The Disclosure Framework Just Got a Major Rewrite

Valve rewrote Steam's AI disclosure rules in February 2026, and the reaction from indie developers has been a mix of relief, confusion, and mild panic. The new framework replaces the binary "yes I used AI / no I didn't" checkbox with a two-tier classification system, adds explicit exemptions for developer efficiency tools, and introduces new requirements for games that use live AI generation at runtime.

If you ship games on Steam — or plan to — you need to understand these changes. Not because they're punitive (they're surprisingly reasonable), but because getting disclosure wrong can delay your store page review, trigger community backlash, or in the worst case, get your game delisted.

This post breaks down exactly what changed, what it means for different types of AI usage, and how to navigate the system without overthinking it.

The Two-Tier System: Pre-Generated vs. Live-Generated

The centerpiece of the new framework is the distinction between pre-generated and live-generated AI content. This isn't just a labeling difference — each tier has different disclosure requirements, review processes, and player-facing labels.

Tier 1: Pre-Generated AI Content

Pre-generated content is anything created using AI tools during development that ships as static assets in the final build. This includes:

• Textures, concept art, or sprites generated or assisted by image models
• Dialogue, lore, or item descriptions written or edited with language models
• Music or sound effects produced with audio generation tools
• 3D models or meshes created with AI geometry tools
• Code generated by AI coding assistants (with an important exception — see below)

The key characteristic: the AI output was reviewed, edited, and approved by a human developer before it reached the player. The player experiences a fixed, curated result.

Disclosure requirement: A single checkbox on the store page setup confirming AI was used in content creation, plus a brief description of which content categories were affected. This appears as a small disclosure tag on your store page.

Review impact: Minimal. Valve reviews the disclosure for completeness but doesn't evaluate the quality of AI-assisted content differently from hand-made content. Your store page review timeline shouldn't be affected.

Tier 2: Live-Generated AI Content

Live-generated content is anything produced by AI models at runtime during gameplay. This includes:

• NPC dialogue generated by language models in real time
• Procedurally generated quests, stories, or narrative content using AI
• Dynamic music or audio generated during gameplay
• AI-generated textures or visual content created on the fly
• Player-facing AI chatbots or companions

The key characteristic: the player experiences AI output that wasn't individually reviewed by the developer. The developer built the system, set the guardrails, but can't guarantee every specific output.

Disclosure requirement: More substantial. You need to specify which AI models are used, describe the guardrails and content filtering in place, explain what types of content are generated live, and acknowledge that outputs may vary. This information appears in a dedicated "AI Content" section on your store page.

Review impact: Moderate. Valve may request additional documentation about your content filtering approach. Games with live-generated text content face the most scrutiny, particularly around hate speech, explicit content, and content that could be mistaken for real information. Expect 1-2 additional review cycles.

Why the Distinction Matters

From Valve's perspective, the risk profile is completely different. A game that used Midjourney to create concept art references (pre-generated, Tier 1) poses zero ongoing content risk. A game with a live LLM-powered NPC that can say anything to any player (live-generated, Tier 2) poses meaningful content moderation challenges.

The two-tier system lets Valve apply proportional scrutiny. It also gives players clearer information — a Tier 1 disclosure tells them "AI helped make this game," while a Tier 2 disclosure tells them "AI is actively generating content while you play."

The Efficiency Tool Exemption

This is the part most relevant to indie developers, and it's the part that's caused the most confusion.

The new framework explicitly exempts developer efficiency tools from disclosure requirements. The language in Valve's documentation reads:

Tools used by developers to accelerate their workflow — including but not limited to AI-powered code editors, build automation, asset pipeline tools, and editor scripting — do not constitute AI-generated content and do not require disclosure.

Let's be precise about what this means and what it doesn't.

What's Exempt

AI coding assistants. If you use Claude Code, Cursor, GitHub Copilot, or similar tools to write game code, that code is not considered AI-generated content. The reasoning is sound: the code is a developer tool, not player-facing content. Players interact with the behavior the code produces, not the code itself. Nobody discloses which IDE they used.

Editor automation and scripting. MCP servers that automate editor tasks — placing actors, configuring materials, setting up lighting, managing assets — are developer efficiency tools. The Unreal MCP Server, the Blender MCP Server, and similar tools that connect AI assistants to editors fall squarely in this category. You're using AI to operate the tools faster, not to generate player-facing content.

