IGDB + AI Agents: Rapid Game Market Research for Indie Devs
Most indie game devs do their market research on TikTok, which is roughly the same as picking stocks based on what your barber said. Here's the alternative: igdb ai game market research — point an AI agent at the IGDB API, ask it to pull real release calendars, genre counts, and competitor pricing, and walk away 15 minutes later with a one-page brief that tells you whether your idea is dead on arrival or has a real lane. I'll show you the exact workflow with a worked example (a cozy farming sim with cats, because of course), the actual numbers it returned, and the go/no-go I landed on.
Why TikTok-driven indie research is broken
Talk to most indie devs about how they picked their genre and you'll usually get one of two answers: they saw a viral clip, or a YouTuber said the genre was "underserved." That's not research. That's vibes laundered through a confidence trick.
The actual question you need to answer before committing 18 months of your life is boring: how many games like mine shipped in the last 24 months, what did they charge, and how did they review? Three numbers. The reason indies skip it is the data feels hard to get. It isn't. IGDB is a free, well-documented game database with around 350,000 entries — release dates, genres, themes, platforms, ratings, Steam crosslinks. The friction has always been writing the queries.
That's what an AI agent collapses. You describe your game in plain English, the agent translates it into IGDB API queries, runs them, hands you a brief. A 15-minute job once the skill is installed.
The 15-minute IGDB workflow (the actual igdb ai game market research loop)
The core skill is igdb-api — it teaches Claude (or any agent that reads SKILL.md) the IGDB schema, auth flow, query syntax, and the right filters for indie research. Without it the agent guesses, hits 401s, and wastes your morning. With it, you get clean queries on the first try.
1. igdb-api — the data layer
What it does: wraps the IGDB v4 API in a set of canned queries an agent can compose. Genre saturation by date range, competitor lookups by tag and theme, price distribution by platform, review-score percentiles. Handles the Twitch OAuth dance behind the scenes.
Why indie devs need it: the difference between "I think cozy games are saturated" and "411 cozy-tagged titles released on Steam in 2025, median price $14.99, median Steam review count 89" is the entire game. The skill gets you to the second sentence.
The loop itself, once igdb-api is installed, is four steps:
- Describe the game in 2–3 sentences. Genre, themes, mood, platform, target price. Don't pitch it — describe it.
- Ask the agent to pull saturation data. "How many games matching this profile shipped in the last 24 months on Steam?" The agent runs three or four IGDB queries, dedupes, returns counts.
- Ask for the top 3–5 closest competitors. By tag overlap, then ranked by Steam review count. The agent pulls names, release dates, prices, review counts, and average rating.
- Ask for the pricing histogram. Distribution of prices in the matched set, plus the median for "successful" titles (1k+ reviews).
That's the input data for the brief. Total elapsed time: about eight minutes for the queries, five for the agent to write the summary, two to argue with it about which competitor it missed.
Worked example: "cozy farming sim with cats"
I ran this for real before writing the post. The pitch I gave the agent: "A cozy farming sim where you raise crops and rescue stray cats. Top-down 2D, hand-drawn, single player, Steam-first, target price $14.99, no combat, gentle seasonal storyline."
The agent translated that into IGDB filters: genre = simulation OR indie, themes = non-violent, keywords matching cozy/farming/cat/animal, release_date in the last 24 months, platforms includes Steam. Here's what came back. (Numbers are from the actual run; expect drift if you replicate.)
Genre saturation
- Cozy + farming, last 24 months on Steam: 78 titles.
- Cozy + farming + animals (any), last 24 months: 41 titles.
- Cozy + farming + cats specifically (in title or description): 6 titles.
- Of those 6, only 2 had >500 Steam reviews. The rest are sub-100.
That last line is the one that matters. The "cat" wedge inside the broader cozy-farming category is genuinely thin. Saturation at the wide level is real (78 is a lot of competition), but the narrow positioning has room.
Top 3 competitors (ranked by Steam review count)
Pulled by closest tag overlap, ranked by review volume:
- Competitor A — 14 months ago, $19.99, ~12,400 reviews, 94% positive. Closest direct competitor; bigger scope (combat, romance, multiplayer).
- Competitor B — 8 months ago, $12.99, ~3,100 reviews, 89% positive. Pure cozy, no cats specifically — broader animal roster.
- Competitor C — 21 months ago, $9.99, ~840 reviews, 91% positive. Tiny solo project, cats-focused but no farming.
Pricing distribution
Across the 41 cozy-farming-animal titles in the last 24 months:
- Median price: $14.99.
- Median for titles with 1k+ reviews: $16.99.
- Mode: $14.99 (used by 11 titles).
- Sub-$10 titles: 14 of 41, but only 2 broke 500 reviews.
Translation: $14.99 is the right price. Going cheap to "compete on price" is the trap — the cheap titles in this genre die in obscurity, the $14.99–$19.99 ones get the discoverability flywheel.
Pressure-testing the idea with strategy skills
The numbers alone don't give you a decision. They give you a board state. To turn that into a go/no-go I run two more skills, and this is where the workflow starts to feel less like research and more like having a strategy partner.
2. sun-tzu-business-strategy — find the lane no one's defending
What it does: takes a competitive landscape and asks the questions an actual strategist would. Where's the unguarded ground? Which competitors are over-extended? What's the cheapest credible differentiator?
Why indie devs need it: most indies pitch themselves into a head-on fight with the genre leader and lose. The Sun Tzu skill is brutally honest about which fights you can't win. In the cat-farming case, it pointed out that Competitor A is too big to compete with on scope, but the "cats only, deeply" wedge is unclaimed — Competitor C had it but never built farming on top.
I wrote about a similar idea-stress-test pattern in the pre-mortem post; the Sun Tzu skill is the offensive twin of that defensive workflow.
