Public Steam disclosure research25,376 disclosures tracked
Where do games disclose AI use?
Search developers’ Steam disclosures, compare classified AI use across
games, and follow recorded changes.
This index adds cross-game classifications and statistics to
statements published on individual store pages.
Groups reflect automatic analysis of public disclosures; mature records are excluded. These counts do not measure AI use across all of Steam.
Missing or unclear statements cannot establish whether a game uses AI.
A Light in the Gloam uses pre-generated AI content. Nothing is generated while the game is running, and the game makes no network calls of any kind.
Visual art: all in-game art was generated with OpenAI's image model (gpt-image-1) from prompts written by hand, then processed by our own scripts, which downscale it, quantise it to a fixed palette and trim it before it is placed in the game. This covers 14 character portraits, 43 enemy and boss sprites, 63 item icons, 27 wall, floor and ceiling textures, 10 dungeon props, 21 achievement emblems, and 14 town and title backdrops. The store capsules and the client icons are crops of the title-screen generation rather than new images.
Music and sound: all 14 music tracks were generated with ElevenLabs' music model and all 47 sound effects with ElevenLabs' sound-effects model, both from written prompts and both post-processed by our own scripts for level and length. There is no voice acting and no synthesised speech anywhere in the game.
Writing and translation: the game's text, meaning item and monster descriptions, dialogue, menus and the story, was written with the assistance of a large language model and then edited by hand. The Brazilian Portuguese translation was produced the same way and reviewed against the English.
Atharia uses generative AI tools under the developer's constant direction. The combat animations are AI-generated video clips, converted into sprites and integrated into custom-built staging. Background environments and interface elements are also created with AI assistance. Every asset is art-directed, selected and validated by the developer before it enters the game.
Generative AI has been used to help work with coding.
All Assets (art, models, animations, sprites/textures, sounds and music have been made by myself or have been bought)
Frontier was conceived, designed, directed, and extensively play tested by a solo developer. Generative AI was used as the primary production tool to implement that vision, including programming, artwork, audio, writing, and other game assets. Every aspect of the game was developed through an iterative human-directed process, with the developer specifying requirements, reviewing results, requesting revisions, making design decisions, integrating systems, and personally testing and approving the final content. No generative AI is used during gameplay.
DAWNBLOOM uses generative AI-assisted tools during the development of certain game content. The in-game “Dawnbloom Spirit” feature connects to a generative AI service to provide real-time text responses to players’ game-related questions, including game knowledge, gameplay explanations, and general guidance. This feature generates auxiliary text content only. It does not generate or modify game code, nor does it directly control game rules, character data, or player assets.
Generative AI was used during development. Part of the graphics (characters, enemies, bosses, icons, room backgrounds and backdrops) was created with image-generation tools: Google Gemini, OpenAI ChatGPT and a locally run Stable Diffusion model. Part of the music was created with audio-generation tools (Google Gemini). Part of the code, room layouts and texts were written with the help of an AI assistant. The developer reviewed, cut, retouched and hand-adjusted all of this material before adding it to the game. The game does not generate any AI content while it is being played.
Some pre-generated 2D artwork, visual effects, icons, maps, promotional images, and achievement artwork were created or refined with generative AI during development. All AI-generated output was reviewed, edited, and integrated by the development team. The game does not use generative AI at runtime and does not generate player-facing text, images, audio, music, code, or other content while you play.
ISKRA currently includes AI-generated artwork prepared during development. We plan to replace this artwork with work created by human artists before leaving Early Access, giving the game the individual artistic identity we envision.
Some of the visual assets in Twin Compass, including enemy and character illustrations, background art, and UI decoration, were created with the assistance of generative AI tools, then selected, edited, composited, and hand-tuned by the developer for use in the game. All game design, code, writing, systems, and final asset integration are human-authored. No AI is used at runtime, the game generates nothing on the player's machine and makes no external requests during play.
All artwork, animation, writing, character designs and other creative content in this game were made by hand. None of it was created with generative AI.
Generative AI was used only in two parts of development:
Programming assistance: helping to write and debug parts of the game code.
Translation assistance: helping to translate the text for non-Chinese language versions.
The artwork and writing that players see in the game were not generated by AI.
Generative AI was used primarily to assist in the creation of some visual assets and images used in the game, including illustrations, UI artwork, and promotional materials. AI-generated images were manually reviewed, selected, edited, and adjusted by the developer before being integrated into the game.
