FAQ:

I expect there to be some decent quality visual novels that are mostly AI-written and/or mostly AI-drawn by that time, but hitting $1 mil with a visual novel is probably pretty hard.

96% of the games on Steam generate less than $1 million in sales: https://gameworldobserver.com/2022/11/29/median-indie-game-earnings-steam-barely-over-1000

So it cannot just be any random experimental indie game where the AI generation is the main draw; it has to actually be a good game that attracts a significant following. Apparently a third of the ~400 games on the Meta Quest store have earned over $1 million (https://uploadvr.com/1-in-3-meta-quest-store-apps-make-millions-in-revenue/), so perhaps the level of quality on the Quest store is a good benchmark for comparison: https://www.oculus.com/experiences/quest/section/258035155854818/

Personally, I am doubtful that a high-quality game could be developed, released, and hit $1M revenue on such a short timespan. (Less than three years from now -- we don't currently even have AI that can generate good 3D models! And I am doubtful that an experimental visual novel or text adventure is going to break $1M.) Videogame creation workflows seem pretty complex, such that even a game with one very impressive AI-generated aspect (ie, all the artwork), other aspects of the game (mechanics, etc) would need to be created by people.

I would appreciate some examples of what would cause this market to resolve to "NO", as I could see some game incorporating AI in a fairly basic / limited but high-salience way (ie, the enemies in in a stealth game are RL-trained agents that realistically search for the player, a la "Hello Neighbor", or all the textures are AI generated but then they are applied to human-made game objects), gets a bunch of overhyped headlines as "the first AI-created game!!", but doesn't truly meet the criteria that this market seems to be laying out.

After consulting with others, I have updated with a FAQ to help to remove ambiguity.

Added the related 2026 market: https://manifold.markets/SneakySly/by-the-end-of-2026-will-there-be-a

I would put high probability on a game (I would consider it a game but you might not hence the question) existing where its AI Dungeon esc with important additions. It generates images for stuff, it prohibits you from narrative breaking / forces you to follow established mechanics but everything is still in written text style, it would have conditions under which you would win or lose.

The most important mechanic is the AI forcing you to stick to rules. That, alongside resolution criteria, would in my mind be worth +$1 million and resolve yes.

AI dungeon is on steam but thats not a "game", would the above be considered a game to you? If not, what are the minimal additions needed so that it would be?

I believe an answer to this question would do a lot for providing clarity on your resolution criteria.

Created a related market since Manifold seems bullish on 2025:

Do existing procedurally-generated roguelikes count?

Does something like AI Dungeon count?