7-Day AI Storefront Disclosure Challenge - One Honest Bullet Per Day Before Q3 2026 Partner Review (2026)

Your Steam page says “AI-generated dialogue.” Your demo still ships placeholder barks from 2024. Your mobile data-safety form says “no AI.” Your build calls OpenAI on the first NPC line. Partner reviewers in Q3 2026 do not treat those as marketing problems—they score dimension 7 red on the mock audit tabletop.

This Challenges & Sprints guide is a seven-day, one-gate-per-weekday sprint that aligns storefront bullets, platform disclosure forms, in-game labels, and demo behavior before intake—not after a rejection email.

Who this is for

• 1–4 person teams shipping LLM NPCs, procedural copy, AI-assisted art, or voice synthesis
• Studios that already ran prompt registry freeze but never synced store pages
• Anyone preparing Steam + mobile disclosure intake alongside cert packets

Time: ~45–60 minutes per weekday (~6 hours total).

Why this matters now (May 2026)

1. Q3 2026 intake density — Gamescom-adjacent partner reviews, Deck Verified autumn refresh, and early holiday console windows overlap; FAQ/demo mismatches bounce faster than crash bugs.
2. Platform disclosure hardening — Steam AI content fields, mobile data-safety questionnaires, and Quest mirrors now expect consistent answers across surfaces.
3. Mock audit scoring — Dimension 7 treats disclosure like governance, not legal footnotes—teams without a sprint discover gaps in ninety-minute table tops.

Direct answer: Run one gate per weekday; each gate produces a signed row in release-evidence/06-governance/ai-disclosure-2026/; do not upload partner packets until Sunday evidence zip passes internal read-aloud.

Before you start — readiness checklist

• [ ] Demo build ID frozen for the week
• [ ] List of every AI touchpoint (dialogue, art, voice, moderation, analytics)
• [ ] Steamworks admin + mobile console access for disclosure forms
• [ ] release-evidence/06-governance/ai-disclosure-2026/ folder created
• [ ] Legal counsel on speed-dial for wording—not for delaying gates
• [ ] Agreement: no new AI features during the challenge week

Rules of the challenge

1. One owner per gate (compliance hat rotates if solo—still sign two distinct reviews where possible).
2. Honest bullets only—if the demo cannot prove it, label planned or remove the claim.
3. Log artifacts daily; no batching proof on Sunday night.
4. Pair with wishlist truth audit only after this week if copy overlaps.
5. Add results to operating review Block 4.

How this differs from the prompt registry freeze sprint

The 14-day prompt registry freeze governs engineering change control for live prompts. This seven-day challenge governs player-facing honesty across every surface that mentions AI. You need both when shipping LLM features: registry first or in parallel, disclosure sprint before partner upload.

Day 1 (Monday) — Inventory every AI surface

Goal: One spreadsheet row per AI touchpoint.

Include: Steam long description, About section, trailer supers, in-game settings text, EULA, mobile data safety, Discord FAQ pin, press kit PDF.

Pass criteria: No orphan claims (every row has demo_proof or planned with date).

Artifact: day1-inventory.csv committed to evidence folder.

Day 1 facilitator script

“Read every Steam bullet aloud. For each, demo owner demonstrates or says planned with month. If neither, row is red until copy changes.”

Day 1 deficiency example

DEF-AI-2026-05-19-01
build: RC-2026-05-18
issue: steam_long_feature_4 claims procedural quests
acceptance: bullet removed OR demo shows procedural quest gen on RC-2026-05-18

Day 2 (Tuesday) — Steam AI content disclosure fields

Goal: Steam backend fields match inventory rows.

Steps:

1. Open Steamworks AI Content Disclosure (and related content survey fields).
2. For each feature toggled yes, link to a demo path or mark not in demo explicitly in internal notes.
3. Screenshot each saved page with build ID in filename.
4. Cross-check Steam + Play intake checklist.

Pass criteria: Zero “yes” without demo proof or planned ticket.

Artifact: day2-steam-disclosure-screenshots/ + day2-diff.md listing changes.

