App Store Optimization skill for Claude Code

Done well means each store listing targets a deliberate set of terms the app can actually rank for, every character-limited field is spent without repetition, the first three screenshots sell the core outcome to someone who will never scroll, and every change is measured one lever at a time in the store’s own analytics or experiment tools.

Ground rules

• The two stores index differently. Treat them as separate projects that share a keyword list, never as one listing copied twice.
• Search ranking follows downloads, retention, and ratings as much as metadata. Metadata gets you considered; conversion and quality keep you ranked.
• Change one lever per release window (metadata, then screenshots, then icon) and wait 2-4 weeks before judging, or you cannot attribute anything.
• Ratings come only from the official prompt APIs and genuine users. No incentives, no pre-prompts that route happy users to the store and unhappy users elsewhere.
• Do not put competitor brand names in any metadata field; both stores reject or penalize it.

Field reference

Field	App Store	Google Play	Indexed for search
App name / title	30 characters	30 characters	Both, highest weight
Subtitle / short description	Subtitle, 30 characters	Short description, 80 characters	Both
Keyword field	100 bytes, comma separated, hidden	None	Apple only
Promotional text	170 characters, editable without a release	None	No
Long description	4,000 characters	4,000 characters	Apple: no. Play: yes
Developer name	Yes	Yes	Both, weakly
In-app purchase names	Promoted IAP display names	Product names visible in listing	Apple: yes
Screenshots	Up to 10 per device size	2-8 per device type	No (conversion only)
Video	Up to 3 app previews, 15-30 seconds each	One YouTube video	No
Feature graphic	None	1024 x 500, required	No

Apple name, subtitle, and keyword changes ship only with a new app version submission; promotional text changes anytime. Play listing changes go through review but need no build.

Workflow

- [ ] 1. Baseline: impressions, page views, conversion, top search terms, ratings
- [ ] 2. Keyword research and target selection per store and locale
- [ ] 3. Write metadata per store within limits, no wasted characters
- [ ] 4. Storyboard screenshots and preview video around the core outcome
- [ ] 5. Fix the ratings pipeline: prompt timing, responses, bug-driven themes
- [ ] 6. Localize the highest-value markets
- [ ] 7. Run listing experiments, one variable each
- [ ] 8. Report and set the next release's lever

1. Baseline

• App Store Connect analytics: impressions, product page views, conversion rate, and downloads split by source (App Store search, browse, web referrer, app referrer).
• Play Console: store listing visitors and acquisitions by traffic source, plus the search terms report showing visitors and installs per query. It is the best free keyword data either store offers; export it first.
• Apple Ads (formerly Search Ads) keyword planning shows a popularity score per term for the App Store; use it even without running ads.
• Current rank for 20-50 candidate terms per store and country from an ASO tool or manual searches on a clean device.
• Rating average, rating count in the last 90 days, and the top five complaint themes.

2. Choose target keywords

Score each candidate term:

Factor	Question	Weight
Relevance	Would a searcher for this term be satisfied by the app’s core feature?	Gate: drop if no
Volume	Popularity score or Play search-term visitors	High
Difficulty	Are the top 10 results giants with hundreds of thousands of ratings?	High
Current rank	Already ranking 10-30 means metadata can move it	Medium
Intent	Problem terms (“budget planner”) convert better than generic ones (“money”)	Medium

Pick per store: 1-2 primary terms for the title, 3-6 secondary for subtitle or short description, and the long tail for Apple’s keyword field or Play’s description. A small app should target terms where the current top results have modest rating counts; ranking 3rd for a mid-volume term beats ranking 60th for a head term.

3. Write metadata

App Store

• Name: brand plus the primary term, read naturally (“Tally: Budget Planner”). Never a keyword list; review rejects it and users skip it.
• Subtitle: the outcome plus secondary terms (“Track spending and save monthly”).
• Keyword field: single words, comma separated with no spaces, all 100 bytes used. Apple combines words across name, subtitle, and keyword field, so never repeat a word already in the name or subtitle, skip plurals of words already present, skip “app” and your category name, and prefer single words over phrases. Non-Latin scripts use more bytes per character.
• Description: written for humans who scroll; not indexed. First three lines appear before “more”, so lead with the benefit.
• Promotional text: seasonal offers, new features; refresh it without a release.

Google Play

• Title: brand plus primary term within 30 characters. Play policy bans terms like “best”, “#1”, “free”, or “top” in titles and icons, plus emoji and all caps unless part of the brand.
• Short description: the outcome and two secondary terms, written as a sentence.
• Full description: indexed, so include primary and secondary terms naturally, two or three times each across 4,000 characters, in short scannable sections. Repetition beyond that reads as stuffing and can trigger a metadata policy rejection.

