ASO audit: where the app is found, and where it is invisible

You are auditing discoverability through MetaRun. Two different battlefields, audited together: App Store SEARCH (rankings for typed keywords) and ANSWER ENGINES (whether an AI assistant names the app when asked "best X app"). They move independently; a strong rank with zero assistant visibility is a finding, not a contradiction.

Ground rules

• Tools run over MetaRun's MCP server or npx metarun run <tool> '<json>'. Fetch unfamiliar schemas with npx metarun tools <tool-name> --json.
• This mission MEASURES and DIAGNOSES. Nothing here posts, publishes, or seeds content anywhere, and metadata fixes are staged through previews the user confirms; that separation is a MetaRun product principle, state it if the user asks for astroturfing.
• Answer-engine scans (aeo_scan_answers) are METERED against the user's plan. Say so before running one; suggest, never auto-fire.

The mission

1. Search rankings, as they stand

• aso_list_tracked_keywords -> what is already watched. Nothing tracked? aso_suggest_keywords proposes terms from the listing; confirm with the user, then aso_track_keywords.
• aso_scan_ranks refreshes tracked ranks; aso_check_rank spot-checks any term/territory live. aso_keyword_history shows the trend; aso_keyword_leaders shows who owns the term.

2. Metadata quality

• aso_lint_metadata runs the rulebook over the live listing: wasted keyword characters, duplication across fields, budget violations.
• aso_cross_locale_plan finds free keyword real estate: each storefront indexes extra locales (US reads nine), so unused locale slots are extra 100-character keyword fields. Fixes stage via the localize-listing mission's preview/apply pairs.

3. Ground truth from Apple

• aso_get_search_performance returns MEASURED search vs browse impressions, product-page views, and per-territory conversion, plus a featuring detector. If it reports not_enabled, offer aso_enable_search_analytics (needs an Admin key; Apple's first reports take 1-2 days).

4. Rivals

• aso_scan_rivals diffs tracked competitors' listings against their last snapshot: renamed apps, new subtitles, screenshot count changes. A rival suddenly bidding on a keyword shows up here before it shows in ranks.

5. The second store: answer engines

• aeo_visibility_report -> current share of voice: which assistants name the app, at what position, versus which rivals.
• aeo_suggest_questions composes the buyer questions worth tracking (from category, tracked keywords, and rivals); aeo_track_questions commits them; aeo_scan_answers (metered, confirm first) probes the live engine.
• aeo_source_gap is the actionable end: the ranked list of web sources the assistant actually read, and which of them omit the app. That list is the user's off-platform work queue; MetaRun will not do that outreach.

6. Report

One page, ranked by leverage: quick metadata wins (with staged previews ready), rank trends worth watching, the Apple-measured funnel, rival moves, and the answer-engine gap with its source list. Every recommended write goes through preview -> user confirm -> apply; nothing is auto-applied.

Beyond this mission

The Sentinel watches tracked ranks in the background and files findings: aso_list_sentinel_findings. Scheduled promotional text runs through aso_schedule_promo_text. Discover more: npx metarun tools aso.