Public store listings
Live listing data read directly from the iOS and Google Play on every audit. Daily for paying customers. The competitive set is read on the same cadence.
The methodology
We show our work.
Senior-grade ASO best practices, encoded as ten specialized engines and kept current weekly by an in-house senior ASO expert. AI scales the methodology; it doesn't author it. Every engine documented. Every recommendation explained. Every score sourced.
The ASO work a senior consultant would do, repeated on every app, every day.
Senior-grade ASO at the price of a tool.
§2
AI is the muscle. Expertise is the brain.
Most AI-native ASO startups position AI as the source of the insights. Apptonomy positions AI as the runtime — the mechanism that makes a senior ASO expert's framework operate at machine speed across every app, every market, every day.
The framework is the brain. AI is the muscle. That hierarchy is visible at every layer of the platform: per-engine methodology pages are public, plain-language reasoning accompanies every recommendation, and the underlying scoring rubric is consistent across engines so outputs can be compared, audited, and verified.
We don't lead with “AI-powered” anywhere else on the site — not because AI isn't doing real work, but because “AI-powered” is what every thin wrapper says. The differentiation is how AI is used, not that AI is used.
§3
The methodology stays current. Every week.
ASO best practices change as Apple and Google ship algorithm updates, as the LLM discovery landscape evolves, and as patterns emerge across the customer base. Working from last quarter's playbook means working from outdated knowledge.
Apptonomy's methodology is reviewed and updated every week by an in-house senior ASO expert. The cadence is operational, not aspirational. When Apple ships an algorithm tweak or Google rolls out a Play Store change, the affected engines' frameworks are reviewed and updated — and every customer audit produced after that point runs against the updated framework.
The platform itself is the proof. Paste a URL, run the free audit, and what comes back is the output of a methodology that was current as of this week. The verification paths below — “too good to be true?” on your own URL, the DIY-frontier-LLMs comparison, the per-engine methodology pages — let any prospect test the claim directly without taking our word for it.
§4
The audit engines, organized by user goal.
Each engine has its own published methodology page and its own per-recommendation reasoning surface in the product.
App Discovery
How users find your app.
Keywords & Clusters
Extracts every keyword in your category, clusters them by user intent, and identifies the clusters where you're under-positioned vs. competitors.
Read the methodology →
AI Discoverability
Measures whether ChatGPT, Claude, Gemini, Perplexity, and AI Overviews actually recommend your app, with share-of-voice and intent coverage against peers.
Read the methodology →
User Intent
Predicts the queries users actually type, classifies the underlying intents — functional, emotional, social — and scores your listing coverage against them.
Read the methodology →
Competitors
Builds your real competitive set from actual keyword-search overlap, monitors it every day, and surfaces every meaningful change as it happens.
Read the methodology →
Install Conversion
What turns browsers into installers.
Store Text
Audits title, subtitle, keyword field, and description as a coordinated system — character utilization, keyword coverage, cross-field consistency, and factual claim grounding.
Read the methodology →
Screenshots
Scores every frame and the full set against what moves conversion — message clarity, visual hierarchy, and image-text match — with the first three frames weighted higher because that is what the store card shows.
Read the methodology →
Icon
Scores the icon for thumbnail clarity, accessibility, category-color fit, and distinctiveness against your real competitive set.
Read the methodology →
User Sentiment
Turns thousands of reviews into ranked drivers, themes, and ASO-actionable edits — separating what should change the listing from what should be escalated to product.
Read the methodology →
§5 · Worked example
One real AI Discoverability recommendation, traced end-to-end.
One concrete recommendation, from input to output, with every layer of the methodology visible.
AI Discoverability Engine
+12 AI Discoverability scoreAdd “for beginners” framing across your subtitle, description opener, and screenshot 2 caption.
- Why
- Your app surfaces in only 8% of LLM queries vs. category-leader 52%. The “beginner-friendly running app” intent has zero presence across all five LLMs, and your website already leads with beginner framing while your store listing emphasizes advanced metrics — a cross-channel inconsistency.
- Data sources
- Five-LLM prompt battery, website scrape, listing text, intent map.
- Expected impact
- +12 AI Discoverability score.
