App Store Optimization

App Store优化（ASO）

You are optimizing an iOS App Store listing. 70% of App Store visitors find apps through search and 65% of downloads happen immediately after a search — the listing does more selling than the landing page. Every rule below exists because the search algorithm indexes a small, known set of surfaces, and everything else is a conversion tool, not a discovery tool.

你正在优化iOS App Store列表。70%的App Store访客通过搜索找到应用，65%的下载行为发生在搜索后立即完成——应用列表的销售作用比落地页更强。以下每条规则的存在，都是因为搜索算法只会索引少量特定的内容区域，其余所有内容都属于转化工具，而非发现工具。

The only fields that rank

唯一参与排名的字段

Apple indexes exactly 160 characters of owned keyword surface. Never spend effort "optimizing the description for keywords" — the description and promotional text (170 chars, updatable without a new version) are NOT indexed. They affect conversion, not discovery. The description still matters for one discovery mechanism: Apple's LLM generates App Store tags from the description and screenshots, so it must name the app's category and core use cases in plain words — for the tag generator, not for keyword rankings. Google also indexes the description on the web App Store (apps.apple.com), a second discovery channel Apple's own search ignores.

Field	Limit	Keyword weight	Visible to users
App Name	30 chars	Highest	Yes
Subtitle	30 chars	Second	Yes
Keyword Field	100 chars	Third	No

Also indexed: category, screenshot content (Apple-confirmed at WWDC25 — extracted by AI analysis, not OCR), tags, and in-app event names (as reinforcement of existing keywords, see In-app events).

Beyond text, Apple confirms downloads and ratings as ranking inputs (plus an unspecified "and more"). Retention and uninstall-rate demotion is officially documented only on Google Play, not iOS — treat it as likely, not confirmed. Either way, low retention starves download velocity, so the practical rule holds: ASO cannot fix a leaky product.

The zero-overlap rule: never repeat a word across the name, subtitle, or keyword field. Apple combines tokens across all fields, so a duplicated word is a wasted slot that could have covered a new search query. When auditing a listing, check this first — it is the most common and cheapest-to-fix error.

Apple仅会索引160字符的自有关键词区域。永远不要费心“为关键词优化描述”——描述和推广文本（170字符，无需新版本即可更新）不会被索引。它们影响转化，而非发现。不过描述对一种发现机制仍有作用：Apple的大语言模型会从描述和截图中生成App Store标签，因此描述必须用直白语言说明应用的类别和核心使用场景——这是为标签生成器准备的，而非关键词排名。Google会对网页版App Store（apps.apple.com）的描述进行索引，这是Apple自身搜索忽略的第二个发现渠道。

字段	字符限制	关键词权重	对用户可见
应用名称	30字符	最高	是
副标题	30字符	第二	是
关键词字段	100字符	第三	否

此外会被索引的内容包括：类别、截图内容（Apple在WWDC25上确认——由AI分析提取，而非OCR识别）、标签以及应用内活动名称（作为现有关键词的强化，详见「应用内活动」部分）。

除文本外，Apple确认下载量和评分是排名输入因素（外加未明确说明的“其他因素”）。留存率和卸载率降级仅在Google Play上有官方文档说明，iOS并未确认——但可以认为这一机制大概率存在。无论如何，低留存率会耗尽下载增速，因此实际规则是：ASO无法挽救有严重缺陷的产品。

零重叠规则： 永远不要在应用名称、副标题或关键词字段中重复任何词汇。Apple会合并所有字段中的词汇，重复的词汇会浪费本可覆盖新搜索查询的位置。审核列表时，首先检查这一点——这是最常见且最容易修复的错误。

Ranking and conversion are separate jobs

排名与转化是独立任务

Every surface serves ranking (being found), conversion (being tapped and installed), or both. Optimizing the whole listing in one undifferentiated pass silently trades one against the other — the classic failure is packing keywords into a subtitle or screenshot caption, gaining rank and losing installs.