Steam AI fields change more often than store page copy—screenshot saved confirmation pages, not half-filled drafts. Filename pattern: steam-ai-disclosure_RC-2026-05-18_page2.png.

Day 3 (Wednesday) — Mobile and console questionnaire parity

Goal: Google Play, App Store, Quest, or Xbox forms do not contradict Steam.

Workflow:

1. Export last submitted answers (PDF or screenshot).
2. Diff against Monday inventory—flag contradictions in red.
3. Update forms or downgrade storefront claims until aligned.
4. Note any regional differences (EU vs US) in regional-notes.md.

Pass criteria: No row where Steam says “uses AI” and mobile says “no AI” without documented scope split (e.g., PC-only LLM).

Artifact: day3-mobile-parity.pdf + contradiction log.

Day 4 (Thursday) — In-game labels and settings honesty

Goal: First-session player sees the same truth as the store page.

Checklist:

• [ ] Settings → AI / Data screen lists providers actually called in demo
• [ ] Offline mode copy matches LLM fallback resource behavior if advertised
• [ ] Credits / legal scroll includes generation tools used in build
• [ ] No debug menu exposes hidden AI calls not disclosed

Record 60-second screen capture of settings + one NPC interaction.

Pass criteria: External reviewer (not implementer) confirms labels match observed behavior.

Artifact: day4-ingame-capture.mp4 + day4-readback.md.

Day 5 (Friday) — Prompt registry and human sign-off linkage

Goal: Disclosure rows point to frozen prompt versions.

1. Open prompt registry freeze artifacts.
2. Map prompt_id → storefront claims (e.g., “dynamic dialogue” → npc-bark-v3).
3. Attach human sign-off names + dates for any prompt used in demo.
4. If sign-off missing, downgrade storefront bullet to planned.

Pass criteria: Every live LLM call in demo has prompt_id + signer in evidence folder.

Artifact: day5-prompt-map.json + sign-off PDFs.

Day 6 (Saturday) — FAQ and support macro alignment

Goal: Public FAQ matches dimension 7 mock audit read-aloud.

1. Pick three FAQ bullets players actually read (not lawyer walls).
2. Read aloud while demo runs—mark true/false per bullet.
3. Update FAQ, Steam Q&A prep, and support macros together.
4. Link AI patch notes workflow if changelog promises AI changes.

Pass criteria: All three bullets true on frozen demo.

Artifact: day6-faq-readback.md with timestamps.

Day 7 (Sunday) — Evidence zip and tabletop preview

Goal: Internal mock audit dimension 7 dry run.

1. Zip release-evidence/06-governance/ai-disclosure-2026/.
2. Run 15-minute dimension 7 script from mock audit guide.
3. File deficiency tickets for any red.
4. Update operating review Block 4: ai_disclosure_challenge: pass | fail.

Pass criteria: Dimension 7 score ≥ 4/5 with no waiver required.

Artifact: day7-score-sheet.md + zip manifest hashes.

Refusal-to-stack rule

If Day 2 fails (Steam fields lie), do not proceed to mobile forms—fix Steam first. Stacking green checks on a false base wastes partner calendar.

Minimum tool stack (all free tiers acceptable)

Common mistakes

1. Legal-only review — Counsel polishes words nobody reads; reviewers read FAQ bullets.
2. Trailer-only disclosure — Supers do not satisfy in-game label requirements.
3. Hiding cloud calls — “Local AI” marketing with silent API fallback fails dimension 7 instantly.
4. Batching screenshots Sunday — Reviewers spot stale build IDs.
5. Skipping mobile when PC-only — Partners cross-check packets you already submitted last quarter.
6. New AI feature mid-week — Invalidates all gates; restart Monday.
7. No second reviewer — Solo self-deception is a known failure mode in tabletop drills.