Build and pipeline tools. CI/CD systems with AI optimization, automated LOD generation, texture compression tools with AI-driven quality settings, automated testing frameworks — all exempt.

Procedural generation with non-AI algorithms. Traditional PCG using noise functions, rule-based systems, L-systems, wave function collapse, and similar algorithmic approaches was never covered by AI disclosure, and the new framework makes this explicit. Tools like the Procedural Placement Tool that use rule-based scatter algorithms aren't AI in the disclosure sense.

What's Not Exempt

The exemption doesn't cover tools whose output is directly player-facing AI-generated content. If you use an AI image generator to create your game's textures, those textures are player-facing content — the tool generated them, not you. The fact that the tool is "efficient" doesn't make the output exempt.

The test is straightforward: Does the player directly experience the AI-generated output? If yes, it needs disclosure. If the AI output is an intermediate step in your development process that you review, modify, and approve, it's Tier 1. If the AI output is generated live and served directly to the player, it's Tier 2.

The Gray Areas

There are genuinely ambiguous cases, and Valve hasn't addressed all of them:

• AI-upscaled textures. If you take a hand-painted 512x512 texture and AI-upscale it to 4K, is the result AI-generated? Valve's current guidance suggests this is Tier 1 if the base content was human-created, but the line isn't perfectly clear.

• AI-assisted animation. Motion capture cleaned up by AI tools, or keyframe animation interpolated by AI models — the human created the performance, but AI refined it. Most developers are treating this as exempt or Tier 1 with disclosure.

• AI-generated placeholder content that ships. You generated temp dialogue with an LLM intending to replace it, but ran out of time. It's now in the shipped game. This is Tier 1 AI-generated content and needs disclosure, even if you didn't plan it that way.

When in doubt, disclose. There's no penalty for over-disclosure, and Valve has explicitly said they'd rather developers err on the side of transparency.

Decision Flowchart: Does Your Game Need AI Disclosure?

Here's how to think through it step by step:

Step 1: Did you use any AI tools during development?

If no, you're done. No disclosure needed.

If yes, continue.

Step 2: Were any AI tools used to generate content that players directly experience?

"Content that players directly experience" means art, text, audio, music, video, or interactive elements that appear in the game. Code doesn't count. Build tools don't count. Editor automation doesn't count.

If no — you only used AI for coding, automation, pipeline tools, or internal workflows — you're exempt. No disclosure needed.

If yes, continue.

Step 3: Is the AI-generated content created during development (pre-generated) or at runtime (live-generated)?

If pre-generated: You need Tier 1 disclosure. Check the appropriate boxes during store page setup, describe which content categories used AI, and you're done.

If live-generated: You need Tier 2 disclosure. Prepare documentation about your AI models, content filtering, and guardrails. Budget extra time for the store page review.

If both: Disclose both tiers separately. Many games will have some pre-generated AI content (concept art references, initial dialogue drafts) and live-generated content (dynamic NPC conversations). Each category gets its own disclosure.

Step 4: Review your disclosure before submission

Read your disclosure from the perspective of a player who is suspicious of AI in games. Does it sound honest and specific? Or does it sound like you're hiding something? Players respond well to transparency and poorly to vagueness.

Good: "Character portraits were created using Stable Diffusion and hand-edited for consistency. NPC dialogue was written by our team with AI-assisted grammar and localization checks."

Bad: "Some AI tools were used during development."

The Gameslop Problem and Why Disclosure Alone Doesn't Solve It

We wrote extensively about the gameslop problem earlier this month, and the new disclosure framework doesn't change the fundamental issue: Steam is flooded with low-effort, AI-generated games that are choking discovery for everyone.

The disclosure system helps players make informed decisions, but it doesn't prevent gameslop from being published. Valve's philosophy remains largely hands-off — they'll label content but won't reject games for being AI-generated, as long as the disclosure is accurate.

This means the discovery problem is still your problem to solve as an indie developer.

The Numbers Are Getting Worse

As of March 2026, over 7,300 games on Steam have disclosed AI content. The actual number using AI is almost certainly higher — many developers either don't know they need to disclose or are choosing not to (which is risky, as Valve has started enforcing with page removals).

More games means more noise. More noise means less visibility for any individual title. The average indie game on Steam already struggled for attention before the AI flood. Now it's harder.

