Skip to main content

Keyword Research

Generate a curated list of high-value ASO keywords for your app — clustered by sub-niche, scored by popularity and difficulty — and write the result to AiGuidelines/keywords.md. The output is designed to feed directly into Metadata Localization (Mode 1: keyword expansion).

This is a skill-driven workflow running through the Claude Code skill. It uses the Astro MCP tools when they're connected for real-time App Store data; without Astro it falls back to a manual brainstorm.

Triggers​

Using kappmaker, research keywords for AI image generator
Using kappmaker, find aso keywords around manga translation
Using kappmaker, keyword research drift coaching competitors="Driftbox, RaceChrono"
Using kappmaker, find sub-niche keywords for photo editor min_popularity=40 max_difficulty=35

Short forms like kappmaker keyword research <topic> and find aso keywords for <topic> route here too.

What it produces​

AiGuidelines/keywords.md — a single Markdown file with:

  1. Header — base keyword, generation date, filter thresholds, and competitor sources.
  2. Recommended primary keywords (top 5) — highest-value picks across all sub-niches. These are the ones to use in your iOS name and subtitle, and to front-load in keywords.txt.
  3. Sub-niche clusters — groups of related keywords with their popularity and difficulty scores, plus a one-line description of the user intent each one captures.
  4. Discarded — keywords that hit the filter cutoff, so you can sanity-check the threshold and see what was rejected.
  5. Ready-to-paste command — a one-liner you can drop straight into localize metadata mode=keyword-expansion.

Example structure:

# ASO Keyword Research

**Base keyword:** AI image generator
**Generated:** 2026-05-13
**Filters:** popularity ≥ 30, difficulty ≤ 45

## Recommended primary keywords (top 5)

| Keyword | Popularity | Difficulty | Why |
|---------|-----------:|-----------:|-----|
| ai art generator    | 72 | 28 | High volume, low competition relative to "ai photo" |
| text to image ai    | 68 | 34 | Direct search intent for the core feature |
| ...

## Sub-niche clusters

### Cluster 1 — AI text-to-image
| Keyword | Popularity | Difficulty | Description |
|---|---:|---:|---|
| ai text to image    | 65 | 38 | Main user search intent for text-to-image |
| prompt to picture   | 42 | 22 | Long-tail variant, very low competition |
| ai art from text    | 38 | 31 | Casual user phrasing |

### Cluster 2 — Style transfer
...

Arguments​

ArgumentDefaultNotes
base / base keyword(required)The anchor topic. Can be derived from AiGuidelines/prd.md / AiGuidelines/app-idea.md (or any product-description *.md under AiGuidelines/), then the project README.md, or the existing en-US name + subtitle as a last resort. The workflow confirms the inferred base with you before proceeding.
competitors(auto-discovered)Comma-separated list of competitor app names or App Store IDs. Auto-discovered via search_app_store if omitted.
min_popularity30Keywords below this are dropped from the main lists (still appear in "Discarded").
max_difficulty45Keywords above this are dropped (still appear in "Discarded").
target_count30–50Approximate size of the final filtered pool. The workflow relaxes max_difficulty by +5 once if it can't hit the target.
outputAiGuidelines/keywords.mdIf the file already exists, the workflow asks once before overwriting; on decline, writes to AiGuidelines/keywords-<date>.md.

Workflow​

  1. Discover competitors — search_app_store({ query: <base> }) to find live apps ranking for the base keyword. Top 5–10 most relevant become the competitor set (or use the explicit competitors= list).
  2. Track competitors in Astro — list_apps to check what's already tracked, then add_app for any missing ones. Free-tier quota errors are surfaced but don't abort the run.
  3. Extract competitor keywords — extract_competitors_keywords / get_app_keywords on each tracked competitor; collect every returned keyword along with its popularity and difficulty scores.
  4. Expand with AI suggestions — get_keyword_suggestions({ base_keyword: <base> }) to broaden the candidate pool with variations the competitor scan didn't surface.
  5. Filter and dedupe — apply popularity ≥ min_popularity and difficulty ≤ max_difficulty, drop case-insensitive duplicates, prefer the higher-popularity entry for near-duplicates (singular/plural).
  6. Cluster into sub-niches — group remaining keywords by semantic theme. Each cluster maps cleanly to one of the 9 indexed locales when you run Mode 1 keyword expansion next.
  7. Write AiGuidelines/keywords.md with the structure above.

Project convention: the AiGuidelines/ folder​

KAppMaker's ASO workflows treat AiGuidelines/ as the home for AI-facing planning docs:

  • AiGuidelines/prd.md — Product Requirements Document
  • AiGuidelines/app-idea.md — short app concept / value prop
  • AiGuidelines/keywords.md — output of this workflow
  • Any other *.md you keep there that describes the app's purpose, target users, or competitors

When this workflow needs to derive a base keyword from your project, it scans AiGuidelines/ first, then falls back to the project README.md, then your existing en-US ASO metadata. You don't have to use this folder — but if you do, the workflow finds your context automatically and you skip the "describe your app" prompt.

If AiGuidelines/ doesn't exist yet, the workflow creates it on first write.

Astro MCP not available?​

If Astro MCP isn't connected in your session, the workflow falls back to a manual brainstorm: it generates 30–50 sub-niche candidates from category knowledge, marks popularity / difficulty columns as ? (unknown), and adds a header warning that you need to validate the numbers on Astro, App Store Connect, AppTweak, or Sensor Tower before using the keywords in production listings.

The cluster structure and the "Recommended primary keywords" section still get filled in, so you still get the same downstream value — just without the scoring confidence.

To connect Astro MCP, follow the setup guide at tryastro.app/docs/mcp/ and add the MCP server to your Claude Code config.

The full ASO chain​

1. Using kappmaker, research keywords for <topic>
   → produces AiGuidelines/keywords.md

2. Pick 10–15 keywords from the Recommended + top-cluster entries

3. Using kappmaker, localize metadata mode=keyword-expansion keywords="<picks>"
   → writes English copy across the 9 US-indexed locales
      with each locale holding a different cluster of keywords

4. kappmaker translate-screenshots --locales ar-SA fr-FR ko pt-BR ru vi zh-CN zh-TW es-MX
   → matches the image-side to the text-side

End-to-end: research → text → images.

See also​

  • ASO Guidelines — character limits, keyword field rules, US-indexed locale strategy explained.
  • Metadata Localization — the natural next step. Feed the keywords from this workflow into Mode 1 keyword expansion.
  • Screenshot Translation — image-side ASO companion.
  • Astro MCP — the data source for popularity / difficulty scores.