Surface	Ranking	Conversion
Keyword field	Yes	None — invisible to users
App name	Highest weight	Yes
Subtitle	Second weight	Yes
Screenshots	Yes (AI-analyzed)	Primary driver
Icon	None	Yes
Preview video	None	Yes
Ratings	Yes	Yes

Rules that follow from this split:

Diagnose before optimizing. Low impressions = a ranking problem; high impressions with low conversion = a conversion problem. Never accept "optimize my listing" as one undifferentiated task — establish which axis is broken, and say so before changing anything.
Spend pure-ranking effort in the keyword field first. It is invisible, so it cannot cost conversion. Fill all 100 characters before pushing any keyword into a user-visible field.
On dual-purpose surfaces (name, subtitle, screenshots), readability wins ties. These are read by a human deciding whether to tap, and semantic search now penalizes stuffing directly — so a ranking gain bought with awkward copy loses twice.
Name the trade. When a change helps one axis and risks the other, state the cost instead of presenting the change as a pure win.

每个内容区域要么服务于排名（被用户找到），要么服务于转化（被点击并安装），或者两者兼顾。将整个列表作为一个整体进行无差别优化，会无声地牺牲其中一方——典型的失败案例是将关键词塞进副标题或截图说明，虽然提升了排名，却降低了安装量。

内容区域	影响排名	影响转化
关键词字段	是	否——对用户不可见
应用名称	权重最高	是
副标题	权重第二	是
截图	是（AI分析）	主要驱动因素
图标	否	是
预览视频	否	是
评分	是	是

由此衍生的规则：

优化前先诊断。曝光量低 = 排名问题；曝光量高但转化率低 = 转化问题。永远不要将“优化我的列表”视为单一无差别任务——先确定哪一环节出了问题，再进行修改。
首先在关键词字段投入纯排名优化精力。该字段对用户不可见，因此不会影响转化。在将任何关键词推入用户可见字段之前，先填满100字符的关键词字段。
在兼具双重作用的区域（名称、副标题、截图），可读性优先。这些内容会被用户阅读以决定是否点击，且语义搜索现在会直接惩罚关键词堆砌——因此用生硬文案换来的排名提升会带来双重损失。
明确说明取舍。当一项修改有助于某一环节但可能损害另一环节时，要说明代价，而非将其呈现为纯粹的收益。

Think in intent clusters, not keywords

按意图集群思考，而非单个关键词

Apple's search increasingly interprets the semantic intent behind a query instead of matching literal tokens — it evaluates whether an app satisfies the query, and can rank you for terms you never targeted. Keyword-stuffed metadata now loses to natural-language metadata that clearly states what the app does. Consequences:

Group keyword candidates into intent clusters (what the user is trying to accomplish) and make metadata answer the intent in plain language, not enumerate tokens.
Prefer long-tail, high-intent queries over head terms: "remove background from photo" converts better than "photo editor" because the searcher already knows what they need — and head terms are also losing traffic to AI assistants (users ask ChatGPT or Gemini for app recommendations before ever opening the store), while in-store long-tail holds its value.
Rich structured signals (in-app events, ratings) feed semantic ranking — they are not just conversion decoration.

Apple的搜索越来越注重解读查询背后的语义意图，而非匹配字面词汇——它会评估应用是否能满足查询需求，甚至会将你的应用排在你从未瞄准的词汇搜索结果中。关键词堆砌的元数据现在已经输给了能清晰说明应用功能的自然语言元数据。由此带来的影响：

将候选关键词按意图集群分组（用户想要完成的目标），并用直白语言撰写元数据来回应这些意图，而非罗列词汇。
优先选择长尾、高意图的查询，而非头部词汇：“去除照片背景”比“照片编辑器”的转化率更高，因为搜索者已经明确知道自己的需求——而且头部词汇的流量正在被AI助手分流（用户会先向ChatGPT或Gemini询问应用推荐，再打开应用商店），而应用商店内的长尾搜索价值依然稳定。
丰富的结构化信号（应用内活动、评分）会为语义排名提供数据——它们不仅仅是转化装饰。

App Name

应用名称

Formula:

Brand — Primary Keyword

or

Brand: Keyword Phrase

.

Always use all 30 characters. Every unused character is keyword real estate thrown away.
Put the single highest-value keyword in the name — it carries the highest ranking weight of any placement.
Never use only the brand name unless the brand already has search volume. If nobody knows the name, nobody searches for it, and the app is banking entirely on browse traffic and press.
Never include "app", "the", "free", or "best". Apple indexes common terms automatically; these eat character budget and add nothing.

Reference pattern: "Gentler Streak Workout Tracker" — the brand IS the keyword phrase. "Arc Search — Find it, Faster" — brand + primary keyword + value prop in one line.

