses; and it accepts free-text actions typed by the player (for example "cut the chandelier loose"), judges them, and narrates the result. Damage, stats, difficulty, and progression are always governed by the game's own rules. Player input is filtered before it reaches the model, every generated line is checked before it is shown, and no art or audio is generated at runtime.

Highlights are auto-detected and may not be fully accurate. Report an issue

This disclosure in context

How comparisons work

Compare this developer’s disclosure with 26,215 non-mature Steam products in the index. These figures describe published disclosures, not all games on Steam or independently verified AI use.

Live AI during play
3.2%
828 of 25,784 disclosures with structured analysis.
Compare this group
Writing or localization
28.2%
7,405 of 26,215 indexed disclosures mention this category.
Compare this category

Categories can overlap. A disclosure may mention several types of AI use.

Source and record

First indexed
Sep 7, 2026
Source last checked
Oct 20, 2026
Automatic analysis updated
Oct 18, 2026
Recorded text changes
1 Read the history

The disclosure is the developer’s statement on Steam. Index dates show when we observed it, not when AI use began. History covers captured versions only. Automatic analysis can be wrong and is not a verification of the game’s content. Read the methodology.

Disclosure History

Captured versions of the developer’s statement. This record can miss changes between checks and does not establish when AI use began.

Pre-generated: Some 2D illustrations and the pre-recorded narration were produced with AI tools and reviewed by the developers before inclusion. Live-generated: An on-device language model (no internet connection) writes in-game text such as narration, dialogue, and the names and lore of items, relics and skills, and selects among game-defined options for certain run content. Damage, stats, difficulty and progression are governed by the game's own rules. Generated text is checked before it is shown, and no art or audio is generated at runtime.

Pre-generated: Some 2D illustrations and the pre-recorded narration were produced with AI tools and reviewed by the developers before inclusion. Live-generated: A language model bundled with the game runs entirely on the player's device, with no internet connection. During play it writes narration, dialogue, and the names and lore of items, relics and skills; it selects among game-defined options for run content such as synergies, oaths and subclasses; and it accepts free-text actions typed by the player (for example "cut the chandelier loose"), judges them, and narrates the result. Damage, stats, difficulty, and progression are always governed by the game's own rules. Player input is filtered before it reaches the model, every generated line is checked before it is shown, and no art or audio is generated at runtime.