# How App Store search matches metadata

> This is the model the openaso server teaches your agent. It is also the part of ASO most tooling gets wrong — worth understanding even if an agent does the work.

## One pool, three fields

For each localization, Apple indexes the **title** (30 characters), the **subtitle** (30), and the hidden **keyword field** (100) as **one combined word pool**. An app can rank for any query assembled from words across those fields: `photo` in the title plus `resize` in the keyword field can rank for *"resize photo"*.

- Title words weigh most, then subtitle, then keyword field.
- Word order and case do not gate what can rank — a phrase kept intact in the title or subtitle is a mild extra relevance signal, nothing more.
- Repeating a word in a second field adds nothing. That budget is wasted.
- Storefronts may index extra localizations — the US storefront also indexes Spanish (Mexico) — expanding the usable pool beyond one localization's budgets.

## Keyword-field mechanics

Commas count toward the 100 characters, and spaces around commas are pure waste: `a,b` not `a, b`. A space that is a word's only separator costs the same as the comma replacing it.

## Optimize the pool, not phrases

A word's search value is the sum of every phrase it can combine into — `photo` serves *photo editor*, *photo maker*, and *take photo* at once. But a phrase's value accrues to each of its words, so compare words by **marginal unlock and at-risk value**, not by summing phrase totals. The `coverage` tool computes all of this locally for a draft: pool, duplicates, budgets, keyword-field waste, per-phrase coverage, and per-word value when you pass volume-weighted phrases.

```json
{
  "phrase": "resize photo",
  "covered": true,
  "via": { "resize": "keywords", "photo": "title" },
  "note": "cross-field combinations tend to rank weaker than same-field phrases — verify with rank"
}
```

## Caveats the tools surface

- `rank` covers the top 200 search results — the public API's cap.
- `volume` returns Apple's real Search Ads popularity when `source: "apple"`; where Apple floors low-demand keywords or carries no signal, `source: "proxy"` is a transparent independent estimate with per-signal components.
- The live keyword field and subtitle are not exposed by any public API — tools infer around that honestly rather than pretending to read them.
- `revenue` is a modeled order-of-magnitude band from public grossing ranks. Agents are told to report it as an estimate, never as fact.
- Google Play is not covered.

## Go deeper

This page is the model in one sitting. The [ASO reference](https://openaso.ai/reference.md) takes each mechanism apart:

- [Title & subtitle weighting](https://openaso.ai/reference/title-subtitle.md) — the two strongest fields, and when changes take effect
- [Keyword field mechanics](https://openaso.ai/reference/keyword-field.md) — the 100-character accounting in full
- [Cross-localization indexing](https://openaso.ai/reference/cross-localization-indexing.md) — the extra budget most listings never claim
- [Search ranking factors](https://openaso.ai/reference/search-ranking-factors.md) — what orders the candidates metadata qualifies
- [Search popularity & autocomplete](https://openaso.ai/reference/search-popularity.md) — where demand numbers actually come from
- [Ratings & reviews](https://openaso.ai/reference/ratings-reviews.md) — per-storefront ratings and the conversion loop
- [Top charts](https://openaso.ai/reference/top-charts.md) — velocity ranking and the one public monetization signal
- [Revenue estimation](https://openaso.ai/reference/revenue-estimates.md) — the power-law methodology, in the open

## Terms

- **combined word pool** — The set of words a storefront indexes for an app across title, subtitle, and keyword field (and any extra indexed localizations) — queries match against the pool, not against fields in isolation.
- **field weight** — The relevance hierarchy among indexed fields: title heaviest, then subtitle, then keyword field.
- **coverage** — Whether a target phrase can be assembled from a draft's word pool, and at what cost — computed per phrase, per word, per field.