What an AI disclosure actually covers

In the 23 October selected PlayRecon ranking slice, I found an AI disclosure for 325 of 1,055 entries: 30.81%. The other 730 had no disclosure found. Those labels are useful starting points, but reading two developers’ actual statements shows why they cannot stand in for a production percentage.

For Dinghai: The Ocean Pillar, the disclosed role is localization and translation. For Hack You, it is static images and interface elements, followed by manual processing and integration. Both carry PlayRecon’s With AI flag. Their statements describe different work.

Read the role behind the flag

I rechecked the English Steam game and demo pages on 23 October. These are developer statements, not an independent audit of their production pipelines.

Dinghai: The Ocean Pillar. Its main page and separate demo page say generative AI assisted localization and translation, with the team reviewing and editing the text. They explicitly exclude AI generation for art, audio and code. The declared contribution is narrow; it does not support calling the game mostly AI-made.

Hack You. Its main page and demo page describe generated static images and UI elements, including a cryptocurrency exchange chart. The studio says it manually processed and integrated those assets, while its code, mechanics, audio and original level design were created without AI generation. How much of the interface those assets cover is unspecified. A visual contribution can matter; the disclosure alone does not quantify its breadth or importance.

Frostrail and ReBlade: The Death Spiral. I found no AI-content disclosure on their current English Frostrail and ReBlade main pages. Their linked demo store URLs redirected to the main games, so separate demo disclosure pages were not independently observable. That is an access and evidence limit—not proof of AI-free development.

These four illustrative cases were selected from the baseline’s ranks 1–100. They do not represent every rank band or the 325 flagged entries. They are production examples, not game recommendations. Under AIM’s own focus, translation alone is insufficient to establish substantial AI-assisted development; Hack You’s production extent remains unverified.

What Steam’s survey asks

Valve’s current Content Survey guidance focuses on AI-created content shipped to and consumed by players, including artwork, sound, narrative and localization. It distinguishes pre-generated content from content generated while a game runs. It explicitly separates this focus from efficiency gains through AI-powered development tools.

That makes the public disclosure a particular kind of evidence. It can describe player-facing generated content without measuring every use of development tools. Conversely, finding no public disclosure cannot establish that no AI assistance was used. Neither outcome supplies a quality score.

The 23 October count, with its denominator

The selection rule is unchanged from the 22 October baseline: PlayRecon’s top-1,000 cutoff by rolling 24-hour demo peak concurrent players, with adult-tagged entries included and every tie at the cutoff retained. At that update, 87 games share the last included displayed rank, 969, at eight concurrent players. Keeping that full tie produces 1,055 entries.

The disclosure share fell by 0.41 percentage points between these selected slices. It is not a measurement of AI use across the whole festival. Nor is it evidence that games disclosed less AI: 931 entries overlap, while 124 entered the selection and 126 left it. Membership and rank bands can change even when a developer’s disclosure does not.

Keep the same games in view, too

The separate exploratory panel freezes 12 baseline games rather than replacing dropouts. Eleven remain inside today’s selected slice. I observed the twelfth, Chinese Pancake Simulator, through its ordinary public card at the same 16:20 UTC provider update: global rank 1,056 and a rolling 24-hour peak of seven, compared with rank 642 and a peak of 14 in the baseline.

This supplies all 12 current rolling-peak observations without putting that outside-cutoff game into the 1,055-entry denominator. The separate card has no disclosure field, so its current classification remains unknown there; its baseline grouping is retained separately. The panel is exploratory, not a set of causally matched pairs.

How to use this information

For a reader, the useful next step after an AI flag is the developer’s description: what was produced, where players encounter it, what humans authored or reviewed, and what remains unspecified. For a developer, a concrete disclosure makes those distinctions easier to understand than a broad production claim.

The ranking adds a different observation: each demo’s highest simultaneous-player count over a trailing 24-hour window. It is not unique players, downloads, sales or the festival’s all-time peak. Rank movement is relative to other games. Genre, prior audience, visibility, demo age, marketing and quality can all differ between games; this comparison cannot isolate an AI advantage or penalty.

Sources and reproducible scope

Source: PlayRecon’s public native table, version 0.1.8, with All and With AI views at the same 23 October 2026, 16:20 UTC update. All was collected at 16:39:09 UTC and With AI at 16:40:07 UTC. The outside-cutoff panel card was captured separately at the same provider update. Steam disclosures and Valve guidance were rechecked on 23 October.

PlayRecon defines With AI as a disclosure on the game or its demo. Its No AI option is treated here as no disclosure found. See the provider’s Metrics and Data explanations. Followers can refresh more slowly than player data; no fresh follower-growth result is inferred. Quantitative wishlist conclusions are excluded while the chart’s stated top-3,000 coverage conflicts with displayed ranks beyond that boundary.

Download the aggregate rank-band CSV · Original chart SVG. No complete provider game catalogue is republished. This is the first daily observation after the baseline, while the festival is still running. No paid coverage.