The game does not use generative AI to create content in real time during gameplay.
The game engine is written with substantial AI assistance.
Some dialogues and quest texts are AI generated, and are successively replaced by human-written dialogue. Every piece of flavor text has a label that states the author (human or ai).
The 3D character models are currently generated with AI, and then transformed through a custom pipeline to work with the text-based graphics and CPU rendering. It is possible to replace them with manually crafted models.
Some promotional artwork on this store page was created using generative AI tools. All in-game assets (characters, environments, effects) are original or licensed third-party content.
Some in-game visual assets (icons and other artwork) were created with generative AI tools during development. Localization text was also drafted with AI and then reviewed by the developer. All shipped images and translations were checked before release. No AI content is generated while the game is running.
Generative AI is used to story board, code, analyze code, analyze performance, create thumbnails and many basic tasks like that to speed up workflows. We pay human artists for things like zombies, guns, props, static meshes etc.
Portraits of characters and creatures are generated using Stable Diffusion. They appear in-game and in store page materials. I plan to replace all generated portraits with hand-drawn art.
Some in-game labels and text descriptions are placeholders generated by an LLM. I am gradually rewriting them by hand.
Time Traveler: Divergence uses locally running generative AI to enable dynamic NPC conversations. AI-generated dialogue is constrained by the game's predefined rules, characters, world knowledge, and simulation state. AI is also used locally to generate voice-over for these dynamic conversations. No external AI service is required during gameplay.
Serv-Bot's Firewall uses AI tools in its art and media pipeline, always under human direction and review:
Sprite art. The game's original hand-authored sprites were upscaled and repainted at native resolution with an image-editing model (Qwen-Image-Edit), then reviewed, corrected and in many cases restored to the originals by hand. Concept art for the six "arsenal" enemy types, the Serv-Bot HUD portraits and the achievement icon motifs were generated with local image models (Z-Image Turbo, Qwen-Image-Edit) from our own designs and reference art, then composited and finished in code. Canon elements such as the Greeb ship in the icons are the game's own sprite art, not generated.
Cutscenes. The in-game story videos come from our SIGNAL series. Keyframes were generated with a local image model trained on our own characters, motion with a local video model (Wan 2.2), narration with a text-to-speech model (Kokoro) and music with local music models (ACE-Step, Stable Audio Open). Scripts, story, character designs, editing and direction are ours.
Localization. Interface text in 28 languages was machine-translated with an AI model from our English source and has not yet been reviewed by native speakers.
Code. Parts of the game's code were written with AI coding assistance, reviewed and tested by the developer.
Sound effects are procedurally synthesised, not AI-generated. All AI generation ran on the developer's own hardware; no copyrighted third-party works were used as inputs beyond our own assets and characters.
I use AI to assist in coding, and some skill icons for combat/talent trees; however all character sprites/animations are hand animated. Using AI allows me to make a game with all of the ideas I have and transform them into something tangible, and viewable. The goal is simple, create a fun game that people enjoy playing.
During development, generative AI tools (such as ChatGPT) were used to assist in writing and optimizing portions of the Godot (GDScript) game code, as well as for brainstorming visual concepts during the early stages of art design. All final art assets, text, audio, and code within the game have been reviewed, modified, and finalized by the developer prior to release. The game does not generate any AI content in real-time during gameplay.
As a solo developer with a big dream and limited resources, I want to be entirely upfront how this game was built. The vast majority of this game is completely human-made: I crafted the pixel characters, animations, code, and story over the past 10 years. I also utilized incredible royalty-free and purchased assets from human creators like gyvr, Eunity, RafaelMatos, penusbmic, Winlu, Geoff Harvey.
The use of generative AI was for touching up and refining the marketing capsule images using Adobe Photoshop tools.
Hero Chains of Destiny includes pre-generated content created with the assistance of generative AI tools during development. This includes selected artwork, music, and sound effects. All generated content is reviewed, curated, and approved by the developer before being included in the game.
Some of the game's artwork, including characters, monsters, items and backgrounds, was created with the help of generative AI tools and then edited and refined by our team; none of it is used as raw AI output. Parts of the in-game text and translations were also produced with AI assistance. No AI content is generated during gameplay.
