Store listings
Listing data read from the App Store and Google Play on every audit, for the app and for its competitors.
Methodology
How the audit works
The Apptonomy audit is ten engines. Each applies a written ASO methodology to store, keyword, competitor, review and AI-assistant data, and each methodology is published on its own page.
§1 · Authorship
Who writes the methodology
An in-house ASO practitioner writes and maintains the methodology for each engine.
A methodology states what the engine checks, which data it reads and how it scores the result.
Revisions are made by hand. An audit runs against the methodology current when it starts.
Language models carry out the methodology on the collected data. They do not decide what it checks.
Every recommendation shows the reasoning and the data behind it.
§2 · The engines
The engines, by the input they measure
Organic installs depend on two inputs: how many people see a listing, and how many of them install. Four of the engines measure the first, three the second, and three cover inputs to both. Each has its own methodology page.
Visibility
How many people see the listing.
Keywords & Clusters
Collects the keywords in the app's category, groups them into semantic clusters by topic and compares the app's coverage of each group with its competitors'.
Read the methodology →
AI Discoverability
Asks AI assistants category questions and records whether they recommend the app, and how often compared with its competitors.
Read the methodology →
User Intent
Predicts the queries users type, classifies the intent behind each one and scores how well the listing answers it.
Read the methodology →
Competitors
Builds the competitor set from store search results for the listing's keywords, the store's similar-apps list (the App Store's “You Might Also Like”, Google Play's “Similar apps”) and the sub-genre's top charts. Apps outside the app's sub-genre are dropped, except for non-game apps on Google Play: their listings name a single store category, so the language-model rating decides instead. On Google Play, and in several App Store categories including games, a language model rates each remaining candidate as a direct rival (same core use and audience), an adjacent app, a same-genre app or unrelated, and direct rivals rank first. Apps whose names borrow a better-known rival's are dropped; in games, only when they are small beside the audited app. The publisher's own other apps are dropped too, except in games, where one of its titles is kept when it is rated a direct rival; for an App Store game filed under a non-game primary genre, only when its title or subtitle also names a subject of the audited app's title or subtitle. The set is ranked by chart position and install volume, and changes to the competitors' listings are recorded.
Read the methodology →
Conversion
How many of them install.
Screenshots
Scores each screenshot and the set for message clarity, visual hierarchy and match between image and caption. The first three carry more weight because they are seen first: on the App Store, search results show the first three portrait screenshots.
Read the methodology →
Icon
Scores the icon for legibility at small sizes, accessibility, fit with the category's colors and difference from competitors' icons.
Read the methodology →
User Sentiment
Groups reviews into themes and separates what the listing can address from what belongs with the product team.
Read the methodology →
Both
Metadata, coverage and compliance.
Store Text
Checks the text fields for character use, keyword coverage, consistency across fields and whether each claim is supported. On the App Store, keyword coverage is read from the title and subtitle, and from the keyword field when App Store Connect is connected; the description is not indexed for search, so it is judged as copy for readers. On Google Play, the title, short description and full description are all indexed.
Read the methodology →
Localization Analysis
Checks which languages have a localized listing and where the store shows the default listing instead.
Read the methodology →
Policy Checker
Flags metadata that conflicts with App Store or Google Play guidelines before it is submitted.
Read the methodology →
§3 · Data
Where the data comes from
The sources are listed here. The prompts and scoring weights are not published.
Keyword data
Search volume is an estimate: Google Ads monthly searches converted to a store popularity scale with a curve fitted to Apple Ads search popularity, applied to both stores. Difficulty comes from the strength of the apps ranking for the keyword in live App Store and Google Play search results.
AI assistants
Category questions sent to ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. Share of voice is measured from their answers; the plan sets which assistants are asked.
Store analytics
When an account is connected, on Pro, Elite and Enterprise. From App Store Connect: impressions, product page views by source type, first-time downloads and conversion rate. From Play Console: store listing visitors and acquisitions, by search term and traffic source, and conversion rate. App Store Connect reports no search terms.
Reviews and the web
Store reviews for User Sentiment. Reddit and other public web pages for AI Discoverability.
Listings, reviews and store analytics are read from these sources, and search volume is the estimate described above. The search intents User Intent scores are named by language models from observed search phrases, and the questions AI Discoverability asks are written by them unless the team supplies its own.
§4 · Language models
How language models are used
- Prompts from the methodology. Each engine's prompts are written from its methodology and maintained with it.
- Structured output. Each engine returns typed fields checked against a schema, not free text. That is what lets a recommendation cite the data behind it.
- One scoring rubric. The engines score on a shared rubric that rolls up into a 0–100 ASO Readiness Score, shown with four pillar scores: Discoverability, Conversion, Trust and AI Discoverability. Successive audits can be compared on it.
- Scripts keyword extraction reads. Literal keyword extraction splits listing text written in Latin script (including accented letters), Chinese, Japanese and Korean, Arabic script and Thai. Listing text in Cyrillic, Greek, Hebrew or Devanagari is not split into literal keywords.
§5 · A recommendation
What a recommendation contains
Every recommendation in the product has the same parts.
- 1
The change
What to change in the listing, and where.
- 2
The finding
What the engine found, with the competitor or market comparison behind it.
- 3
The data
The sources the finding was read from.
- 4
The expected effect
The pillar score the change is expected to move.
- 5
The methodology
The engine and the part of its methodology that applies.
The AI Discoverability methodology page shows one engine in full →
§6 · From finding to change
From audit to published change
Audits run daily on Elite and Enterprise, twice a week on Pro, weekly on Basic and monthly on Free. Each audit's findings go through the same three steps.
Audit
How it works
The engines score the listing on the schedule the plan sets.
Plan and publish
Publishing
Findings become a ranked revision plan. On Pro, Elite and Enterprise, approved changes are written to the App Store version being prepared for submission and committed to Google Play, which reviews them.
Measure
Analytics
Later audits, and store analytics on Pro, Elite and Enterprise, show the movement after each published change.
§7 · Checking the claims
Compare with doing it on ChatGPT or Claude
The comparison page lists what building the same audit on ChatGPT or Claude would take: the prompt library, the store data feeds, the model orchestration and the schedule that keeps it running.
§8 · Any public app
Run the audit on any public app
The free audit runs the published methodology on any public app. No account is needed.
Get a free quick audit for your app.
FAQ
Product questions
The methodology, the data, publishing, and how Apptonomy fits an existing ASO workflow.