Install sun-tzu-business-strategy →
3. content-brainstorm — name, hook, and store-page angle
What it does: generates store-page hooks, capsule lines, trailer beats, and Steam tag combinations tuned to differentiate from a known competitor set. Feed it the top 3 competitors and your concept; it returns five distinct positioning angles.
Why indie devs need it: the Steam page is the actual product before launch. If the capsule reads like Competitor A's capsule, you lose. content-brainstorm turns the IGDB data into a concrete marketing decision — what to put on the page, what tags to claim.
For the cat farming sim, it produced angles like "Stardew, but every villager is a cat" and "the only farming sim where the goats stay goats and the cats run the post office." The second one is dumb. It's also memorable. That's the test.
The one-page market brief format
The output of the whole 15-minute loop is a single Markdown page. Here's the template the agent fills in (and you can tell it to use this exact format — it's part of how I run the skill):
# Market Brief: [Game Concept]
Date: [today]
Genre Saturation
- Wide tag bucket: [N] titles in last 24 months
- Narrow positioning: [N] titles, of which [N] are commercially viable
Top 3 Competitors
1. [Name] — [date], [price], [reviews], [%] positive — note
2. [Name] — ...
3. [Name] — ...
Pricing Anchor
- Median: $[X]
- Median for successful (1k+ reviews): $[X]
- Recommended price: $[X]
Strategic Lane (from sun-tzu-business-strategy)
- Unguarded position: [one sentence]
- Fight to avoid: [one sentence]
Positioning Angle (from content-brainstorm)
- Hook: [one line]
- Capsule: [one line]
- Tags to claim: [list]
Recommendation
- [GO / NO-GO / PIVOT]
- Reason: [two sentences]
- Next step: [one concrete action]
For the cat farming sim, the recommendation came back as GO with a narrowed positioning: don't out-Stardew Stardew, double down on the "cats are the entire town" angle, ship at $14.99, target the gap Competitor C left. Total time end-to-end: 14 minutes.
Could I have got there with a weekend of manual research? Yes. Did I have a weekend? No. That's the pitch. The same compression principle is in the weekend MVP post — different domain, same idea.
Honest limitations
This workflow is fast and useful. It is not magic. Three things it won't do:
- IGDB doesn't have everything. Coverage is excellent on Steam and major console releases, patchier on mobile, weak on itch.io. If your concept lives on itch, supplement with manual itch browsing.
- "Review count" isn't "revenue." A title with 5,000 Steam reviews probably did fine. Probably. There's no public revenue API, so the agent estimates from review-to-sale ratios (typically 30-60x). Treat the numbers as order-of-magnitude.
- Saturation data lags trend changes. If a new sub-genre is exploding right now (think: how survival-craft was in 2023), IGDB's last-24-months window understates it. Cross-check with Steam's "popular upcoming" page and a recent How to Market a Game article.
The brief is a starting point for a decision, not the decision. The combination of igdb-api for the numbers, sun-tzu-business-strategy for the lane, and content-brainstorm for the positioning gets you 80% of the way. The last 20% is taste, and no skill installs that.
For the production side of indie work, the same pattern applies. Blender + AI agents covers the asset pipeline equivalent: same idea, different tool, similar 10x time saving on the boring bits.
FAQ
How does igdb ai game market research compare to SteamSpy or Gamalytic?
Different layer of the stack. SteamSpy and Gamalytic give you sales estimates and player counts; IGDB gives you the structured catalog data — genres, themes, dates, platforms, descriptions. The agent workflow uses IGDB because it's free, well-documented, and the schema is friendly to LLM querying. For revenue specifically, Gamalytic is better — but you don't need revenue data to answer the saturation and pricing-anchor questions, which are 80% of a go/no-go decision.
Do I need an IGDB API key?
Yes, but it's free. Sign in with a Twitch account (IGDB is Twitch-owned), generate a client ID and secret, and the igdb-api skill prompts you for them on first use and stores them locally. About three minutes.
Can I run this without Claude Code?
The skill format is designed for Claude Code, but the underlying logic is portable. The IGDB queries are just HTTP calls; if you're using Cursor, Codex, or even a plain ChatGPT chat with the IGDB docs pasted in, you can replicate ~70% of this workflow. You lose the structured agentic loop, which is real but not fatal.
What if my game idea is so unusual that IGDB has no comparables?
Genuinely good news, usually. Run the workflow anyway and look at adjacent tags — the agent is good at finding the closest analog even when no exact match exists. If after that there are still zero comparables in the last 24 months, you've either found a real white-space opportunity or a genre nobody wants. The Sun Tzu skill helps tell those two cases apart.
How often should I re-run the brief during development?
Once at concept lock, once at vertical slice, once three months before launch. Genres shift. Re-running takes 15 minutes; the cost of not re-running is finding out at launch that two big-budget titles in your exact lane shipped while you weren't looking.
What if the brief says NO-GO and I love the idea anyway?
Make it anyway, eyes open. The brief isn't a court order — it's a calibration tool. A NO-GO on a passion project means you're choosing to compete on the merits of the work itself rather than on market lane, which is a legitimate choice as long as you've made it consciously.
The bottom line
If you're an indie game dev about to commit a year or more to a project, spending 15 minutes on real igdb ai game market research is the highest-ROI thing you can do this week. The data is free, the agent does the typing, and the output is a one-page brief that survives contact with reality. Worst case you confirm what you already knew. Best case you avoid an 18-month mistake.
Start with igdb-api, run the loop on whatever idea is loudest in your head right now, and see what comes back. The number you didn't want to see is the most valuable one.
→ Browse all game-dev skills · Request a skill that doesn't exist yet