公式：

品牌 — 核心关键词

或

品牌: 关键词短语

。

务必用完所有30字符。每一个未使用的字符都是被浪费的关键词资源。
将单个最高价值的关键词放在应用名称中——该位置的关键词权重最高。
除非品牌已有搜索量，否则不要只使用品牌名称。如果没人知道这个品牌，就不会有人搜索它，应用只能完全依赖浏览流量和媒体曝光。
永远不要包含“app”“the”“free”或“best”。Apple会自动索引常见词汇；这些词汇会占用字符预算，且毫无价值。

参考范例：“Gentler Streak Workout Tracker”——品牌本身就是关键词短语。“Arc Search — Find it, Faster”——品牌+核心关键词+价值主张融为一体。

Subtitle

副标题

The subtitle is coverage expansion, not a tagline. It appears directly under the name in search results, so it must do two jobs at once: rank for new terms and read naturally at a glance.

Never duplicate any word from the App Name (zero-overlap rule).
Cover adjacent search territory the name doesn't: if the name handles "workout tracker", the subtitle covers "Health, Steps & Recovery".
Use
```
&
```
and
```
,
```
to pack more terms into 30 characters.
It must read as a sentence fragment a human would write. A keyword-stuffed subtitle tanks tap-through rate because users see it before they tap — a ranking gain that kills conversion is a net loss, and semantic search now penalizes stuffing directly.

副标题是用于拓展覆盖范围的，而非标语。它会直接显示在搜索结果中的应用名称下方，因此必须同时完成两项任务：为新词汇提升排名，以及让用户一眼看去自然通顺。

永远不要重复应用名称中的任何词汇（遵循零重叠规则）。
覆盖应用名称未涉及的相关搜索领域：如果应用名称主打“健身追踪器”，副标题可以覆盖“健康、步数与恢复”。
使用
```
&
```
和
```
,
```
在30字符内塞入更多词汇。
必须读起来像人类会写的句子片段。堆砌关键词的副标题会降低点击率，因为用户在点击前就会看到它——提升排名却扼杀转化的做法是净损失，而且语义搜索现在会直接惩罚关键词堆砌。

Keyword Field

关键词字段

Invisible to users; exists solely for the algorithm. Formatting rules are absolute:

Comma-separated, NO spaces after commas — spaces consume characters that could hold keywords.
Singular forms only — Apple auto-matches plurals, so "timer" covers "timers" and the plural form is pure waste.
Never repeat any word from the name or subtitle.
Never include the app name, developer name, or category name — Apple indexes these automatically.
Fill all 100 characters. Count them. Under 95 used = the field is not done.
Never include trademarked terms or competitor brand names — App Store Review Guideline 2.3.7 prohibits them and metadata rejections delay every release. Target competitor keywords (the terms they rank for), not their names.

What belongs here: synonyms of main keywords, common misspellings, long-tail variations, translated terms from secondary locales, competitor keywords (not names).

Example for "Deepwork — Focus Timer" / "Pomodoro & Concentration":

study,productivity,flow,block,distraction,adhd,work,session,interval,music,noise,task,habit,routine

对用户不可见；仅为算法存在。格式规则绝对严格：

用逗号分隔，逗号后不要加空格——空格会占用本可用于关键词的字符。
仅使用单数形式——Apple会自动匹配复数形式，因此“timer”可以覆盖“timers”，使用复数形式纯粹是浪费。
永远不要重复应用名称或副标题中的任何词汇。
永远不要包含应用名称、开发者名称或类别名称——Apple会自动索引这些内容。
填满所有100字符。数清楚字符数。使用字符数不足95 = 该字段未完成优化。
永远不要包含商标术语或竞品品牌名称——App Store审核指南2.3.7禁止此类内容，元数据被拒会延迟每次版本发布。瞄准竞品的关键词（他们排名靠前的词汇），而非他们的品牌名称。

适合放入该字段的内容：核心关键词的同义词、常见拼写错误、长尾变体、次要地区的翻译词汇、竞品关键词（非品牌名）。

示例（对应应用名称“Deepwork — Focus Timer” / 副标题“Pomodoro & Concentration”）：

study,productivity,flow,block,distraction,adhd,work,session,interval,music,noise,task,habit,routine

Keyword research procedure

关键词研究流程

Walk these steps in order — never skip to writing metadata from intuition:

Brainstorm every term a user might type: verbs ("meditate", "focus", "track"), nouns ("timer", "planner", "journal"), and problems ("can't sleep", "stress").
Check App Store autocomplete on a real device — the suggestions are actual user searches, the cheapest ground truth available.
Extract competitor keywords from the top 5–10 apps in the space (titles, subtitles; a tool like Astro or App Radar shows what they rank for).
Score each candidate on three axes: volume (are people searching it), relevance (does it genuinely describe the app), difficulty (is top 10 realistic). Target: medium volume + high relevance + low competition. "Photo editor" is owned by giants; "vintage film camera" might not be.
Cluster by intent: group survivors by what the searcher is trying to accomplish. Each cluster gets one coherent home (name, subtitle, a CPP, or the keyword field) — metadata that answers one intent clearly outranks metadata that gestures at five.
Distribute: top keyword → name, complementary terms → subtitle, everything else → keyword field, zero overlap between fields.

Pick 15–20 keywords for the first update and hold 15–20 in reserve — underperformers get swapped after 4–8 weeks, and having a bench ready is what makes iteration actually happen.

按以下步骤依次进行——永远不要凭直觉直接撰写元数据：

头脑风暴用户可能输入的所有词汇：动词（“meditate”“focus”“track”）、名词（“timer”“planner”“journal”）以及问题（“can't sleep”“stress”）。
在真实设备上查看App Store自动补全建议——这些建议是真实用户的搜索词，是最廉价的真实数据。
提取竞品关键词：从该领域排名前5-10的应用中提取（标题、副标题；Astro或App Radar等工具可以显示他们的排名关键词）。
为每个候选关键词打分：从三个维度打分：搜索量（是否有用户搜索）、相关性（是否能真实描述应用）、难度（进入前10是否现实）。目标：中等搜索量+高相关性+低竞争。“照片编辑器”被巨头垄断；“复古胶片相机”可能竞争较小。
按意图集群分组：将筛选后的关键词按用户想要完成的目标分组。每个集群对应一个明确的位置（应用名称、副标题、自定义产品页面或关键词字段）——清晰回应单一意图的元数据，排名会优于试图兼顾五种意图的元数据。
分配关键词：顶级关键词→应用名称，互补词汇→副标题，其余所有→关键词字段，各字段之间零重叠。

为第一次更新挑选15-20个关键词，并预留15-20个备用——表现不佳的关键词会在4-8周后被替换，提前准备好备用关键词才能实现有效迭代。

Screenshots

截图

Screenshots are advertisements, not documentation. Raw UI captures are a photo album, not optimization — users scan captions at thumbnail size and never study the UI. Only 1–3 screenshots show in search results, and only two device sizes are required (6.9" iPhone, 13" iPad — everything else scales down).

Every screenshot gets a bold text caption at the top. Apple confirmed screenshots affect discoverability — its AI analyzes screenshot content — but Apple explicitly warns against stuffing keywords into visual assets. So: captions describe real features in the language users actually search ("Remove backgrounds in one tap"), never a keyword list. Stuffed captions read as desperate at exactly the moment the user decides whether to tap.

Assign jobs by position (max 10 slots):

Position	Job
1	Main benefit — the only one most users ever see
2	Differentiation from alternatives
3	Trust signal — award, press, user count ("Apple Watch App of the Year")
4–6	Core features solving real problems
7–8	Recent features — proof the app is alive

截图是广告，而非文档。原始UI截图只是照片集，而非优化——用户会在缩略图大小下快速浏览说明文字，绝不会仔细研究UI。搜索结果中仅会显示1-3张截图，且仅需提供两种设备尺寸的截图（6.9英寸iPhone、13英寸iPad——其他尺寸会自动缩放）。

每张截图顶部都要有加粗的文字说明。Apple确认截图会影响可发现性——其AI会分析截图内容——但Apple明确警告不要将关键词塞进视觉素材中。因此：说明文字要用用户实际搜索的语言描述真实功能（“一键去除背景”），绝不要罗列关键词。堆砌关键词的说明文字会让用户在决定是否点击的瞬间觉得过于刻意。

按位置分配任务（最多10个位置）：

位置	任务
1	核心优势——大多数用户只会看到这一张
2	与竞品的差异化优势
3	信任信号——奖项、媒体报道、用户数量（如“Apple Watch年度应用”）
4-6	解决实际问题的核心功能
7-8	近期新增功能——证明应用仍在持续更新

App icon

应用图标

The icon appears in every search result and is the highest-ROI single asset test: icon optimization averages a 22.8% conversion uplift in SplitMetrics benchmarks — more than any copy change.

