Yes—when a game ships AI-created content for players or uses AI services in the product, Steam requires the developer to disclose that use in the Steamworks Content Survey. Valve’s survey has a dedicated “Generative Artificial Intelligence Content” section. Developers must explain the implementation, and live, in-game generation requires additional information about safeguards intended to prevent illegal output.
Using an AI tool privately to speed up coding, brainstorming or other development work is different. Valve says that efficiency use is not the focus of this section; the key question is whether AI-generated material reaches players or an AI service is part of the shipped product. Disclosure is part of Steam’s review process, not permission to violate Steam’s content rules.
What exactly does Steam require developers to disclose?
Steam requires applicable developers to report generative-AI use through the Steamworks Content Survey before submitting the store page and product build for review. The survey asks about AI-created content that ships with the game and is consumed by players, as well as AI services incorporated into the product.
The obligation is broader than a single type of asset. Depending on how the game was made and what players receive, the relevant content can include:
- Artwork, such as characters, environments, textures, icons or promotional material included in the product.
- Sound, music, voice, dialogue or other audio generated with AI assistance.
- Narrative text, quests, item descriptions, dialogue and other written material.
- Localization or translated content produced with AI tools.
- Other player-facing material created with generative systems during development.
- AI services that operate as part of the product rather than merely helping the development team.
Valve’s documentation says the developer must describe the implementation in detail. That means a short statement such as “AI was used” is unlikely to communicate the information the survey is asking for. The useful description identifies what the system generated, where that output appears, whether it was created before release or during play, and how the feature is controlled.
Do I have to tell Steam if I used AI art while making my game?
If the AI-created art ships in the game and players can consume it, yes, it falls within the disclosure focus of Steam’s survey. If an AI tool was used only as a private productivity aid and none of its generated output reaches players, Valve’s documentation distinguishes that activity from the player-facing content covered by this section.
Player-facing output is the key test
Consider an artist who uses an image model to produce a texture, then includes that texture in the final build. Players encounter the texture, so the developer should report the use and explain how the asset was incorporated. The same principle applies to generated dialogue, music, localization and other content that remains in the shipped game.
By contrast, using an AI assistant to suggest code, summarize bug reports or draft an internal design outline does not, by itself, make the tool a player-facing feature. The distinction is not whether a developer touched an AI system at some point; it is whether AI-created material ships for players or an AI service is part of the product.
Mixed workflows need a complete description
Many projects combine human-authored and AI-assisted work. A developer can describe the combination rather than treating the project as wholly AI-generated. For example, a team might explain that an AI system produced initial concept variations, artists revised the selected designs, and only the revised assets appear in the game. The survey’s purpose is to disclose the implementation, not to force every asset into an “all AI” or “no AI” label.
What is the difference between pre-generated and live-generated AI content?
Pre-generated AI content is created during development and packaged with the game. Live-generated content is produced by an AI system while the game is running. Steam applies its general content expectations to both categories, but live generation carries an extra disclosure requirement for guardrails against illegal output.
| AI category | When content is created | How players receive it | What the developer must explain | Additional review concern |
|---|---|---|---|---|
| Pre-generated | During development, before release | Included in the shipped build or other player-facing game content | What was generated, how it was used and where it appears | Output must comply with Steam’s rules and match the game’s marketing |
| Live-generated | While the game is running | Created in response to player actions, prompts or other runtime events | The implementation plus the controls and guardrails used to limit illegal output | The runtime system must still comply with Steam’s content rules as outputs are generated |
How pre-generated content is reviewed
Valve treats pre-generated AI output under the same general pre-release approach used for non-AI content. The fact that an image, line of dialogue or sound effect came from a model does not remove the developer’s responsibility for what ships. Content must not be illegal or infringing, and it should not contradict promises made in the game’s store materials.
Why live generation requires more information
A live system can produce new output after release, so the review question is not limited to a fixed set of files. Steam requires the developer to explain the guardrails intended to prevent illegal content. The disclosure should therefore cover the runtime design and its controls, not just the model’s name.
Useful implementation details can include where prompts originate, what inputs are blocked or filtered, whether outputs pass through moderation before display, what happens when a request is rejected, and how the developer handles abuse reports or unsafe responses. Valve’s requirement is to describe how the implementation works; developers should provide the concrete controls that actually exist rather than promising safeguards that are not implemented.
Does Steam allow games with AI-generated assets?
Disclosure does not make otherwise prohibited material acceptable. Steam’s official survey guidance says products must follow the platform’s content rules regardless of whether AI use was disclosed. A developer cannot rely on a completed survey as a waiver, exemption or guarantee of approval.
For pre-generated assets, the normal content obligations still apply. Generated art, text, audio and other material must not be illegal or infringing, and the shipped content should be consistent with the claims made on the store page. If marketing presents a game as containing a particular style, feature or level of content, the actual product should not materially contradict that presentation.
Live generation adds operational risk because the developer may not know every output in advance. The game’s safeguards need to address the foreseeable ways a system could produce illegal material. Even with safeguards, the developer remains responsible for meeting Steam’s rules; disclosure alone does not transfer that responsibility to Valve.
When must the survey be completed?
The Steamworks Content Survey must be completed before the developer submits the store page and product build for review. Treat the survey as a release-readiness task, not something to complete after approval or after launch.
- Inventory the AI use. List every AI system used to create material that will ship, and every AI service that will operate in the product. Separate internal productivity use from player-facing output.
