Brainstorm App Store Keywords with AI (and Check What It Gives You)

• Expands one head term into dozens of variants, audiences and use cases.
• Rephrases a problem the way different users would say it.
• Suggests local-language phrasing, which you still need to check; see translate or transcreate keywords.

It has no live App Store data. It cannot say how many people search a phrase, how strong the top 10 is or whether you could rank. Apple publishes no search counts, and an assistant that quotes a number is inventing one. Treat any claim of volume from a chat as a guess.

Use a prompt like this, with your own details:

My app: a habit tracker for people with ADHD (iOS).
Audience: adults who struggle to keep routines.
List 40 phrases of 2 to 4 words that a person might type into the
App Store search bar to find an app like mine.
Group them by: problem, feature, audience.
Rules: no competitor names, no trademarks, no word shorter than
3 characters, no claims like "best" or "number one".

1. Competitor names and trademarks. Delete them. Apple states other apps' and companies' names are not allowed in the keyword list, and prohibits unauthorized trademarked terms and competing app names. See competitor names in keywords.
2. Length. Each keyword must be longer than two characters. Check non-Latin text: the field limit is 100 bytes, not characters.
3. Relevance. Remove phrases that describe features you do not have. AI often adds them.
4. Generic filler. Remove words like "app" and "best".
5. Duplicates and variants. Merge plural and singular pairs; see singular or plural.

For every phrase that survives, search the App Store and read the top 10. Does it show apps like yours? How many have large rating counts? The method is in choosing keywords you can rank for. This step is what separates a list of words from a keyword strategy.

Put the phrases that pass into the name, subtitle and keyword field, and track their rank; see what to put in the keyword field.

Assistants often return phrases that sound natural but nobody types, such as "habit formation platform". They repeat the same stem in ten variants and mix in competitor names without being asked. They may also claim a phrase has "high search volume" with no data behind it. Ask for phrases in groups, run the filter, and keep a record of which phrases you accepted and why, so your next brainstorm starts with your own history.

LocalASO uses the AI you bring: Claude Code, an Anthropic API key, Ollama, Codex or Gemini CLI. Brainstorm takes a one-sentence description, asks your AI for keywords, then checks demand and difficulty in the countries you pick. Those numbers are estimates from rating counts.

If you work in an AI tool with the MCP server, the tools are board_add_keywords to put candidates on the board, keywords_check to queue rank, volume and difficulty checks for keywords already on the board, and board_get to read the results. The AI brainstorms; LocalASO measures. Setup is in the MCP docs and AI providers docs, and the Suggestions page describes Brainstorm.