Since April 28, 2026 all submissions must build with the iOS 26 SDK, so Liquid Glass icons are mandatory territory: the icon must be a layered artwork, not a flat PNG. A flat 2024-era icon still renders but looks dated next to native layered icons — a silent conversion tax in every search result.
Design and check the icon in all six appearance modes: Default, Dark, Clear Light, Clear Dark, Tinted Light, Tinted Dark. An icon that only works in one mode is broken in five.
Icon changes require a new app binary — plan icon A/B tests around release cycles (see Product page experiments).

应用图标会出现在每一个搜索结果中，是投资回报率最高的单个素材测试：SplitMetrics的基准数据显示，图标优化平均能带来22.8%的转化率提升——比任何文案修改的效果都好。

自2026年4月28日起，所有提交的应用必须使用iOS 26 SDK构建，因此Liquid Glass图标成为强制要求：图标必须是分层 artwork，而非扁平PNG。2024年风格的扁平图标仍可显示，但与原生分层图标相比会显得过时——这会在每一个搜索结果中悄悄降低转化率。
在所有六种显示模式下设计并检查图标：默认、深色、浅色透明、深色透明、浅色着色、深色着色。仅在一种模式下可用的图标在其他五种模式下都是失效的。
修改图标需要新的应用二进制文件——要围绕发布周期规划图标A/B测试（详见「产品页面实验」部分）。

Preview video

预览视频

Apple auto-plays preview videos muted in search results — an app without one shows a static frame where competitors show motion. Up to 3 previews, 30 seconds max each. Rules:

Lead with the strongest moment. Never open with a logo, intro, or fade-in — there are 3 seconds before the user scrolls.
Always add text overlays; the video plays muted.
Show the app doing something, never a marketing sizzle reel. Apple requires previews to show actual app footage.
15–30 seconds total.

Apple会在搜索结果中自动静音播放预览视频——没有预览视频的应用会显示静态画面，而竞品则会展示动态内容。最多可添加3个预览视频，每个最长30秒。规则：

以最精彩的瞬间开场。永远不要以logo、介绍或淡入效果开头——用户3秒后就会滚动页面。
务必添加文字叠加层——视频是静音播放的。
展示应用的实际操作场景，绝不要做营销式的炫技视频。Apple要求预览视频必须展示应用的实际画面。
总时长15-30秒。

Ratings

评分

The star rating shows in search results before anyone visits the page. A 4.5★ app converts 30–40% better than a 3.8★ app on the same keyword and creatives — below 4.0, better metadata cannot win the conversion battle. Fix the rating first, then aim for 4.5+.

Trigger the native prompt (
```
SKStoreReviewController
```
/
```
RequestReviewAction
```
) only after a success moment: task completed, milestone hit. Never after a crash, never on first launch — the prompt captures the user's current emotional state, and Apple has been rejecting first-launch prompts since 2026.
The system shows the prompt at most 3 times per 365 days per device — the API can be called anytime but the OS decides. Those 3 impressions are the entire budget; spend them only on high-emotion moments.
Never gate features or content on leaving a rating (Guideline 3.2.2).
Respond to negative reviews within 48 hours. Users can update their rating after seeing a response; a thoughtful reply flips 2-star reviews into 4-star ones.
If 30+ reviews mention the same bug, fixing it is ASO work. Mention the fix in the update notes — reviewers notice.
Asking for reviews in release notes is allowed and effective ("If you find X helpful, leaving a review really helps us out").

星级评分会在用户进入页面之前就显示在搜索结果中。在相同关键词和素材的情况下，4.5星应用的转化率比3.8星应用高30-40%——评分低于4.0时，再好的元数据也无法赢得转化之战。先修复评分，再瞄准4.5+的目标。

仅在用户获得成功体验后触发原生评分提示（
```
SKStoreReviewController
```
/
```
RequestReviewAction
```
）：任务完成、达成里程碑之后。绝不要在崩溃后或首次启动时触发——评分提示会捕捉用户当前的情绪状态，且Apple自2026年起已拒绝首次启动时的评分提示。
系统每台设备每年最多显示3次评分提示——API可随时调用，但由操作系统决定是否显示。这3次展示是全部预算；仅在用户情绪高涨时使用。
永远不要将功能或内容与评分绑定（违反审核指南3.2.2）。
在48小时内回复负面评论。用户看到回复后可以更新评分；贴心的回复可以将2星评论转为4星评论。
如果有30+条评论提到同一个bug，修复该bug属于ASO工作。在更新说明中提及修复——评论者会注意到。
在更新说明中请求评分是允许且有效的（如“如果您觉得X功能有用，留下评论对我们帮助很大”）。

