The AI-Era Game-Building Playbook

The one rule

The AI revolution in game-making is, overwhelmingly, an asset-and-boilerplate revolution. It compresses the *production* of a game — art, audio, text, plumbing — by an order of magnitude, and the *design* of a game barely at all. Every strategic move in this playbook follows from that one fact.

So the rule is simple to state and hard to live by: build where AI builds, and bet where it can't. Pick a game whose content AI can generate for free, and pour your entire human budget into the ingredients AI can't author — feel, balance, netcode, taste. The sections below turn that into four concrete steps.

Everything here rests on our buildability model: 45 reusable ingredients scored for how much AI can do today, rolled up into a difficulty score for each of 30 genres. Those scores are our own editorial assessment of current tooling, not a measurement — read them as where the labour collapses, not where the craft does.

Step 1 — Pick a genre by its kitchen, not its reputation

A genre's buildability is decided by which discipline its core ingredients come from. Audio and art are the most automated kitchens; narrative is split; code scaffolds fast but tunes slowly; and feel is the hardest kitchen of all. So judge a genre by what it's *made of*, not by how it sounds.

The JRPG is the clearest example of the trap working in your favour: it has a fearsome "hard RPG" reputation, but it's really a writing/art/music genre over a simple turn-based skeleton — which is why it scores 2.77, far easier than its action-RPG cousin at 3.61. The single ingredient between them, real-time combat feel, is the whole difference.

• ✓Write your target game as a recipe: list its features, and weight each by how central it is to the experience.
• ✓Look up where each ingredient sits on the 1–5 scale — generation (art, audio, writing) is 1–2; feel, netcode and design are 4–5.
• ✓If your game's core identity rests on a 4 or 5 ingredient, accept that AI changes your cost, not your difficulty.
• ✓Distrust reputation: a genre that sounds hard but is built from 1–2 ingredients (like the JRPG) is a gift.

Step 2 — Let AI own the production layer

Once you've chosen, hand AI everything it does well — without guilt. This is the half of the game where the labour genuinely collapses, and clinging to it by hand is how solo projects die in the asset mines.

Voice acting is the single most-automated craft in games; backgrounds, concept art, textures, flavour text, branching dialogue, worldbuilding, inventory systems, save/menu shells and platform glue are all in AI's strong zone. Generate broadly, then curate ruthlessly — the human value in this layer is selection and consistency, not creation.

• ✓Voice, music, SFX and ambience: generate, don't record or licence-hunt — it's the most-automated kitchen.
• ✓2D art, backgrounds, textures and UI icons: generate sets, then enforce a single consistent style.
• ✓Branching dialogue, lore and flavour text: draft at volume, then edit for voice — the final pass is yours.
• ✓Inventory, save/settings, menus, achievements and procgen: let AI scaffold the boilerplate it has seen a thousand times.
• ✓Treat the production layer as curation: your job is which 50 of 500 to keep, not making all 500.

Step 3 — Spend your whole human budget on the craft AI can't touch

Whatever you saved in Step 2, reinvest here. Four things recur as the ceiling across every hard genre, and they're the things that decide whether a finished game is actually good: game feel, real-time netcode, level/puzzle/encounter design, and the "is it fun?" balance loop.

They share a property — each is judged in motion, against a player, not in isolation — which is exactly why no tool generates them. This is where your taste, your playtests and your iteration go. A game that nails one of these on a fully AI-built production layer beats a beautifully-produced game that feels dead.

• ✓If your genre is feel-bound (platformer, action RPG, racing): budget weeks for game-feel juice — input, animation, camera, sound, screen-shake.
• ✓If it's design-bound (puzzle, metroidvania, tower defense): hand-author the difficulty curve; AI's layouts look done and play hollow.
• ✓If it's balance-bound (deckbuilder, roguelike, management): instrument playtests and tune the numbers — this is the work that eats the schedule.
• ✓If it's netcode-bound (any competitive multiplayer): treat real-time authoritative sync as a specialist hire, not a prompt — AI helps least here.
• ✓Playtest early and often: the four ceilings only reveal themselves in motion.

Step 4 — Aim at the overlap of easy and growing

Buildability alone is a trap: everyone can build the easy genres now. The signal is the *overlap* of AI-easy and real, recent demand. Filter the catalogue for build difficulty at or below 2.7, a thousand-plus titles of proven demand, and at least 40% released since 2023, and five genres survive: interactive fiction, hidden object, clicker, walking sim and card game.

Each is a genre where AI can do most of the production *and* a live audience keeps arriving. Two honourable mentions sit just outside: farming sim (the hottest momentum of any easy genre) and the JRPG. Pick from the overlap and you start with both a low build cost and a buyer base that's still growing — the rare case where the easy choice is also the smart one.

• ✓Don't choose on buildability alone — the easiest genres are the most crowded by definition.
• ✓Cross buildability with momentum: favour genres where 40%+ of titles are recent.
• ✓Cross both with proven demand: a thousand-plus existing titles means the audience is real, not theoretical.
• ✓Shortlist from the overlap: interactive fiction, hidden object, clicker, walking sim, card game — plus farming sim and JRPG.

The contrarian play — the defensible niche

There's a smarter, harder move hiding in the data. Deckbuilding is the catalogue's fastest-growing genre — 64% of titles released since 2023 — yet it scores 3.18 to build, because its soul is *balance*, the AI-hardest content work there is.

That's not a reason to avoid it; it's the reason to choose it. Let AI generate every card and every piece of art, and spend 100% of your human effort on the one thing it can't do: making the numbers sing, the way a Slay-the-Spire-class game does. Highest demand × hardest-resisting core is the place a skilled human still wins — because AI just commoditised everyone else's production, but not your balance.

That's the general principle in miniature. As AI flattens the cost of production, value concentrates in the ingredients AI can't author. Find a genre whose content AI builds for free and whose craft AI can't touch, and put your whole budget on the craft.

What this playbook doesn't promise

Two honest limits. First, "buildable" is not "good": these scores measure where the labour collapses, not whether the result is fun. AI can generate every asset of a competent platformer in a weekend, and it will still feel dead without the human feel pass. The score tells you where to save effort, not where you're allowed to skip it.

Second, the numbers are what they are. AI-build difficulty is our editorial judgement of current tooling, and the frontiers move — 3D characters, skeletal animation and adaptive music could each ease within a year. The demand figures are banded popularity estimates from our own compiled dataset, a relative-popularity proxy, not unit sales — there is no revenue figure here and we never treat one as if there were. Use this to choose where to spend your scarce human time. That's the whole game now.

FAQ