The outline#
| # | Slide | What goes on it | The test | Limit |
|---|---|---|---|---|
| 1 | Title | Game name, genre, platforms, one-line hook | Could a stranger repeat the hook after one read? | ≤ 15 words |
| 2 | The moment | One GIF or screenshot of the best 10 seconds of play | Does it show a player action, not a vista? | Caption ≤ 12 words |
| 3 | Core loop | 30-second / 10-minute / session loops as a diagram | Can you point at the verb in each loop? | Diagram only |
| 4 | Why it needs a model | What breaks if you remove the AI | Is the answer a mechanic, not a technology? | ≤ 40 words |
| 5 | Authorship | What's hand-made: rules, tone bible, fallbacks, art | Would a sceptic believe humans are directing it? | 4 bullets |
| 6 | Comparables | 3–5 similar games, price, rough performance | Are they realistic for your team size? | Table |
| 7 | Audience | Who buys it, where they gather, what they pay | Named communities, not demographics | ≤ 50 words |
| 8 | Art direction | Target look, made by your artists | Is every image human-made or clearly labelled? | Images |
| 9 | Systems & tech | Engine, model stack, hosting, offline, safety layer | Does it name the failure modes? | Diagram + 5 bullets |
| 10 | Inference economics | Cost per player-hour, lifetime cost per player, plan to cut it | Is lifetime cost < 15% of price (premium)? | Small table |
| 11 | Team | Roles, shipped titles, gaps you'll hire for | Is someone clearly responsible for each discipline? | Photos + 1 line each |
| 12 | Roadmap | Milestones with dates and playable deliverables | Is each milestone something a publisher can play? | Timeline |
| 13 | Budget & ask | Total budget, the ask, cost per milestone | Does headcount × months × cost explain the number? | Table |
| 14 | Status & next step | What exists today, other interest, what you want from this meeting | Is the next step specific? | ≤ 40 words |
Slides 4, 5, 9 and 10 are specific to AI-native games. Leave them out for a conventional game and the rest still holds. For the reasoning behind each slide, read how to pitch a game to a publisher.
Copy-paste version#
01 TITLE
Put: Game name, genre, platforms, one-line hook
Test: Could a stranger repeat the hook after one read?
Limit: ≤ 15 words
02 THE MOMENT
Put: One GIF or screenshot of the best 10 seconds of play
Test: Does it show a player action, not a vista?
Limit: Caption ≤ 12 words
03 CORE LOOP
Put: 30-second / 10-minute / session loops as a diagram
Test: Can you point at the verb in each loop?
Limit: Diagram only
04 WHY IT NEEDS A MODEL
Put: What breaks if you remove the AI
Test: Is the answer a mechanic, not a technology?
Limit: ≤ 40 words
05 AUTHORSHIP
Put: What's hand-made: rules, tone bible, fallbacks, art
Test: Would a sceptic believe humans are directing it?
Limit: 4 bullets
06 COMPARABLES
Put: 3–5 similar games, price, rough performance
Test: Are they realistic for your team size?
Limit: Table
07 AUDIENCE
Put: Who buys it, where they gather, what they pay
Test: Named communities, not demographics
Limit: ≤ 50 words
08 ART DIRECTION
Put: Target look, made by your artists
Test: Is every image human-made or clearly labelled?
Limit: Images
09 SYSTEMS & TECH
Put: Engine, model stack, hosting, offline, safety layer
Test: Does it name the failure modes?
Limit: Diagram + 5 bullets
10 INFERENCE ECONOMICS
Put: Cost per player-hour, lifetime cost per player, plan to cut it
Test: Is lifetime cost < 15% of price (premium)?
Limit: Small table
11 TEAM
Put: Roles, shipped titles, gaps you'll hire for
Test: Is someone clearly responsible for each discipline?
Limit: Photos + 1 line each
12 ROADMAP
Put: Milestones with dates and playable deliverables
Test: Is each milestone something a publisher can play?
Limit: Timeline
13 BUDGET & ASK
Put: Total budget, the ask, cost per milestone
Test: Does headcount × months × cost explain the number?
Limit: Table
14 STATUS & NEXT STEP
Put: What exists today, other interest, what you want from this meeting
Test: Is the next step specific?
Limit: ≤ 40 wordsFilling in the inference economics slide#
This is the slide most AI-native pitches skip and the one we look at first. Fill in four numbers:
| Line | How to get it |
|---|---|
| Calls per player-hour | Log every model call in a 30-minute playtest, all systems included, ×2 |
| Tokens per call | Average input + output from the same logs |
| Cost per player-hour | Calls × tokens × your blended price; see the inference cost guide |
| Lifetime cost per player | Cost per hour × expected hours (use the median, then show the 90th percentile) |
Then one line on how you'll reduce it: caching, summarisation, routing or on-device models, with a target. Our research on caching and summary intervals gives defensible numbers to cite.
Five ways decks lose the room#
- Market-size slides. "Games is a $200B market" tells a publisher nothing about your game. Cut it.
- Feature lists. Twenty bullets says you haven't decided what the game is.
- Trailer-only evidence. A trailer proves you can edit video. Raw capture proves the game works.
- Aspirational comparables. Comparing a team of four to the year's biggest hit undermines every other number you show.
- A budget without milestones. A number with no breakdown reads as a guess.
Frequently asked questions#
How many slides should a game pitch deck be?
12–15. This outline has 14; drop the four AI-specific slides for a conventional game.
What should a game pitch deck include?
Hook, core loop, comparables, audience, art direction, tech, team, roadmap and budget, plus for AI-native games: why it needs a model, authorship, and inference economics.