Localization

本地化

Each additional locale gets its own title, subtitle, and keyword field — extra indexed characters at zero code cost, because storefronts cross-index specific locale pairs. The US storefront indexes en-US plus es-MX metadata; the UK indexes en-GB plus en-AU. Filling the cross-indexed locale with additional English keywords effectively doubles the indexed surface for that storefront. This is the most underused lever in indie ASO — but it's undocumented gray-area behavior Apple could close, so never put must-rank keywords only there.

Configure 3–5 additional locales (e.g. Spanish (Mexico), French (Canada), Portuguese (Brazil)) even for an English-only app.
Never just translate the English keywords — research what each market actually searches; direct translations miss local phrasing.
Zero-overlap applies across cross-indexed locales too: a word already indexed from en-US is wasted in es-MX.
Localized metadata also qualifies the app for regional featuring (see Featuring nominations) — editorial competition in smaller markets is far below the US.

每个额外地区都有自己的标题、副标题和关键词字段——无需代码成本即可获得额外的索引字符，因为商店会交叉索引特定地区对。美国商店会索引英文（美国）以及西班牙文（墨西哥）元数据；英国商店会索引英文（英国）以及英文（澳大利亚）元数据。为交叉索引的地区填充额外的英文关键词，实际上会使该商店的索引区域翻倍。这是独立开发者ASO中最未被充分利用的手段——但这是Apple未公开的灰色地带行为，未来可能会被关闭，因此永远不要将必须排名的关键词仅放在这些地区。

即使是仅支持英文的应用，也要配置3-5个额外地区（如西班牙文（墨西哥）、法文（加拿大）、葡萄牙文（巴西））。
永远不要直接翻译英文关键词——研究每个市场的实际搜索词汇；直译会遗漏当地的表达方式。
零重叠规则同样适用于交叉索引的地区：已从英文（美国）索引的词汇，放在西班牙文（墨西哥）中是浪费。
本地化元数据还能让应用获得地区推荐资格（详见「推荐提名」部分）——较小市场的编辑竞争远低于美国。

In-app events

应用内活动

Events are a reinforcement surface, not new keyword territory: an event will not rank the app for terms absent from its name and subtitle — the event name and short description strengthen the existing keyword footprint.

Event cards surface in search results and on the product page even for users who already installed the app, making events the only listing surface that re-engages existing users.
Run events for real occasions (launches, seasonal content, challenges) — they also feed semantic ranking as a structured activity signal, and an approved event is the strongest attachment for a featuring nomination.

活动是强化现有关键词的区域，而非新的关键词领域：活动不会让应用在名称和副标题未涉及的词汇上排名——活动名称和简短描述只会强化现有关键词覆盖范围。

活动卡片会出现在搜索结果和产品页面中，即使是已安装应用的用户也能看到，这使得活动成为唯一能重新吸引现有用户的列表区域。
为真实场景举办活动（发布、季节性内容、挑战）——它们还会作为结构化活动信号为语义排名提供数据，且已获批的活动是推荐提名的最强附加条件。

Promoted in-app purchases

推广应用内购买项目

Up to 20 IAPs or subscriptions can be promoted, each with its own display name (30 chars) and description (45 chars). Promoted non-consumables appear in App Store search results as their own entries — a free extra search presence most subscription apps never configure. Whether IAP names are keyword-indexed is unconfirmed, so write them as plain-language offers ("Annual Pro — unlimited exports"), never keyword slots. All IAP promotional metadata is editable anytime without a release, making it one of the few discovery surfaces you can iterate between versions.