Pro tips

• Mirror inventory columns in Asia-EU handoff receipts when nightly builds touch prompts.
• Use planned labels with month specificity (“LLM barks — September 2026 RC”) instead of vague “coming soon.”
• Keep disclosure screenshots at 1080p—illegible captures fail diligence.
• Run truth audit the following week if marketing copy repeats AI claims.
• Attach LLM NPC resource list to Day 5 for fallback net proof.

Pairing with cert and publisher workflows

Narrative walkthrough (compressed)

Monday: Team discovers Steam long description still promises “fully voiced AI companions” while demo uses recorded lines only—inventory row flagged demo_proof: no.

Tuesday: Steam AI fields corrected; screenshot saved with RC-2026-05-18 build ID.

Wednesday: Play data safety updated from “no AI” to “AI-generated text, optional cloud”—matches PC demo.

Thursday: Settings screen now names provider; capture proves offline fallback works.

Friday: npc-bark-v3 linked to signed prompt registry entry.

Saturday: FAQ bullets edited; read-aloud passes with two-person sign-off.

Sunday: Dimension 7 tabletop scores 4/5; one waiver drafted for art gen not yet in demo with signed roadmap.

That week prevented an intake loop—not a lawyer bill.

Scaling for larger teams

Day-by-day time budget (realistic)

Block calendars now—not “we will catch up Sunday.”

Inventory starter rows (copy into CSV)

surface,claim,demo_proof,data_sent_external,owner
steam_short_bullet_2,AI dialogue,yes,openai_api,@alice
steam_long_feature_4,procedural quests,no,none,@alice
trailer_super_00_08,AI voices,planned,elevenlabs,@bob
ingame_settings_ai,uses cloud LLM,yes,openai_api,@alice
play_data_safety_ai_text,yes,openai_api,@carol
discord_faq_pin_3,local only,no,none,@bob
press_kit_page_2,AI art gen,yes,midjourney_api,@dana

Extend until every public sentence about AI has a row.

Engine and middleware notes

Unity 6.6 + LLM plugins

• Search project for UnityWebRequest, vendor SDKs, and ScriptableObject prompt assets.
• List each in inventory with scene path proof.
• If Input preflight week overlaps, do not confuse controller work with disclosure work—separate evidence folders.

Godot 4.5 + HTTP dialogue

• Grep HTTPRequest and third-party AI addons.
• Capture export preset name on Day 4 video—web demos need extra cloud disclosure.
• Link Godot web export tutorial if browser build is the demo SKU.

Unreal + OnlineSubsystem

• Blueprint nodes calling remote inference must appear in Day 1 inventory even if hidden behind dev flags.
• Disable -dev shortcuts before Day 4 capture.

Moderation and safety disclosure

If you use OpenAI Moderation, Perspective API, or local filters, add rows:

Partners increasingly ask what happens to player prompts—answer in FAQ bullet 2 on Day 6.

Offline and fallback honesty

Fallback net resources are not optional branding if you advertise “playable offline.” Day 4 must show:

1. Airplane mode or firewall block.
2. Game still delivers acceptable NPC lines via local template or on-device model.
3. Settings text describes degraded mode honestly.

If fallback is not built, Steam must not say offline AI dialogue.

Evidence folder layout

release-evidence/06-governance/ai-disclosure-2026/
  day1-inventory.csv
  day2-steam-disclosure-screenshots/
  day3-mobile-parity.pdf
  day4-ingame-capture.mp4
  day5-prompt-map.json
  day6-faq-readback.md
  day7-score-sheet.md
  manifest.json

Hash the folder on Sunday before upload to partner portals.

Cold-open drill (five minutes)

Facilitator opens random Steam bullet from last month: “Prove this in today’s demo in sixty seconds.” If the team opens settings first instead of playing, your labels are not player-visible enough—fix Day 4 before intake.

When to waive vs fix

Unsigned “we will fix copy later” waivers fail mock audit dimension 2.

After the challenge — maintenance

Add to Friday Block 5:

• Grep release-evidence/06-governance/ai-disclosure-2026/ for stale build IDs.
• Compare latest prompt registry hash to Day 5 map.
• One FAQ bullet read-aloud if any AI PR merged that week.