AI tools were used during development to create pre-generated character art and animations, backgrounds, interface artwork, and sound effects, as well as Steam store descriptions and promotional artwork, including the animated store banner. Assets were selected, reviewed, processed or edited as needed, and integrated into the game or its Steam presentation. Music was composed and synthesized with AI-assisted development tools. The shipped game uses these prepared assets; it does not generate AI content or connect to AI services during gameplay.
Generative AI tools are used during development to create some 3D assets as Place Holders and assist with Art/ music Content. These materials are prepared in advance and incorporated into the game or its store page. The game does not generate content using generative AI during gameplay.
Some character voice recordings are also created using generative AI.
Most of the game's visual art and sound was pre-generated with AI tools and then selected, edited and placed by the developer: the dungeon panoramas, room props, floors, town and building plates, character portraits, walk and combat sprite sheets, monster, boss and villager sprites, item, skill and stack icons, combat and reward effects, and the store page art were generated with the FLUX.1 image model through fal.ai and animated with AutoSprite; the sound effects, ambience beds, voice barks and music were generated with ElevenLabs. Every generated file, the tool that made it and the prompt that asked for it are recorded in the game's files (assets/README.md and assets/PROMPTS.md). All game text, design, rules and code are written by people. Nothing is generated while you play: the game contains no live AI generation and calls no AI service at runtime.
Generative AI was used during development to assist with the creation and refinement of some visual assets, including character artwork, environmental artwork, icons, and promotional imagery. AI-assisted assets were reviewed, edited, and integrated by the developer to maintain a consistent visual style. Generative AI is not used during gameplay and does not generate content in real time.
Generative AI was used during development to create pre-generated visual assets for the game, including character artwork, card artwork, backgrounds, visual effects, textures, UI artwork, and promotional materials.
All AI-generated content is reviewed and curated by the developer before being included in the game.
The game does not use live generative AI during gameplay.
As with most VTT systems, users are welcome to import assets made anywhere. Phroller makes it easier than ever. You can just drag the image or model files into the VTT directly from your browser or file system!
I want to encourage supporting real artists where we can though. As such I do my best to collaborate with real creatives on the vast Phroller libraries. Each one of the assets in the library has a field showing the artist behind it. If you like their work, please consider checking them out. Many of them are open for commissions as well! However, I cannot guarantee that assets or part of those assets were created with no AI use at all. A large amount of the assets are made before 2022 though, so we should be safe there.
While I do my best to make Phroller as user friendly as possible, it does still require a ton of instructions and manuals. Frontier models are used to keep some of the text up to date across rapid early access development. Sorry if it sounds robotic!
A Light in the Gloam uses pre-generated AI content. Nothing is generated while the game is running, and the game makes no network calls of any kind.
Visual art: all in-game art was generated with OpenAI's image model (gpt-image-1) from prompts written by hand, then processed by our own scripts, which downscale it, quantise it to a fixed palette and trim it before it is placed in the game. This covers 14 character portraits, 43 enemy and boss sprites, 63 item icons, 27 wall, floor and ceiling textures, 10 dungeon props, 21 achievement emblems, and 14 town and title backdrops. The store capsules and the client icons are crops of the title-screen generation rather than new images.
Music and sound: all 14 music tracks were generated with ElevenLabs' music model and all 47 sound effects with ElevenLabs' sound-effects model, both from written prompts and both post-processed by our own scripts for level and length. There is no voice acting and no synthesised speech anywhere in the game.
Writing and translation: the game's text, meaning item and monster descriptions, dialogue, menus and the story, was written with the assistance of a large language model and then edited by hand. The Brazilian Portuguese translation was produced the same way and reviewed against the English.
Atharia uses generative AI tools under the developer's constant direction. The combat animations are AI-generated video clips, converted into sprites and integrated into custom-built staging. Background environments and interface elements are also created with AI assistance. Every asset is art-directed, selected and validated by the developer before it enters the game.
Generative AI has been used to help work with coding.
All Assets (art, models, animations, sprites/textures, sounds and music have been made by myself or have been bought)
Frontier was conceived, designed, directed, and extensively play tested by a solo developer. Generative AI was used as the primary production tool to implement that vision, including programming, artwork, audio, writing, and other game assets. Every aspect of the game was developed through an iterative human-directed process, with the developer specifying requirements, reviewing results, requesting revisions, making design decisions, integrating systems, and personally testing and approving the final content. No generative AI is used during gameplay.