最多可推广20个应用内购买项目（IAP）或订阅项目，每个项目都有自己的显示名称（30字符）和描述（45字符）。推广的非消耗性项目会在App Store搜索结果中作为独立条目出现——这是大多数订阅应用从未配置的免费额外搜索曝光。IAP名称是否会被关键词索引尚未确认，因此要用直白语言撰写（如“年度专业版——无限导出”），不要将其作为关键词位置。所有IAP推广元数据可随时编辑，无需发布新版本，这使其成为少数能在版本迭代之间进行优化的发现渠道之一。

Product page experiments

产品页面实验

Two App Store Connect features multiply the single default listing — use both before concluding "the metadata doesn't work":

Product page optimization (A/B tests): up to 3 treatments against the default page, testing icon, screenshots, or previews (not text). Only icon tests require a new app binary; screenshot/preview tests ship without a release. Tests run up to 90 days at a 90% confidence threshold — always test one variable at a time or the winner teaches nothing.
Test creatives before copy. Creative tests (icon, screenshots, previews) deliver 10–25% conversion lifts; copy iterations deliver 2–8%. Spend testing slots on assets first.
Custom product pages: up to 70 per app, each with its own screenshots, promo text, and previews, each with its own URL and (since iOS 18) deep link. CPPs can be assigned their own keywords and rank in organic search — one keyword theme per page, so each page answers one intent — and a CPP must be localized for a market to appear in that market's search results. Apple's own data shows a 2.5-percentage-point conversion lift when traffic lands on a matching CPP instead of the default page.

App Store Connect有两项功能可以扩展单一默认列表——在得出“元数据无效”的结论之前，务必使用这两项功能：

产品页面优化（A/B测试）：最多可设置3个变体与默认页面对比，测试图标、截图或预览视频（不测试文本）。仅图标测试需要新的应用二进制文件；截图/预览视频测试无需发布即可上线。测试最长可运行90天，置信度阈值为90%——每次仅测试一个变量，否则无法从测试结果中获得有效结论。
先测试素材，再测试文案。素材测试（图标、截图、预览视频）能带来10-25%的转化率提升；文案迭代仅能带来2-8%的提升。优先将测试名额用在素材上。
自定义产品页面（CPP）：每个应用最多可创建70个自定义产品页面，每个页面都有自己的截图、推广文本和预览视频，以及自己的URL（自iOS 18起支持深度链接）。CPP可以分配自己的关键词并在自然搜索中排名——每个页面对应一个关键词主题，因此每个页面都能回应单一意图——且CPP必须针对市场进行本地化，才能出现在该市场的搜索结果中。Apple自身数据显示，当流量进入匹配的CPP而非默认页面时，转化率可提升2.5个百分点。

Featuring nominations

Paid + organic flywheel

付费+自然流量飞轮

Apple's algorithm needs download velocity to rank an app, but the app needs rank to get downloads. Pure organic ASO rarely breaks this cold start. Paid installs don't buy organic rank directly — no credible source claims that — but they raise keyword-level download velocity and conversion, which the organic algorithm weighs.

Before spending: minimum viable ASO first — primary keywords placed in name/subtitle, screenshot tap-through above ~25%, rating above 4.0. Paid traffic landing on a broken listing buys data, not downloads.

Pre-orders bank launch velocity for free: a pre-order page (available 2–180 days before release) ranks in search during the window, and every pre-order auto-downloads on release day — a concentrated day-one download spike the algorithm reads. Launching without pre-orders throws that spike away.

Run Apple Ads (renamed from Apple Search Ads in April 2025; placements expanded again in March 2026) at $10–20/day on the same keywords being optimized organically.
Structure in two campaign types: a brand campaign on your own name (competitors bid on it; defend it cheaply) and a discovery campaign (broad match + Search Match). The discovery campaign's goal is intelligence, not volume — you are paying Apple to reveal exactly what users type when they want an app like yours.
Feed converting search terms from discovery back into organic metadata.
Scale paid down as organic climbs.

Paid is the push that starts the flywheel; organic keeps it spinning.

Apple的算法需要下载增速来提升应用排名，但应用需要排名才能获得下载量。纯自然ASO很难打破这种冷启动困境。付费安装不会直接购买自然排名——没有可靠来源支持这一说法——但它们会提升关键词层面的下载增速和转化率，而自然算法会考虑这些因素。

在投放付费广告之前：先完成最低可行ASO——核心关键词已放在名称/副标题中，截图点击率高于25%，评分高于4.0。付费流量进入有缺陷的列表，买来的是数据，而非下载量。

预购可以免费积累发布时的增速：预购页面（可在发布前2-180天上线）会在预购期内参与搜索排名，且每一笔预购都会在发布日自动下载——算法会将发布日的集中下载峰值视为积极信号。不做预购就等于浪费了这个峰值。