Contradictions with other marketing sprints

Metrics to track

Key takeaways

• Q3 2026 scores FAQ and AI disclosure against the demo, not your pitch deck.
• Seven days, seven gates—inventory, Steam, mobile, in-game, prompts, FAQ, evidence zip.
• Refusal-to-stack when Steam fields lie.
• Freeze demo build ID for the whole week.
• planned labels beat silent over-claims.
• Day 7 previews mock audit dimension 7.
• Pair with prompt registry freeze for live LLM titles.
• Log everything under release-evidence/06-governance/ai-disclosure-2026/.
• Mobile and Steam must not contradict without documented scope splits.
• Run before RC freeze when AI is in scope.
• Treat storefront AI honesty as a ship gate, not a legal afterthought.
• Re-run inventory after any trailer or FAQ edit.

FAQ

We have no LLM in the demo yet—skip the challenge?

If store copy mentions AI, run Days 1–2 and remove or label planned claims. If zero AI anywhere, skip—but verify with inventory anyway.

Does this replace legal review?

No. Legal polishes; this sprint proves demo parity reviewers actually score.

Can we run parallel with wishlist truth audit?

Not the same week—copy edits collide. Do disclosure first if AI claims exist.

What about AI-generated art only?

Include art tools in inventory; Day 4 credits and Steam fields must list generators used in shipped assets.

How does this relate to Lesson 172?

Lesson 172 dimension 7 is the rubric; this challenge is the week-before habit builder.

Voice synthesis counts?

Yes—disclose providers, on-device vs cloud, and demo proof paths.

What if platform forms lack a field?

Document gap in platform-gap-notes.md and compensate with in-game label + FAQ bullet reviewers can grep.

Solo dev—who is the second reviewer?

Friend playtester, publisher contact, or future-you after 24-hour break—never same-day implementer sign-off.

Does generative fill on marketing art count?

Yes if the asset appears on the store page or trailer. Disclose tools used and whether the demo build contains the same asset pipeline output.

Conclusion

AI disclosure is no longer a checkbox at the bottom of a submission wizard—it is a cross-surface honesty system partners test in minutes. Seven disciplined days cost less than one intake rejection that forces you to re-export every governance CSV while fest clocks run. Schedule the sprint on your calendar before the next partner email—not after the first red dimension. Honest bullets are cheaper than one cert resubmission loop.

Start Monday with inventory. End Sunday with a zip your mock audit facilitator can open without asking “which build was that?”

Appendix — partner question cheat sheet

Keep this table in the Sunday zip README so reviewers stop re-asking in email threads.

Seven-day calendar invite block (copy to team)

Mon 09:00 — AI inventory (owner: compliance hat)
Tue 10:00 — Steam disclosure screenshots
Wed 10:00 — Mobile/console parity diff
Thu 14:00 — In-game capture + friend review
Fri 11:00 — Prompt registry map
Sat 10:00 — FAQ read-aloud (two reviewers)
Sun 15:00 — Evidence zip + dimension 7 tabletop

Send as recurring holds before marketing schedules trailer edits that reintroduce AI claims.

Translation and localization note

If you ship non-English store pages, Day 6 extends one day—each locale FAQ must pass read-aloud. Machine-translated bullets without human review are a common 2026 intake failure mode. Keep English inventory as source of truth; mark translated rows derived_from_en_row_12 for audit traceability.

Steam Next Fest timing

Teams targeting October 2026 Next Fest should finish this challenge by August 2026 so fest traffic never hits a page that over-promises AI features still marked planned in inventory. Pair fest prep with Next Fest calendar week that covers demo scope labels—not AI disclosure alone.

Appendix — integration with save and replay governance

Games with save fuzz testing sometimes store prompt metadata in saves. If so, add inventory rows for serialized fields and prove they match disclosed providers. Link save fuzz resource list when auditing serialization of AI state.

Overnight Asia-EU handoffs that touch prompts must set governance_exports_changed when disclosure inventory rows change—reviewers in other timezones should see diffs in BUILD_RECEIPT scope lines.