DAWNBLOOM uses generative AI-assisted tools during the development of certain game content. The in-game “Dawnbloom Spirit” feature connects to a generative AI service to provide real-time text responses to players’ game-related questions, including game knowledge, gameplay explanations, and general guidance. This feature generates auxiliary text content only. It does not generate or modify game code, nor does it directly control game rules, character data, or player assets.
Generative AI was used during development. Part of the graphics (characters, enemies, bosses, icons, room backgrounds and backdrops) was created with image-generation tools: Google Gemini, OpenAI ChatGPT and a locally run Stable Diffusion model. Part of the music was created with audio-generation tools (Google Gemini). Part of the code, room layouts and texts were written with the help of an AI assistant. The developer reviewed, cut, retouched and hand-adjusted all of this material before adding it to the game. The game does not generate any AI content while it is being played.
Some pre-generated 2D artwork, visual effects, icons, maps, promotional images, and achievement artwork were created or refined with generative AI during development. All AI-generated output was reviewed, edited, and integrated by the development team. The game does not use generative AI at runtime and does not generate player-facing text, images, audio, music, code, or other content while you play.
ISKRA currently includes AI-generated artwork prepared during development. We plan to replace this artwork with work created by human artists before leaving Early Access, giving the game the individual artistic identity we envision.
Some of the visual assets in Twin Compass, including enemy and character illustrations, background art, and UI decoration, were created with the assistance of generative AI tools, then selected, edited, composited, and hand-tuned by the developer for use in the game. All game design, code, writing, systems, and final asset integration are human-authored. No AI is used at runtime, the game generates nothing on the player's machine and makes no external requests during play.
All artwork, animation, writing, character designs and other creative content in this game were made by hand. None of it was created with generative AI.
Generative AI was used only in two parts of development:
Programming assistance: helping to write and debug parts of the game code.
Translation assistance: helping to translate the text for non-Chinese language versions.
The artwork and writing that players see in the game were not generated by AI.
Generative AI was used primarily to assist in the creation of some visual assets and images used in the game, including illustrations, UI artwork, and promotional materials. AI-generated images were manually reviewed, selected, edited, and adjusted by the developer before being integrated into the game.
The game does not use generative AI to create content in real time during gameplay.
The game engine is written with substantial AI assistance.
Some dialogues and quest texts are AI generated, and are successively replaced by human-written dialogue. Every piece of flavor text has a label that states the author (human or ai).
The 3D character models are currently generated with AI, and then transformed through a custom pipeline to work with the text-based graphics and CPU rendering. It is possible to replace them with manually crafted models.
Some promotional artwork on this store page was created using generative AI tools. All in-game assets (characters, environments, effects) are original or licensed third-party content.
Some in-game visual assets (icons and other artwork) were created with generative AI tools during development. Localization text was also drafted with AI and then reviewed by the developer. All shipped images and translations were checked before release. No AI content is generated while the game is running.
Generative AI is used to story board, code, analyze code, analyze performance, create thumbnails and many basic tasks like that to speed up workflows. We pay human artists for things like zombies, guns, props, static meshes etc.
Portraits of characters and creatures are generated using Stable Diffusion. They appear in-game and in store page materials. I plan to replace all generated portraits with hand-drawn art.
Some in-game labels and text descriptions are placeholders generated by an LLM. I am gradually rewriting them by hand.
Time Traveler: Divergence uses locally running generative AI to enable dynamic NPC conversations. AI-generated dialogue is constrained by the game's predefined rules, characters, world knowledge, and simulation state. AI is also used locally to generate voice-over for these dynamic conversations. No external AI service is required during gameplay.
Serv-Bot's Firewall uses AI tools in its art and media pipeline, always under human direction and review:
Sprite art. The game's original hand-authored sprites were upscaled and repainted at native resolution with an image-editing model (Qwen-Image-Edit), then reviewed, corrected and in many cases restored to the originals by hand. Concept art for the six "arsenal" enemy types, the Serv-Bot HUD portraits and the achievement icon motifs were generated with local image models (Z-Image Turbo, Qwen-Image-Edit) from our own designs and reference art, then composited and finished in code. Canon elements such as the Greeb ship in the icons are the game's own sprite art, not generated.
Cutscenes. The in-game story videos come from our SIGNAL series. Keyframes were generated with a local image model trained on our own characters, motion with a local video model (Wan 2.2), narration with a text-to-speech model (Kokoro) and music with local music models (ACE-Step, Stable Audio Open). Scripts, story, character designs, editing and direction are ours.