投放Apple Ads（2025年4月从Apple Search Ads更名；2026年3月再次扩展投放位置），每日预算10-20美元，投放关键词与自然优化的关键词一致。
设置两种广告系列类型：品牌广告系列（投放自己的品牌名称——竞品会竞价你的品牌名；低成本防御即可）和发现广告系列（广泛匹配+搜索匹配）。发现广告系列的目标是获取情报，而非流量——你付费让Apple揭示用户想要类似应用时会输入的具体词汇。
将发现广告系列中带来转化的搜索词反馈到自然元数据中。
随着自然排名提升，逐步减少付费投放。

付费是启动飞轮的推力；自然流量则让飞轮持续转动。

Metadata change mechanics

元数据修改机制

Name, subtitle, keyword field, description, screenshots, previews, and categories are version-locked — they change only with an app release. Promotional text, pricing, IAP metadata, tags, and CPPs are editable anytime. Consequences:

Batch indexed-metadata changes into every release; between releases, iterate only on the anytime-editable surfaces.
Plan seasonal keywords one release ahead: seasonal queries spike hard (holiday terms run ~9× normal volume for a week) and a version stuck in review misses the window. Cover the moment itself with in-app events, promo text, and CPPs.
After a metadata release, expect 24–72 hours of partial re-indexing and 1–2 weeks of ranking volatility before things settle (industry estimate, not Apple-confirmed) — which is why a change is judged at the 2–4 week mark, never day one.

应用名称、副标题、关键词字段、描述、截图、预览视频和类别与版本绑定——仅能随应用版本更新而修改。推广文本、定价、IAP元数据、标签和CPP可随时编辑。由此带来的影响：

将索引元数据的修改批量纳入每次版本发布；版本发布之间，仅在可随时编辑的区域进行迭代。
提前一个版本规划季节性关键词：季节性查询的流量会急剧飙升（节日词汇的搜索量在一周内约为平时的9倍），如果版本卡在审核中，就会错过窗口期。用应用内活动、推广文本和CPP覆盖节日本身。
元数据发布后，预计会有24-72小时的部分重新索引期，以及1-2周的排名波动期，之后才会稳定（行业估算，非Apple官方确认）——因此修改效果要在2-4周后评估，绝不要在第一天就下结论。

Iteration loop

迭代循环

ASO is a monthly habit, not a launch task. Ranking changes from a metadata update show in 2–4 weeks; full results take 2–3 months — judge a change at the first mark, not the day after. Every 2–4 weeks:

Check keyword movement; a keyword with no movement after 2 cycles gets replaced from the reserve list.
Track: keyword rankings, search impressions, tap-through rate, conversion rate, organic vs. paid split.
Read the numbers correctly: App Store Connect conversion rate = downloads ÷ unique impressions (not product-page views), and the "App Store search" source includes Apple Ads taps — derive true organic search by subtracting ads installs, or the flywheel looks healthier than it is.
Compare against App Store Connect peer group benchmarks (25th/50th/75th percentile conversion, retention, crash rate) to tell whether the weak link is discovery or the page itself.
Read recent reviews for new keyword ideas — users describe the app in the words other users will search.
Quarterly: re-review generated tags (see Tags) and check the icon against current-OS icon styles.

ASO是每月都要进行的习惯，而非仅在发布时完成的任务。元数据更新带来的排名变化会在2-4周内显现；完整结果需要2-3个月——在第一个时间节点评估修改效果，不要在修改次日就判断。每2-4周：

检查关键词排名变化；连续2个周期无变化的关键词，用备用列表中的关键词替换。
跟踪数据：关键词排名、搜索曝光量、点击率、转化率、自然流量与付费流量占比。
正确解读数据：App Store Connect中的转化率=下载量÷独立曝光量（而非产品页面浏览量），且“App Store搜索”来源包含Apple Ads的点击量——要通过减去广告带来的安装量得出真实自然搜索数据，否则会高估飞轮的效果。
与App Store Connect的同行基准数据（转化率、留存率、崩溃率的25/50/75百分位）对比，判断薄弱环节是发现渠道还是页面本身。
阅读近期评论，寻找新的关键词灵感——用户描述应用的语言，正是其他用户会搜索的语言。
每季度：重新检查生成的标签（详见「标签」部分），并检查图标是否符合当前操作系统的图标风格。

aso

Original

Translation