ASO Metadata Optimizer

Craft title, subtitle, and keywords field for one locale so they both rank and convert.

Preflight: see the aso router skill.

Iron law

Never repeat a word across title, subtitle, and keywords. Apple indexes all three independently and combines their tokens into phrases: repetition burns character budget for zero extra ranking benefit. Validate every draft with aso lint before presenting it.

Step 1: Gather inputs

• Current metadata for the locale (ask the user, or read their fastlane metadata/<locale>/{name.txt,subtitle.txt,keywords.txt} if the project has one).
• A scored keyword shortlist. If you don't have one, run aso-keyword-research first: don't invent keywords here.
• Constraints: brand name, terms the user insists on keeping or dropping.

Step 2: Token budget

The keywords field limit is bytes, not characters (see aso for why non-Latin scripts cost more per byte); check aso lint's byte count rather than eyeballing character count. See aso-localization for locale-by-locale handling.

Total is roughly 22-26 unique tokens across the whole locale (fewer for non-Latin keyword fields). Every token must earn its place: it should form at least one compound phrase (with another token in the same or a different field) that has real popularity and a realistic ranking path for this app (see the difficulty ceiling table in aso-keyword-research).

Step 3: Build the compound-phrase matrix

List candidate tokens per field, then enumerate the phrases their combinations form, and score each with aso keywords:

Title tokens:    [security, camera]
Subtitle tokens: [home, monitor, baby, cam]
Keyword tokens:  [surveillance, motion, detect, night, vision, wifi, ...]

Phrases to score: "security camera", "home security", "baby cam", "home monitor", "motion detect", ...

aso keywords "security camera" "home security" "baby cam" "home monitor" "motion detect" --app <appId> -c US

Keep tokens whose phrases clear the bar; drop tokens that only form low-opportunity or irrelevant phrases. Sum the phrases' opportunity as your rough score to maximize per draft. Skip any candidate flagged brand: true, that's someone else's brand term, not a real ranking opportunity.

Step 4: Draft 2-3 full options

For each draft:

• Title: Brand - Natural Phrase (or just the brand for very well-known single-word brands). It must read as a real product name a human would say out loud, not a keyword string. Fold in your #1 phrase naturally.
• Subtitle: a genuine benefit statement, not a second keyword dump, e.g. "Home Monitor & Baby Cam", not "Camera Security Monitor Home". Connectors (&, +, -, "and") are fine; Apple strips them for indexing but they make the phrase readable.
• Keywords field: everything else, comma-separated, no spaces after commas, singular forms (Apple stems plurals automatically, "camera" already covers "cameras"), no stop words ("the", "and", "for"...), packed toward 95-100 of 100 chars. Never include competitor brand names or Apple trademarks (iPhone, FaceTime, etc.), App Review risk, not just a ranking non-issue.

Step 5: Validate every draft

aso lint --title "<title>" --subtitle "<subtitle>" --keywords "<keywords>" --locale en-US

This checks lengths, duplicate words across fields, spaces after commas, stop words, plurals, wasted characters, and competitor-brand risk, returning pass, score, issues[], and stats. Fix every issue it raises before presenting the draft: don't hand-check what aso lint already checks for you. Re-run after every edit.

Present drafts side by side with their lint score and character usage so the user picks one; don't silently pick for them unless asked to.

Step 6: Promotional text and What's New

Two more fields worth deliberate use, neither indexed for search, both purely for conversion and timely messaging:

• Promotional text (170 chars): editable anytime without a new app build or App Review wait. The safest field for time-sensitive messaging: a sale, a seasonal push, a feature just shipped. Write it, ship it, revert it, with no release cycle in the way.
• What's New (up to 4,000 chars): lead with the user-facing benefit of the update, not an engineering log. Avoid generic filler like "bug fixes and performance improvements" with nothing else, it wastes a visible slot that recently-updated-app browsers actually read.

Step 7: Screenshot captions (brief)

Apple has never confirmed that screenshot captions are read for search, so treat them as a conversion lever, not a keyword-placement opportunity. Don't tell users captions are indexed: that steers effort into image text instead of the fields that are known to rank. Write captions that lead with user benefit ("See who's at your door from anywhere"), one idea each, under ~40 characters for thumbnail readability. For full icon/screenshot/preview-video/Custom-Product-Page guidance, see aso-conversion.

Step 8: Apply and log

aso never uploads metadata. To ship:

• fastlane deliver (preferred): write fastlane/metadata/<locale>/name.txt, subtitle.txt, keywords.txt, then the user runs fastlane deliver.
• Manual: paste the exact strings into App Store Connect.

Never push or submit metadata without the user's explicit approval of the final text.

Once shipped, log it so aso-tracking can measure before/after:

aso log add --app <appId> --locale en-US --field keywords --old "<old value>" --new "<new value>" --note "why"

Log title/subtitle/keywords separately if more than one field changed. See aso log list --app <appId> to review history.