What Gameslop Looks Like in Practice

The tell-tale signs haven't changed: AI-generated store page art that looks polished until you play the game, asset-flipped environments with inconsistent art direction, dialogue that reads like unedited LLM output, and gameplay loops that feel procedurally assembled rather than designed.

The new disclosure system actually makes it easier to spot gameslop — games with extensive Tier 1 disclosure but no evidence of human creative direction are red flags for review-savvy players.

Strategies for Standing Out in a Flooded Marketplace

If you're a developer who uses AI responsibly as part of a genuine creative process, here's how to differentiate yourself from the gameslop flood.

Lead with Your Creative Vision

The single most important differentiator is design intent. Gameslop has no point of view. Your game should have one, and it should be visible in every element — art direction, gameplay design, narrative voice, audio design.

When you talk about your game on store pages, devlogs, and social media, lead with what makes it unique as a creative work. Not "we used AI to make it faster" but "here's the specific experience we're building and why it matters to us."

Be Transparently Specific About AI Usage

Counter-intuitively, the best way to handle AI disclosure in 2026 is to be more transparent than required. Players who are skeptical of AI respond well to developers who explain exactly how they used it and why.

"We used AI coding assistants to prototype gameplay systems faster, freeing up time to focus on handcrafted level design" is a message that resonates. It positions AI as a means to a creative end, not the end itself.

Invest in the Things AI Can't Do Well

AI is good at generating individual assets. It's bad at creative coherence — making sure every element of a game feels like it belongs in the same world, serves the same design goals, and creates a unified player experience.

Focus your marketing and development effort on:

• Art direction consistency. A game with a strong, consistent visual identity stands out immediately against AI-generated visual noise.
• Gameplay feel. The moment-to-moment experience of playing your game — controls, feedback, pacing — is something AI can't replicate. Polish it obsessively.
• Narrative voice. If your game has writing, give it a distinctive voice. Players can tell the difference between authored prose and LLM output, even if they can't articulate how.
• Community engagement. Gameslop developers don't do devlogs. They don't engage in Discord. They don't stream development. Being visibly human and passionate about your project is the strongest possible anti-gameslop signal.

Use AI to Compete on Quality, Not Volume

The gameslop model is about volume — ship more, faster, cheaper. The anti-gameslop model is about quality — use AI to make your one game better, not to make ten games that are all mediocre.

AI-powered developer tools genuinely help here. Using the Unreal MCP Server to automate tedious editor tasks means more time for creative decisions. Using the Blender MCP Server to speed up repetitive modeling setup means more time for art direction. Using the Blueprint Template Library for foundational gameplay systems means more time for unique mechanics.

The key distinction: these tools accelerate your creative process. They don't replace it.

Optimize Your Store Page for Trust

Practical tips for your Steam store page in the current climate:

• Use real gameplay footage. Not renders, not AI-generated marketing art, not concept art. Actual gameplay. This is the strongest trust signal available.
• Show development process. A short "making of" section or linked devlog demonstrates human involvement better than any disclosure statement.
• Be specific in your description. Vague marketing copy reads like AI-generated text. Specific, opinionated descriptions of your game's mechanics and design philosophy read like a human who cares about their work.
• Leverage demo builds. Steam Next Fest demos are more valuable than ever. A playable demo is the ultimate proof that your game is a real, designed experience.

Common Questions from Indie Developers

"I used GitHub Copilot for all my code. Do I need to disclose?"

No. Code generated by AI coding assistants is explicitly exempt under the efficiency tool clause. This applies to all AI coding tools — Copilot, Claude Code, Cursor, Cody, and similar assistants.

"I used Midjourney for concept art, then hand-painted the final assets."

No disclosure needed. The concept art was a development reference, not player-facing content. Your final assets are hand-painted. This is equivalent to using a reference photo — the reference isn't the content.

If Midjourney output appears anywhere in the shipped game (loading screens, art book DLC, marketing materials on the store page), that specific usage needs Tier 1 disclosure.

"My game has an AI-powered NPC that generates dialogue."

This is Tier 2 live-generated content. You need full disclosure including the model used, content filtering approach, and clear player-facing indication that the dialogue is AI-generated.

"I used an MCP server to automate level building in the editor."

No disclosure needed. MCP servers are developer efficiency tools. They automate editor operations — placing actors, configuring materials, setting up scenes. The player experiences a designed level, not AI output.

"I'm worried that disclosing AI will hurt my sales."

