Methodology
1. Prompts
For each App Store category we write merchant-style questions with an AI model: plain “best app for…” questions, budget limits, store size, a must-have feature, an industry, and “instead of” questions. Prompts never name an app. Leaderboards use 15 prompts per category; instant checks use 10; tracked apps use up to 40 (Indie) or 100 (Studio), editable by the developer.
2. Assistants
- · ChatGPT: OpenAI GPT-6 Luna
- · Claude: Anthropic Claude Haiku 5.5
- · Gemini: Google Gemini 3.8 Flash
- · Perplexity: Perplexity Sonar (web search)
Each prompt is sent as a fresh conversation with no system instructions beyond “You are a helpful assistant”. Each assistant runs exactly the model listed; if it is unavailable, that part of a run fails and is shown as an error rather than silently using another model. API answers differ from the consumer ChatGPT app (memory, personalization) and Shopify Sidekick cannot be queried, so read results as a repeatable signal, not a screenshot of any one merchant’s chat.
3. Reading the answers
A separate model lists the Shopify apps each answer names, in order of first appearance. Names are matched to App Store listings through an alias table built from each listing’s name, its short name and its developer. Unmatched names still count as competitors under the name the assistant used.
4. Scores and ranges
AI share = answers that name the app ÷ all answers, across prompts, assistants and samples. Average position is the app’s mean rank among the apps named, when it is named. Leaderboards ask every prompt twice; tracked apps three times, and the range shown is the lowest and highest share across samples. Instant checks ask once, so the range shown is a 90% statistical (Wilson) interval.
5. Freshness
Category leaderboards refresh monthly. Listings are re-read at most every 3 days and category pages every 7 days. An instant check for an app is shared for 24 hours.
Request a correction
Think an answer was matched to the wrong app, or your listing is out of date? Tell us and we’ll check it, usually within 2 business days. Rankings reflect what assistants said, never our opinion of an app.
See it in action: run a free check.