The usual way to research mobile game UX is to install a dozen games, play each for twenty minutes and screenshot whatever you remember to. It is slow, you lose the order of events, and the evidence ends up as a camera roll nobody else can use. Here is a faster workflow built on recorded sessions.

It will not replace playing games. Feel, pacing and sound need hands on a device. But for the structural questions (what does this game teach first, when does it show a price, how does it bring players back) you do not need fifty installs.

Step 1: start with a question, not a game

"Look at Royal Match" is not a research question. "How do puzzle games introduce their first booster?" is. A good question names a system and a moment. Write it down before you open anything, because it decides what you ignore.

Step 2: pick the system

Game design already has a vocabulary for this, and a research library should use it. Gamedexy organizes its captures by game-design systems: onboarding, shops and offers, daily rewards, energy and lives, progression and upgrades, and more. Each system page collects real examples, so you start with a shortlist instead of a blank search box.

Step 3: search by what is on screen

When you know the words a screen would contain, search for them. The screen library searches on-screen text, and you can narrow by category, game or pattern tag. Try a phrase like "daily reward" or "battle pass", then filter to one category to cut the noise. Searching by text finds UI that no filename or store description would reveal.

Step 4: read flows in order

This is the step that single screenshots cannot do. A flow is a sequence of screens in the order a player saw them, so you can read what came before and after each moment. Order is the design: a shop after a win is a different message from the same shop after a loss.

Take onboarding. Colony Flow! puts a terms gate before the first tap and goes straight into a tap tutorial. Last Asylum opens with a playable story moment. Meowdoku teaches one rule per screen. Three games, three theories about what a new player needs first, visible in minutes.

Step 5: compare across games

One game is an anecdote. Three or four on the same question start to show a range. Line them up on the same dimension, such as how soon the first unlock arrives, how locked features are presented, or how the level map frames the long game.

For progression, Match Factory drips boosters out one per band of levels, while Magic Sort! shows locked features early as goals. Same problem, opposite pacing choices. Write down which you would try and why.

Step 6: record what the evidence cannot tell you

This habit separates research from collecting. For every conclusion, note the limit. A capture is one session on one device in one store region. It does not prove a design is fun, that it caused anyone's revenue, or that you saw every state the game can reach. Gamedexy's own research articles end with a limits section for exactly this reason, and yours should too.

Step 7: give the evidence to your AI agent

If you design or build with an AI coding agent, it can work from the same library instead of guessing what other games do. Gamedexy offers an MCP server that lets Claude Code or Codex search screens and read full flows while it helps you. It is included with Pro. You add the endpoint, sign in through the browser and ask in plain language, for example:

• Show me how mobile puzzle games introduce energy.
• Find first-session starter offers.
• Compare daily reward structures across games.

Answers come from recorded gameplay, and the same limits apply: the agent can show you what games do, not whether it works. Treat it like a competitor teardown you can query.

Three mistakes that waste research time

• Collecting screens with no question. A folder of 300 shop screenshots answers nothing until you know what you are comparing.
• Judging a game by one screen. A paywall or pop-up means little without the screens that led to it.
• Copying the surface. Borrow the structure and the reasoning, not the art or wording. Game content belongs to its publisher, and your players deserve your own design.

A one-hour version

• 10 minutes: write the question and pick the system.
• 20 minutes: skim the system page, search by on-screen text, shortlist four games.
• 20 minutes: read one flow per game, in order, noting the sequence and the dismiss path.
• 10 minutes: write three observations, one hypothesis and one limit.

Start researching

Gamedexy is a library of real screens and complete player flows from live mobile games, tagged by game-design system. Create a free account to browse previews, or browse player flows and search the screen library now. See pricing for what Pro opens up.