It might, in the current climate. The 85% negative sentiment figure from Quantic Foundry is real. But not disclosing when you should is worse — Valve has started pulling store pages for inaccurate disclosure, and community backlash from a disclosure scandal is far more damaging than upfront transparency.

The developers seeing the least negative impact from disclosure are the ones who pair it with clear evidence of human creative direction. Show your work, be transparent, and let the quality speak for itself.

"What if I'm not sure whether something counts?"

Disclose it as Tier 1. Over-disclosure has no penalty and signals good faith. Under-disclosure is risky. When in doubt, be transparent.

Looking Forward: What's Likely to Change

Valve's disclosure framework will continue evolving. Based on the trajectory so far, here's what we expect:

More granular categories. The current "art, text, audio, code" categories are broad. Expect subcategories — texture generation vs. concept reference, primary dialogue vs. flavor text, original music vs. adaptive soundtrack layers.

Automated detection. Valve is likely investing in AI content detection tools. Not to punish developers, but to catch games that should be disclosing but aren't. If you're tempted to skip disclosure, assume you'll eventually be caught.

Quality signals beyond disclosure. The long-term solution to gameslop isn't better labeling — it's better curation. Expect Steam to invest more in algorithmic quality signals that downrank games showing gameslop characteristics, regardless of their disclosure status.

Industry-wide standards. Steam isn't the only platform dealing with this. PlayStation, Xbox, and Nintendo are all developing their own frameworks. An industry-wide standard is likely within 2-3 years. Developers who build transparent AI workflows now will be well-positioned for whatever comes.

Practical Checklist: Before You Submit Your Store Page

Here's a condensed checklist to walk through before you submit your Steam store page for review.

Inventory Your AI Usage

Go through your entire development process and list every AI tool you used:

• AI coding assistants (Copilot, Claude Code, Cursor, etc.)
• AI image generators (Midjourney, Stable Diffusion, DALL-E, etc.)
• AI text generators (ChatGPT, Claude, etc. for dialogue, descriptions, lore)
• AI audio generators (Suno, Udio, Soundraw, etc.)
• AI 3D model generators (Tripo, Meshy, etc.)
• MCP servers for editor automation
• AI-powered upscaling or enhancement tools
• AI-based animation or motion capture cleanup
• Runtime AI systems (NPC dialogue, procedural narrative, dynamic content)

Classify Each Usage

For each item on your list, determine:

1. Is it exempt as an efficiency tool? (coding assistants, editor automation, build tools)
2. If not exempt, is the output pre-generated (Tier 1) or live-generated (Tier 2)?
3. If Tier 1, was the output reviewed and edited by a human?
4. If Tier 2, what guardrails and content filtering are in place?

Write Your Disclosure

Draft disclosure text that is:

• Specific — name the content categories affected, not just "some AI was used"
• Honest — don't minimize or obscure significant AI usage
• Contextual — explain how AI fit into the human creative process
• Complete — cover all non-exempt AI usage, don't selectively disclose

Review for Completeness

• All Tier 1 AI content is disclosed with category descriptions
• All Tier 2 AI content is disclosed with model, guardrail, and filtering details
• No exempt tools are unnecessarily disclosed (avoid confusion)
• Disclosure text reads clearly from a player's perspective
• Store page screenshots and videos show actual gameplay (not AI-generated marketing art)

Submit and Monitor

After submission, monitor your store page review for any Valve follow-up questions. Respond promptly and transparently. If Valve asks for additional detail about your AI usage, provide it — they're not trying to penalize you, they're ensuring the disclosure is complete for players.

The Bottom Line

The new disclosure rules are mostly good news for indie developers who use AI responsibly. The efficiency tool exemption means your development workflow isn't scrutinized. The two-tier system means proportional requirements based on actual player impact. And the increased transparency helps differentiate genuine games from gameslop.

The bad news is that disclosure alone doesn't solve discovery. The gameslop flood continues, and standing out requires more than compliance — it requires visible creative intent, community engagement, and relentless focus on quality.

Use AI tools to work faster and better. Use the Unreal MCP Server and Blender MCP Server to automate the tedious parts of development. Use the Cinematic Spline Tool to create trailers that showcase real gameplay. Use the Blueprint Template Library to skip months of boilerplate system building.

But never lose sight of the creative vision that makes your game worth playing. That's what AI can't generate, and it's what players are desperately looking for in a marketplace full of noise.