4. Screenshots and video

Most users decide from the first screen of results, where only the first two or three screenshots (or a preview video) show. Storyboard before designing:

1. Screenshot 1: the core outcome in one line of caption over the real screen that delivers it.
2. Screenshot 2: the main differentiator versus the top competing apps for the primary term.
3. Screenshot 3: the second most requested use case or social proof (award, rating count).
4. Screenshots 4-8: remaining features, one per frame, one short caption each.

Rules: captions of 3-6 words readable at thumbnail size, real UI (not mockups of features that do not exist), consistent visual style, dark-mode or device-frame choices tested rather than assumed. Localize captions, not just metadata. A preview video autoplays muted in Apple search results, so the first 3 seconds must show the product in use without relying on audio.

5. Ratings and reviews

• Apple: requestReview through StoreKit; the system shows it at most three times per user in 365 days and decides whether to display it. Custom review prompts are not allowed.
• Google Play: the In-App Review API; quotas are undisclosed and the dialog may not appear, so never attach it to a button that promises a review.
• Trigger after a success moment (task completed, streak reached, third successful session), never on launch, during onboarding, or after an error.
• Apple lets you reset the summary rating when you release a version; use it only after a release that fixes the cause of a bad rating period.
• Reply to negative reviews within 2 days with a specific fix or workaround; users can update their rating after a reply. Feed recurring complaint themes into the roadmap; fixing the top theme usually lifts the rating more than any prompt tactic.

6. Localization

Localize by market value: downloads and revenue potential, not language count. For each target locale, redo keyword research in that language (direct translation misses how people search), localize screenshot captions, and check the field fits after translation. Some App Store storefronts index more than one localization, which gives extra keyword space; confirm Apple’s current storefront-to-localization list before relying on it.

7. Experiments and segmented pages

Tool	Store	What it tests	Notes
Product Page Optimization	App Store	Icon, screenshots, previews; up to 3 treatments vs original	Runs up to 90 days; icon variants must ship in the binary
Custom product pages	App Store	Alternate screenshots, previews, promo text per audience	Linkable from ads and campaigns, and since 2025 can be assigned keywords for organic search; Apple raised the page cap in 2025, check App Store Connect
Store listing experiments	Google Play	Icon, graphics, short and full description	Up to 3 variants vs current; localized or default listing
Custom store listings	Google Play	Listing per country, pre-registration, or deep-link audience	Use for markets with different positioning

Experiment discipline: one variable per test, pick the metric before starting (first-time installers or retained installers, not page views), run at least 7 days to cover weekday cycles, and do not call a winner below the tool’s confidence threshold. Small apps often lack traffic for significance; then run sequential before/after changes with a 2-week window each and compare conversion, not downloads.

In-app events (Apple) and promotional content (Play) appear in search and browse; schedule them around real updates or seasonal moments, a few per quarter.

Gotchas

• Repeating the brand name in the keyword field, or words already in the name, wastes bytes.
• Ranking jumps right after a release can come from the update itself; wait before crediting metadata.
• Changing the app name loses brand-search familiarity; keep the brand token stable and change only the descriptive part.
• A new rating prompt shipped to all users at once can surface pent-up negative ratings. Roll out gradually and watch the average daily.
• Screenshots showing features gated behind paywalls without saying so raise refund rates and review complaints.
• Category choice affects charts and competitors in browse; switching it is a positioning decision, not a quick ranking hack.

Verification

• Every field fits its limit in every locale (count bytes for Apple’s keyword field).
• No word repeats across Apple name, subtitle, and keyword field.
• After the release window, compare per-store: impressions from search, conversion rate, and rank for the target terms against the baseline from step 1.
• Experiment results recorded with variant, metric, duration, and confidence; losing variants documented so they are not retried.

Output template

## ASO plan — <app>, <store>, <locale> (YYYY-MM-DD)

Baseline: impressions <n>, page views <n>, conversion <x%>, rating <x.x> (<n> in 90 days)

| Term | Volume signal | Current rank | Target field |
|---|---|---|---|

**Name / Title (<=30):** ...
**Subtitle (<=30) / Short description (<=80):** ...
**Keyword field (<=100 bytes):** word,word,word  (bytes used: n)
**Screenshot storyboard:** 1 ..., 2 ..., 3 ...
**Rating prompt trigger:** ...
**Next experiment:** variable, hypothesis, metric, duration