- Classify the output. Mark each relevant use as pre-generated or live-generated. A project can contain both categories.
- Map the player experience. Record where players encounter the output: in the build, in downloadable content, in generated conversations, in localization or through another product feature.
- Document the implementation. Prepare a plain-language explanation of what the system does, what data or prompts it receives, what it generates and how the result reaches players.
- Document live-generation safeguards. If output is produced during play, describe the guardrails designed to prevent illegal content, including filtering, moderation, refusal handling and other controls that are actually in use.
- Complete the Generative Artificial Intelligence Content section. Enter the information in the Steamworks Content Survey before sending the store page and build for review.
- Check the product against Steam’s ordinary rules. Review the final content and marketing claims independently of the disclosure. A truthful survey answer cannot cure prohibited content or an inaccurate store description.
What information should a developer prepare for the survey?
Valve’s requirement is a detailed description, so preparation should focus on the facts a reviewer needs to understand the feature. Keep an internal record that answers these questions:
- Which assets or features were created with generative AI?
- Were the outputs generated before release, during play, or both?
- Are players able to see, hear, read or otherwise consume the output?
- Is an AI service called by the game after installation, or is all generated material packaged locally?
- What parts were edited, selected, filtered or approved by people?
- What inputs can players provide, if any?
- For live generation, what blocks illegal requests or unsafe output?
- What does the game do when a request is refused, filtered or unavailable?
This record also helps resolve borderline cases. For example, an AI-generated placeholder used only during development is not the same as an AI-generated character portrait that remains in the final build. Likewise, a game that calls an AI service to create dialogue at runtime has a different disclosure burden from a game that includes a fixed set of dialogue files generated months earlier.
Is using an AI coding assistant different from shipping AI-generated content?
Yes. Valve specifically says that efficiency gains from using AI tools during development are not the focus of the generative-AI section. Private use for coding, planning, debugging or similar productivity work is therefore distinct from shipping AI-created material or incorporating an AI service into the game.
The distinction can change if the development tool’s output becomes part of the player experience. Code suggestions that help implement a conventional feature are not themselves player-facing generated content. But if the finished product includes an AI feature that generates dialogue, images, music or other material for players, that product feature is within the survey’s scope even if the team also used AI privately for ordinary development tasks.
What happens if a game uses live AI generation?
The developer must disclose the live system and explain its guardrails against illegal output. The requirement applies because runtime generation can create new material after the game has been submitted, so the implementation and its controls are part of the review context.
A useful description distinguishes the model or service from the surrounding safety design. State how player input reaches the generator, what moderation or filtering occurs before output is shown, how blocked requests are handled, and what fallback appears when generation fails. Do not imply that a model is safe simply because it was trained for general-purpose use; the survey asks how the game’s implementation is controlled.
Does disclosure guarantee that Steam will approve the game?
No. Disclosure is a compliance step, not an approval promise. Steam’s content rules continue to apply whether a game uses AI, traditional tools or a mixture of both.
Review risk remains tied to the content and the way the game is presented. Illegal or infringing material can cause problems in either a conventional or AI-assisted workflow, and a store page that promises something the product does not deliver can create a separate issue. Live systems also require the developer to account for foreseeable unsafe output through the guardrails described in the survey.
What is still unclear about Steam’s AI disclosure policy?
Valve’s documentation establishes the survey requirement and the distinction between pre-generated and live-generated content, but it does not clearly establish that AI answers are displayed to customers in a dedicated public store-page field. Developers should not promise players a particular public label unless Steamworks provides that information for their product.
The available guidance also does not set out country-by-country variations in this AI-disclosure rule. A developer releasing in multiple regions should follow the Steamworks instructions presented for its account and product, and separately consider any legal disclosure, copyright or consumer-protection duties that may apply in the relevant markets.
What should developers do before submitting an AI-assisted game?
Use a short compliance check before submission:
- Confirm which AI-created outputs remain in the shipped game.
- Separate fixed, pre-generated assets from runtime-generated output.
- Identify every AI service that the product itself calls.
- Write a detailed, accurate implementation description.
- Describe live-generation guardrails and fallback behavior where applicable.
- Check generated material for illegal or infringing content.
- Compare the final game with every AI-related claim on the store page.
- Complete the Generative Artificial Intelligence Content section before submitting the page and build for review.
The most defensible approach is to disclose the player-facing implementation plainly and keep evidence of how the system works. That makes the survey answer consistent with the build, the store page and the controls that operate after launch.
Frequently Asked Questions
Will Steam show my AI disclosure publicly on the store page?
Valve’s available Content Survey guidance does not clearly state that every AI answer is displayed to customers in a dedicated public field. Complete the survey accurately, but do not advertise a specific public badge or label unless Steamworks shows one for your product.
Do AI-generated assets have to be credited individually in the game?
The survey requirement is about reporting the implementation to Steam. The cited guidance does not establish a universal in-game credit format for each generated asset, so any credits should follow the game’s own attribution practices and the licenses governing the tools and source material.
What if my game uses an external AI API after launch?
That is an AI service incorporated into the product, so the implementation belongs in the relevant survey disclosure. Explain when the game calls the service, what players receive and what safeguards limit illegal output if generation happens at runtime.
Can Steam reject a game even after I disclose its AI use?
Yes. Disclosure does not override Steam’s other content requirements. Illegal or infringing output, or store-page claims that do not match the product, can remain review problems regardless of the survey answer.