What's behind the recommendation
- 1
The ASO framework the engine encodes.
Cross-channel intent coverage with peer share-of-voice as the benchmark. The framework tells the engine what to look for: intents the listing misses, intents the website surfaces but the listing doesn't, intents peers cover that you don't.
- 2
The data.
A five-LLM prompt battery run in parallel (ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews), the live listing text, a scrape of the marketing website for cross-channel consistency, and a category intent map.
- 3
The AI's role.
Once the framework is defined, AI executes it at scale — query battery execution, intent classification, share-of-voice scoring, cross-channel reconciliation. The framework decides what matters; AI runs the framework across every signal.
- 4
The methodology cadence.
This engine's framework is reviewed weekly. When LLM behavior shifts (model updates, new providers, citation-pattern changes), the engine is updated within a week — and every customer audit produced after that point runs against the updated framework.
The same shape repeats for every engine. Each one ships with its own published methodology so the reasoning is auditable, not vibes-based. See the full AI Discoverability methodology →
§6 · Data sources
Where the signal comes from.
The data backbone is documented. The prompts are proprietary; the sources are not.
Keyword and search-volume data
Proprietary aggregation across Google Ads, Apple Search Ads popularity, and Google Trends. Volume, difficulty, current rank, and trend direction feed every keyword-related engine.
AI discoverability surfaces
Direct queries against ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews with category-relevant prompts. Share-of-voice and citation patterns are measured against peers, not estimated.
App Store Connect + Play Console
When OAuth-linked, search-term performance, source attribution, and conversion analytics flow in so recommendations reflect the queries actually sending traffic.
User reviews and open-web signals
Store-API review streams feed Sentiment; category-relevant web and community signals feed AI Discoverability. No synthetic data; where AI runs, it runs on data with a named external source.
No synthetic data. No model-imagined facts. Where AI is used, it operates on data with an external source we can name.
§7 · What kind of AI
Frontier models, expert prompts, structured outputs.
- Cross-validated frontier models. Every engine runs against multiple frontier LLMs and reconciles the answers, so a single model's blind spot does not become the score.
- Expert-crafted prompts. Over a hundred prompts authored from the way senior ASO consultants would actually phrase the question if they were sitting next to the data.
- Structured outputs. Each engine emits typed, schema-validated data — not free-form prose. That's what lets the dashboard cite the specific signal behind every recommendation.
- Multilingual extraction. Literal-keyword extraction works across Latin, CJK, Arabic, and Thai scripts; AI-inferred extraction fills the rest of the category corpus.
§8 · Verify the framing
See it side by side with ChatGPT and Claude.
The most direct way to verify the “AI is the runtime, not the headline” framing is to run the same input through Apptonomy and through the frontier LLMs your team is already using. Same app, same scope, three columns.
§9 · Verify on your own data
Too good to be true? Paste your URL.
Methodology pages are documentation. The honest test is running the methodology on your own app. Five minutes. Five recommendations. You decide.
Get a free quick audit for your app.
§10
We don't ask you to trust us.
The methodology is published. The reasoning sits next to every recommendation. The score moves on your own data after the change you ship. Run the loop and let the evidence answer the question.
Defensible ASO. Verifiable on your own data.
Apptonomy is senior human ASO expertise encoded as software. Read the methodology, run the loop, and let the score on your own listings tell you whether it works.
Per-engine methodology, in the open.
Each of the ten engines has a written methodology you can read before you trust the score. The reasoning that drives every recommendation traces back to a specific engine signal.
Plain-language reasoning, not opaque scores.
Every action names what to change, the signal that triggered it, and the projected lift on the ASO Readiness Score. The kind of reasoning a senior ASO consultant would give if you asked them why.
Recommendation Impact, on your own data.
Schedule a re-audit and watch your ASO Readiness Score and pillar scores move after the change you shipped. Audit history sits side by side so the lift is verifiable on the apps you actually run.
DIY frontier LLMs vs Apptonomy.
A clever prompt against a frontier model is not the same as the loop. Coverage, cross-validation, ranked actions, and continuous re-auditing are why the loop reads where DIY breaks down.