Localization. Interface text in 28 languages was machine-translated with an AI model from our English source and has not yet been reviewed by native speakers.
Code. Parts of the game's code were written with AI coding assistance, reviewed and tested by the developer.
Sound effects are procedurally synthesised, not AI-generated. All AI generation ran on the developer's own hardware; no copyrighted third-party works were used as inputs beyond our own assets and characters.
I use AI to assist in coding, and some skill icons for combat/talent trees; however all character sprites/animations are hand animated. Using AI allows me to make a game with all of the ideas I have and transform them into something tangible, and viewable. The goal is simple, create a fun game that people enjoy playing.
During development, generative AI tools (such as ChatGPT) were used to assist in writing and optimizing portions of the Godot (GDScript) game code, as well as for brainstorming visual concepts during the early stages of art design. All final art assets, text, audio, and code within the game have been reviewed, modified, and finalized by the developer prior to release. The game does not generate any AI content in real-time during gameplay.
As a solo developer with a big dream and limited resources, I want to be entirely upfront how this game was built. The vast majority of this game is completely human-made: I crafted the pixel characters, animations, code, and story over the past 10 years. I also utilized incredible royalty-free and purchased assets from human creators like gyvr, Eunity, RafaelMatos, penusbmic, Winlu, Geoff Harvey.
The use of generative AI was for touching up and refining the marketing capsule images using Adobe Photoshop tools.
Hero Chains of Destiny includes pre-generated content created with the assistance of generative AI tools during development. This includes selected artwork, music, and sound effects. All generated content is reviewed, curated, and approved by the developer before being included in the game.
Some of the game's artwork, including characters, monsters, items and backgrounds, was created with the help of generative AI tools and then edited and refined by our team; none of it is used as raw AI output. Parts of the in-game text and translations were also produced with AI assistance. No AI content is generated during gameplay.
AI tools were used during development to create pre-generated character art and animations, backgrounds, interface artwork, and sound effects, as well as Steam store descriptions and promotional artwork, including the animated store banner. Assets were selected, reviewed, processed or edited as needed, and integrated into the game or its Steam presentation. Music was composed and synthesized with AI-assisted development tools. The shipped game uses these prepared assets; it does not generate AI content or connect to AI services during gameplay.
Generative AI tools are used during development to create some 3D assets as Place Holders and assist with Art/ music Content. These materials are prepared in advance and incorporated into the game or its store page. The game does not generate content using generative AI during gameplay.
Some character voice recordings are also created using generative AI.
Most of the game's visual art and sound was pre-generated with AI tools and then selected, edited and placed by the developer: the dungeon panoramas, room props, floors, town and building plates, character portraits, walk and combat sprite sheets, monster, boss and villager sprites, item, skill and stack icons, combat and reward effects, and the store page art were generated with the FLUX.1 image model through fal.ai and animated with AutoSprite; the sound effects, ambience beds, voice barks and music were generated with ElevenLabs. Every generated file, the tool that made it and the prompt that asked for it are recorded in the game's files (assets/README.md and assets/PROMPTS.md). All game text, design, rules and code are written by people. Nothing is generated while you play: the game contains no live AI generation and calls no AI service at runtime.
Generative AI was used during development to assist with the creation and refinement of some visual assets, including character artwork, environmental artwork, icons, and promotional imagery. AI-assisted assets were reviewed, edited, and integrated by the developer to maintain a consistent visual style. Generative AI is not used during gameplay and does not generate content in real time.
Generative AI was used during development to create pre-generated visual assets for the game, including character artwork, card artwork, backgrounds, visual effects, textures, UI artwork, and promotional materials.
All AI-generated content is reviewed and curated by the developer before being included in the game.
The game does not use live generative AI during gameplay.
As with most VTT systems, users are welcome to import assets made anywhere. Phroller makes it easier than ever. You can just drag the image or model files into the VTT directly from your browser or file system!
I want to encourage supporting real artists where we can though. As such I do my best to collaborate with real creatives on the vast Phroller libraries. Each one of the assets in the library has a field showing the artist behind it. If you like their work, please consider checking them out. Many of them are open for commissions as well! However, I cannot guarantee that assets or part of those assets were created with no AI use at all. A large amount of the assets are made before 2022 though, so we should be safe there.
While I do my best to make Phroller as user friendly as possible, it does still require a ton of instructions and manuals. Frontier models are used to keep some of the text up to date across rapid early access development. Sorry if it sounds robotic!
